From 1f58f58cef21af545f17044038a9bd789ba2005e Mon Sep 17 00:00:00 2001
From: mickalchen <mickalchen@tencent.com>
Date: Wed, 26 Aug 2026 19:09:56 +0800
Subject: [PATCH 0001/2012] add Hy4 preview

---
 models/tencent/hy4-preview.toml                | 18 ++++++++++++++++++
 providers/tencent-token-plan/models/hy3.toml   |  2 +-
 .../tencent-token-plan/models/hy4-preview.toml | 18 ++++++++++++++++++
 providers/tencent-tokenhub/models/hy3.toml     |  2 +-
 .../tencent-tokenhub/models/hy4-preview.toml   | 18 ++++++++++++++++++
 5 files changed, 56 insertions(+), 2 deletions(-)
 create mode 100644 models/tencent/hy4-preview.toml
 create mode 100644 providers/tencent-token-plan/models/hy4-preview.toml
 create mode 100644 providers/tencent-tokenhub/models/hy4-preview.toml

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..19f5c9a36ec
--- /dev/null
+++ b/models/tencent/hy4-preview.toml
@@ -0,0 +1,18 @@
+name = "Hy4 preview"
+description = "A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities."
+family = "Hy"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_024_000
+output = 64_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/tencent-token-plan/models/hy3.toml b/providers/tencent-token-plan/models/hy3.toml
index f44d010abe8..312ca996a61 100644
--- a/providers/tencent-token-plan/models/hy3.toml
+++ b/providers/tencent-token-plan/models/hy3.toml
@@ -8,7 +8,7 @@ type = "toggle"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high"]
+values = ["none", "high"]
 
 [cost]
 input = 0
diff --git a/providers/tencent-token-plan/models/hy4-preview.toml b/providers/tencent-token-plan/models/hy4-preview.toml
new file mode 100644
index 00000000000..ca2ab4a6d71
--- /dev/null
+++ b/providers/tencent-token-plan/models/hy4-preview.toml
@@ -0,0 +1,18 @@
+# Per-token USD pricing is not published yet; zeros match other Token Plan entries.
+# Toggle: thinking.type = enabled|disabled (default disabled)
+# Effort: reasoning_effort = low|medium|high (default low), same as hy3 / hy3-preview
+# https://cloud.tencent.com/document/product/1823/131208
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
diff --git a/providers/tencent-tokenhub/models/hy3.toml b/providers/tencent-tokenhub/models/hy3.toml
index f44d010abe8..312ca996a61 100644
--- a/providers/tencent-tokenhub/models/hy3.toml
+++ b/providers/tencent-tokenhub/models/hy3.toml
@@ -8,7 +8,7 @@ type = "toggle"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high"]
+values = ["none", "high"]
 
 [cost]
 input = 0
diff --git a/providers/tencent-tokenhub/models/hy4-preview.toml b/providers/tencent-tokenhub/models/hy4-preview.toml
new file mode 100644
index 00000000000..e4c9ab43a21
--- /dev/null
+++ b/providers/tencent-tokenhub/models/hy4-preview.toml
@@ -0,0 +1,18 @@
+# Per-token USD pricing is not published yet; zeros match other TokenHub plan entries.
+# Toggle: thinking.type = enabled|disabled (default disabled)
+# Effort: reasoning_effort = low|medium|high (default low), same as hy3 / hy3-preview
+# https://cloud.tencent.com/document/product/1823/131208
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 9d1df0eddcd9249e7b8175666136e26d83236501 Mon Sep 17 00:00:00 2001
From: mickalchen <mickalchen@tencent.com>
Date: Wed, 26 Aug 2026 22:27:10 +0800
Subject: [PATCH 0002/2012] add Hy4 preview TokenHub and Token Plan pricing

Convert China-site list rates to USD per million tokens so the catalog no longer ships placeholder zeros.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 .../tencent-token-plan/models/hy4-preview.toml     | 14 ++++++++------
 providers/tencent-tokenhub/models/hy4-preview.toml | 14 ++++++++------
 2 files changed, 16 insertions(+), 12 deletions(-)

diff --git a/providers/tencent-token-plan/models/hy4-preview.toml b/providers/tencent-token-plan/models/hy4-preview.toml
index ca2ab4a6d71..2a75bc2426a 100644
--- a/providers/tencent-token-plan/models/hy4-preview.toml
+++ b/providers/tencent-token-plan/models/hy4-preview.toml
@@ -1,6 +1,9 @@
-# Per-token USD pricing is not published yet; zeros match other Token Plan entries.
+# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
+# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
+# Catalog is USD / million tokens (×1000).
 # Toggle: thinking.type = enabled|disabled (default disabled)
-# Effort: reasoning_effort = low|medium|high (default low), same as hy3 / hy3-preview
+# Effort: reasoning_effort = none|high
+# https://cloud.tencent.com/document/product/1823/130055
 # https://cloud.tencent.com/document/product/1823/131208
 base_model = "tencent/hy4-preview"
 
@@ -12,7 +15,6 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
+input = 0.834
+output = 2.501
+cache_read = 0.042
diff --git a/providers/tencent-tokenhub/models/hy4-preview.toml b/providers/tencent-tokenhub/models/hy4-preview.toml
index e4c9ab43a21..2a75bc2426a 100644
--- a/providers/tencent-tokenhub/models/hy4-preview.toml
+++ b/providers/tencent-tokenhub/models/hy4-preview.toml
@@ -1,6 +1,9 @@
-# Per-token USD pricing is not published yet; zeros match other TokenHub plan entries.
+# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
+# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
+# Catalog is USD / million tokens (×1000).
 # Toggle: thinking.type = enabled|disabled (default disabled)
-# Effort: reasoning_effort = low|medium|high (default low), same as hy3 / hy3-preview
+# Effort: reasoning_effort = none|high
+# https://cloud.tencent.com/document/product/1823/130055
 # https://cloud.tencent.com/document/product/1823/131208
 base_model = "tencent/hy4-preview"
 
@@ -12,7 +15,6 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
+input = 0.834
+output = 2.501
+cache_read = 0.042

From 121342f09705c2b1c55f14a9b3cdad4aa98a5912 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:56:42 +0000
Subject: [PATCH 0003/2012] chore(sync): update EmpirioLabs AI model catalog
 (#5570)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../empiriolabs/models/qwen3-8-flash.toml     | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/empiriolabs/models/qwen3-8-flash.toml

diff --git a/providers/empiriolabs/models/qwen3-8-flash.toml b/providers/empiriolabs/models/qwen3-8-flash.toml
new file mode 100644
index 00000000000..b0a68f51f8c
--- /dev/null
+++ b/providers/empiriolabs/models/qwen3-8-flash.toml
@@ -0,0 +1,19 @@
+base_model = "alibaba/qwen3.8-flash"
+temperature = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 262_144
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.16

From 2bd6c954f477bdacd3aad9c19772e0538a467795 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:56:54 +0000
Subject: [PATCH 0004/2012] chore(sync): update OpenRouter model catalog
 (#5572)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/openai/gpt-5.1.toml             | 1 +
 providers/openrouter/models/openai/gpt-5.2.toml             | 1 +
 providers/openrouter/models/openai/gpt-5.3-codex.toml       | 1 +
 providers/openrouter/models/openai/gpt-5.4-mini.toml        | 1 +
 providers/openrouter/models/openai/gpt-5.4-nano.toml        | 1 +
 providers/openrouter/models/openai/gpt-5.4.toml             | 1 +
 providers/openrouter/models/z-ai/glm-5.3-flash.toml         | 3 +--
 9 files changed, 13 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 38f2e9b4e85..b899aaefaad 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08106
-output = 0.16212
-cache_read = 0.016212
+input = 0.07952
+output = 0.15904
+cache_read = 0.015904
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 0442fcfba2d..309ffa254af 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 1.122
+output = 3.366
+cache_read = 0.0374
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-5.1.toml b/providers/openrouter/models/openai/gpt-5.1.toml
index 75e6b010bde..931ba99f4e0 100644
--- a/providers/openrouter/models/openai/gpt-5.1.toml
+++ b/providers/openrouter/models/openai/gpt-5.1.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.1"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/openai/gpt-5.2.toml b/providers/openrouter/models/openai/gpt-5.2.toml
index d8761780d97..95265894e8a 100644
--- a/providers/openrouter/models/openai/gpt-5.2.toml
+++ b/providers/openrouter/models/openai/gpt-5.2.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.2"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/openai/gpt-5.3-codex.toml b/providers/openrouter/models/openai/gpt-5.3-codex.toml
index 970482bc4de..f7d62896ab3 100644
--- a/providers/openrouter/models/openai/gpt-5.3-codex.toml
+++ b/providers/openrouter/models/openai/gpt-5.3-codex.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.3-codex"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/openai/gpt-5.4-mini.toml b/providers/openrouter/models/openai/gpt-5.4-mini.toml
index b71b67ad193..b3277f7996c 100644
--- a/providers/openrouter/models/openai/gpt-5.4-mini.toml
+++ b/providers/openrouter/models/openai/gpt-5.4-mini.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.4-mini"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/openai/gpt-5.4-nano.toml b/providers/openrouter/models/openai/gpt-5.4-nano.toml
index 2ca6f6a161e..957f3136cbe 100644
--- a/providers/openrouter/models/openai/gpt-5.4-nano.toml
+++ b/providers/openrouter/models/openai/gpt-5.4-nano.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.4-nano"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/openai/gpt-5.4.toml b/providers/openrouter/models/openai/gpt-5.4.toml
index 5ace7ba4f07..a5b83ea76eb 100644
--- a/providers/openrouter/models/openai/gpt-5.4.toml
+++ b/providers/openrouter/models/openai/gpt-5.4.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.4"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index 894e0ea18fb..d5918c075a2 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3-flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -12,7 +11,7 @@ output = 0.25
 cache_read = 0.015
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 
 [modalities]
 input = ["text", "image", "video"]

From 24677d2b3e54035e8165a8c80328d8ba491dd565 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:56:57 +0000
Subject: [PATCH 0005/2012] chore(sync): update NanoGPT model catalog (#5569)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-flash.toml         | 15 +------------
 providers/nano-gpt/models/claw-high.toml      | 21 +++++++++++--------
 providers/nano-gpt/models/claw-low.toml       | 17 ++++++++-------
 providers/nano-gpt/models/claw-medium.toml    | 15 +++++++------
 providers/nano-gpt/models/hermes-high.toml    | 21 +++++++++++--------
 providers/nano-gpt/models/hermes-low.toml     | 17 ++++++++-------
 providers/nano-gpt/models/hermes-medium.toml  | 15 +++++++------
 7 files changed, 63 insertions(+), 58 deletions(-)

diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml b/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
index 603cf086b94..6b08ac34660 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
+++ b/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
@@ -1,12 +1,4 @@
-name = "Qwen3.8 Flash"
-description = "Qwen3.8 Flash is Alibaba's latest fast multimodal model, with a million-token context window for coding, agentic workflows, visual understanding, long documents, codebases, and videos."
-release_date = "2026-08-26"
-last_updated = "2026-08-26"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
+base_model = "alibaba/qwen3.8-flash"
 
 [[reasoning_options]]
 type = "effort"
@@ -21,8 +13,3 @@ cache_write = 0.2
 [limit]
 context = 991_808
 input = 991_808
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/claw-high.toml b/providers/nano-gpt/models/claw-high.toml
index d8fe54b453b..7de86921a6d 100644
--- a/providers/nano-gpt/models/claw-high.toml
+++ b/providers/nano-gpt/models/claw-high.toml
@@ -3,23 +3,26 @@ name = "Claw High"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
 release_date = "2026-05-11"
 last_updated = "2026-05-11"
-attachment = true
+attachment = false
 reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 5
-output = 25
-cache_read = 2.5
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
-context = 1_000_000
-input = 1_000_000
-output = 128_000
+context = 1_048_576
+input = 1_048_576
+output = 131_072
 
 [modalities]
-input = ["text", "image", "pdf"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/claw-low.toml b/providers/nano-gpt/models/claw-low.toml
index 791e35d647d..1da6864013a 100644
--- a/providers/nano-gpt/models/claw-low.toml
+++ b/providers/nano-gpt/models/claw-low.toml
@@ -3,24 +3,27 @@ name = "Claw Low"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
 release_date = "2026-05-11"
 last_updated = "2026-05-11"
-attachment = true
+attachment = false
 reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.25
-output = 1.5
-cache_read = 0.025
+input = 1.4
+output = 4.4
+cache_read = 0.26
 cache_write = 0.08333
 
 [limit]
 context = 1_048_576
 input = 1_048_576
-output = 65_536
+output = 131_072
 
 [modalities]
-input = ["text", "image", "video", "pdf"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/claw-medium.toml b/providers/nano-gpt/models/claw-medium.toml
index 49f28819c78..da9a2c4bb73 100644
--- a/providers/nano-gpt/models/claw-medium.toml
+++ b/providers/nano-gpt/models/claw-medium.toml
@@ -8,16 +8,19 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.315
-output = 1.26
-cache_read = 0.1575
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
-context = 204_800
-input = 204_800
+context = 1_048_576
+input = 1_048_576
 output = 131_072
 
 [modalities]
diff --git a/providers/nano-gpt/models/hermes-high.toml b/providers/nano-gpt/models/hermes-high.toml
index 338d60661ce..0a1235c971b 100644
--- a/providers/nano-gpt/models/hermes-high.toml
+++ b/providers/nano-gpt/models/hermes-high.toml
@@ -4,23 +4,26 @@ description = "Compact GPT model for low-latency assistance and high-volume work
 family = "hermes"
 release_date = "2026-05-11"
 last_updated = "2026-05-11"
-attachment = true
+attachment = false
 reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 5
-output = 25
-cache_read = 2.5
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
-context = 1_000_000
-input = 1_000_000
-output = 128_000
+context = 1_048_576
+input = 1_048_576
+output = 131_072
 
 [modalities]
-input = ["text", "image", "pdf"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/hermes-low.toml b/providers/nano-gpt/models/hermes-low.toml
index a5e52e6189e..51d8dc85b5c 100644
--- a/providers/nano-gpt/models/hermes-low.toml
+++ b/providers/nano-gpt/models/hermes-low.toml
@@ -4,24 +4,27 @@ description = "Compact GPT model for low-latency assistance and high-volume work
 family = "hermes"
 release_date = "2026-05-11"
 last_updated = "2026-05-11"
-attachment = true
+attachment = false
 reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.25
-output = 1.5
-cache_read = 0.025
+input = 1.4
+output = 4.4
+cache_read = 0.26
 cache_write = 0.08333
 
 [limit]
 context = 1_048_576
 input = 1_048_576
-output = 65_536
+output = 131_072
 
 [modalities]
-input = ["text", "image", "video", "pdf"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/hermes-medium.toml b/providers/nano-gpt/models/hermes-medium.toml
index bbee1c0c489..687214f1c58 100644
--- a/providers/nano-gpt/models/hermes-medium.toml
+++ b/providers/nano-gpt/models/hermes-medium.toml
@@ -9,16 +9,19 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.315
-output = 1.26
-cache_read = 0.1575
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
-context = 204_800
-input = 204_800
+context = 1_048_576
+input = 1_048_576
 output = 131_072
 
 [modalities]

From df2675d7bd77bf9e6576d928b07d1edca275fe7d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:56:59 +0000
Subject: [PATCH 0006/2012] chore(sync): update Merge Gateway model catalog
 (#5566)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml     | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml | 7 ++++---
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..faa70b7c1e0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -11,8 +12,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index b4bdd294444..0a7a2254ee8 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 8ce0a3dc5e1259244f427478744b3cb57b894642 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:57:01 +0000
Subject: [PATCH 0007/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5565)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/gemini-3.5-transcribe.toml  | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/vercel/models/google/gemini-3.5-transcribe.toml

diff --git a/providers/vercel/models/google/gemini-3.5-transcribe.toml b/providers/vercel/models/google/gemini-3.5-transcribe.toml
new file mode 100644
index 00000000000..94262e42d64
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.5-transcribe.toml
@@ -0,0 +1,21 @@
+name = "Gemini 3.5 Transcribe"
+description = "Speech transcription model for accurate audio-to-text and captioning workflows"
+family = "gemini"
+release_date = "2026-08-26"
+last_updated = "2026-08-26"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 2
+output = 12
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["audio"]
+output = ["text"]

From 775d34f173da58f7bc1be657901586d46c092828 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:57:10 +0000
Subject: [PATCH 0008/2012] chore(sync): update Kilo model catalog (#5568)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-5.1.toml       | 1 +
 providers/kilo/models/openai/gpt-5.2.toml       | 1 +
 providers/kilo/models/openai/gpt-5.3-codex.toml | 1 +
 providers/kilo/models/openai/gpt-5.4-mini.toml  | 1 +
 providers/kilo/models/openai/gpt-5.4-nano.toml  | 1 +
 providers/kilo/models/openai/gpt-5.4.toml       | 1 +
 providers/kilo/models/xiaomi/mimo-v2.5-pro.toml | 6 +++---
 providers/kilo/models/z-ai/glm-5.3-flash.toml   | 1 -
 8 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-5.1.toml b/providers/kilo/models/openai/gpt-5.1.toml
index 5758647a980..4fb1c961279 100644
--- a/providers/kilo/models/openai/gpt-5.1.toml
+++ b/providers/kilo/models/openai/gpt-5.1.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.1"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/openai/gpt-5.2.toml b/providers/kilo/models/openai/gpt-5.2.toml
index 5561e674b1c..8c11e11820c 100644
--- a/providers/kilo/models/openai/gpt-5.2.toml
+++ b/providers/kilo/models/openai/gpt-5.2.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.2"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/openai/gpt-5.3-codex.toml b/providers/kilo/models/openai/gpt-5.3-codex.toml
index 93f23f1c621..bff5995bbd5 100644
--- a/providers/kilo/models/openai/gpt-5.3-codex.toml
+++ b/providers/kilo/models/openai/gpt-5.3-codex.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.3-codex"
 description = "Coding-optimized GPT model for repository edits, reviews, and agentic software work"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/openai/gpt-5.4-mini.toml b/providers/kilo/models/openai/gpt-5.4-mini.toml
index 23fef5244fc..d9e7888cb4d 100644
--- a/providers/kilo/models/openai/gpt-5.4-mini.toml
+++ b/providers/kilo/models/openai/gpt-5.4-mini.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.4-mini"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/openai/gpt-5.4-nano.toml b/providers/kilo/models/openai/gpt-5.4-nano.toml
index 5357765dea0..162cffcdc29 100644
--- a/providers/kilo/models/openai/gpt-5.4-nano.toml
+++ b/providers/kilo/models/openai/gpt-5.4-nano.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.4-nano"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/openai/gpt-5.4.toml b/providers/kilo/models/openai/gpt-5.4.toml
index 2ac608ee27a..0c60795a42c 100644
--- a/providers/kilo/models/openai/gpt-5.4.toml
+++ b/providers/kilo/models/openai/gpt-5.4.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.4"
 description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
+temperature = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/xiaomi/mimo-v2.5-pro.toml b/providers/kilo/models/xiaomi/mimo-v2.5-pro.toml
index 904393e08ad..378ff87497c 100644
--- a/providers/kilo/models/xiaomi/mimo-v2.5-pro.toml
+++ b/providers/kilo/models/xiaomi/mimo-v2.5-pro.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 1
-output = 3
-cache_read = 0.2
+input = 0.435
+output = 0.87
+cache_read = 0.004
 
 [[cost.tiers]]
 tier = { type = "context", size = 256_000 }
diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 8493dc1fa79..0ed9d5f31a9 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3-flash"
 description = "GLM-5.3-Flash is a native multimodal model from Z.ai. It is suited for efficient coding and long-horizon agent tasks. Its hybrid sparse and linear attention architecture maintains accurate long-context behavior while..."
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"

From 985e3b62fca5f08c28642cd0fd7722ee6eb1af67 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 18:57:22 +0000
Subject: [PATCH 0009/2012] chore(sync): update Charm Hyper model catalog
 (#5567)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.1.toml                |  6 +++---
 providers/hyper/models/glm-5.toml                  |  6 +++---
 providers/hyper/models/gpt-oss-120b.toml           |  6 +++---
 providers/hyper/models/kimi-k2.5.toml              |  6 +++---
 providers/hyper/models/llama-3.3-70b-instruct.toml |  6 +++---
 providers/hyper/models/minimax-m2.7.toml           |  6 +++---
 providers/hyper/models/qwen3.8-flash.toml          | 11 ++---------
 7 files changed, 20 insertions(+), 27 deletions(-)

diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 76d2acf71f4..4c86ffebd03 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.36
-output = 4.4
-cache_write = 0.68
+input = 1.29
+output = 4.22
+cache_write = 0.645
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 10f3e251574..5ce60bdae29 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.91
-output = 2.814
-cache_write = 0.455
+input = 0.92
+output = 2.976
+cache_write = 0.46
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index ad08deed2d7..b8abd9a4a66 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.188
-output = 0.7
-cache_write = 0.094
+input = 0.178
+output = 0.68
+cache_write = 0.089
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index debc53b807a..81e762294c4 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5444
-output = 2.855
-cache_write = 0.2722
+input = 0.544
+output = 2.76
+cache_write = 0.272
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/llama-3.3-70b-instruct.toml b/providers/hyper/models/llama-3.3-70b-instruct.toml
index de40b2bad9b..1620dd2fae7 100644
--- a/providers/hyper/models/llama-3.3-70b-instruct.toml
+++ b/providers/hyper/models/llama-3.3-70b-instruct.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.598
-output = 0.738
-cache_write = 0.299
+input = 0.638
+output = 0.768
+cache_write = 0.319
 
 [limit]
 output = 12_800
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index a0af9119e84..de482c56069 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.41
-output = 1.52
-cache_write = 0.205
+input = 0.424
+output = 1.612
+cache_write = 0.212
 
 [limit]
 context = 262_100
diff --git a/providers/hyper/models/qwen3.8-flash.toml b/providers/hyper/models/qwen3.8-flash.toml
index 7429fb9c848..abd863768ed 100644
--- a/providers/hyper/models/qwen3.8-flash.toml
+++ b/providers/hyper/models/qwen3.8-flash.toml
@@ -1,11 +1,6 @@
-name = "Qwen3.8-Flash"
+base_model = "alibaba/qwen3.8-flash"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
-release_date = "2026-08-26"
-last_updated = "2026-08-26"
-attachment = true
-reasoning = false
-tool_call = true
-open_weights = false
+reasoning_options = []
 
 [cost]
 input = 0.16
@@ -13,9 +8,7 @@ output = 0.47
 cache_read = 0.016
 
 [limit]
-context = 1_000_000
 output = 128_000
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]

From da2f2a8d84ffb735183f83471a1dab12f01f3fed Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:16:34 -0500
Subject: [PATCH 0010/2012] chore(sync): update Deep Infra model catalog
 (#5547)

* chore(sync): update Deep Infra model catalog

* fix(deepinfra): declare GLM-5.3-Flash reasoning effort

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../deepinfra/models/zai-org/GLM-5.3-Flash.toml     | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/deepinfra/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/deepinfra/models/zai-org/GLM-5.3-Flash.toml b/providers/deepinfra/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..60adafc59ec
--- /dev/null
+++ b/providers/deepinfra/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning_effort = low|high|max; thinking cannot be disabled.
+# https://docs.z.ai/guides/llm/glm-5.3-flash
+# https://docs.deepinfra.com/api-reference/chat-completions/openai-chat-completions
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_576

From b82943b8d9cfefd4d5c6dd5f14d778507a9873a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:37:02 -0500
Subject: [PATCH 0011/2012] fix(openrouter): sync optional reasoning toggles
 without effort lists (#5575)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/openrouter.ts     | 18 ++++++---
 packages/core/test/sync.test.ts               | 39 +++++++++++++++++++
 2 files changed, 51 insertions(+), 6 deletions(-)

diff --git a/packages/core/src/sync/providers/openrouter.ts b/packages/core/src/sync/providers/openrouter.ts
index c9ea734d4d1..58764e308a9 100644
--- a/packages/core/src/sync/providers/openrouter.ts
+++ b/packages/core/src/sync/providers/openrouter.ts
@@ -127,9 +127,13 @@ export const openrouter = {
       const authored = context.authored(model.id);
       return authored === undefined ? undefined : { id: model.id, model: authored as SyncedModel };
     }
+    const translated = buildOpenRouterModel(model, context.existing(model.id));
     return {
       id: model.id,
-      model: buildOpenRouterModel(model, context.existing(model.id)),
+      model: translated,
+      header: translated.reasoning_options?.some((option) => option.type === "toggle")
+        ? "# Toggle: reasoning.enabled = true|false\n# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens\n"
+        : undefined,
     };
   },
 } satisfies SyncProvider<OpenRouterModel>;
@@ -214,8 +218,9 @@ export function buildOpenRouterModel(
   // Prefer OpenRouter's live reasoning metadata over authored options so aliases
   // and rotated models pick up new efforts/budget support. Fall back to authored
   // only when the API omits a reasoning object.
-  const reasoning_options = openRouterReasoningOptions(model.reasoning)
-    ?? (reasoning ? existing?.reasoning_options : undefined);
+  const reasoning_options = reasoning
+    ? openRouterReasoningOptions(model.reasoning) ?? existing?.reasoning_options
+    : undefined;
   const context = model.context_length;
   const family = inferFamily(model, name);
   const releaseDate = dateFromTimestamp(model.created);
@@ -320,10 +325,11 @@ function openRouterReasoningOptions(reasoning: OpenRouterModel["reasoning"]): Sy
     ? ["max", "xhigh", "high", "medium", "low", "minimal", "none"] as const
     : reasoning.supported_efforts;
 
+  if (!reasoning.mandatory && !efforts?.includes("none")) {
+    options.push({ type: "toggle" });
+  }
+
   if (efforts !== undefined) {
-    if (!reasoning.mandatory && !efforts.includes("none")) {
-      options.push({ type: "toggle" });
-    }
     options.push({
       type: "effort",
       values: reasoning.mandatory ? efforts.filter((value) => value !== "none") : [...efforts],
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 54b77036934..0371d2fa8a9 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2790,6 +2790,45 @@ test("syncs OpenRouter reasoning efforts from model metadata", () => {
   });
 });
 
+test("syncs OpenRouter toggles without an effort selector", () => {
+  for (const supports_max_tokens of [undefined, true]) {
+    const source = openRouterModel({
+      reasoning: { mandatory: false, supports_max_tokens },
+    });
+    const translated = openrouter.translateModel(source, {
+      existing: () => undefined,
+      authored: () => undefined,
+    });
+    expect(translated?.model.reasoning_options).toEqual([
+      { type: "toggle" },
+      ...(supports_max_tokens ? [{ type: "budget_tokens" }] : []),
+    ]);
+    expect(translated?.header).toStartWith("# Toggle: reasoning.enabled = true|false\n");
+  }
+});
+
+test("does not derive OpenRouter controls for non-reasoning models", () => {
+  const model = buildOpenRouterModel(openRouterModel({
+    supported_parameters: ["temperature"],
+    reasoning: { mandatory: false, supports_max_tokens: true },
+  }), { reasoning_options: [{ type: "toggle" }] });
+  expect(model.reasoning).toBe(false);
+  expect(model.reasoning_options).toBeUndefined();
+});
+
+test("does not add OpenRouter toggles to mandatory or effort-none models", () => {
+  for (const reasoning of [
+    { mandatory: true, supports_max_tokens: true },
+    { mandatory: true, supported_efforts: ["none", "high"] as const },
+    { mandatory: false, supported_efforts: ["none", "high"] as const },
+    { mandatory: false, supported_efforts: null },
+  ]) {
+    const [source] = openrouter.parseModels({ data: [{ ...openRouterModel(), reasoning }] });
+    const model = buildOpenRouterModel(source!, undefined);
+    expect(model.reasoning_options?.some((option) => option.type === "toggle")).toBe(false);
+  }
+});
+
 test("uses OpenRouter model context when top provider reports a shorter context", () => {
   const model = buildOpenRouterModel(openRouterModel({
     context_length: 1_048_576,

From aeabb86bd5ece15e66fab3faf59429aebd4d12d8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:37:21 -0500
Subject: [PATCH 0012/2012] fix(merge-gateway): sync explicitly advertised
 thinking budgets (#5574)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/merge-gateway.ts  | 13 ++++-
 packages/core/test/sync.test.ts               | 50 +++++++++++++++++++
 2 files changed, 62 insertions(+), 1 deletion(-)

diff --git a/packages/core/src/sync/providers/merge-gateway.ts b/packages/core/src/sync/providers/merge-gateway.ts
index 21fbfe5fcfd..1a589afaecb 100644
--- a/packages/core/src/sync/providers/merge-gateway.ts
+++ b/packages/core/src/sync/providers/merge-gateway.ts
@@ -158,7 +158,14 @@ export const mergeGateway = {
   translateModel(model, context) {
     const existing = context.existing(model.model);
     const translated = buildMergeGatewayModel(model, existing, context.authored(model.model));
-    return translated === undefined ? undefined : { id: model.model, model: translated };
+    return translated === undefined ? undefined : {
+      id: model.model,
+      model: translated,
+      header: translated.reasoning_options?.some((option) => option.type === "toggle")
+        && translated.reasoning_options.some((option) => option.type === "budget_tokens")
+        ? '# Toggle: thinking.type = "enabled"|"disabled"; enabled requires thinking.budget_tokens.\n# https://docs.merge.dev/merge-gateway/features/reasoning\n'
+        : undefined,
+    };
   },
 } satisfies SyncProvider<MergeGatewayModel>;
 
@@ -181,6 +188,10 @@ export function mergeGatewayReasoningOptions(
     options.push({ type: "effort" as const, values: [...effortValues] });
   }
 
+  if (controls.includes("thinking.budget_tokens")) {
+    options.push({ type: "budget_tokens" });
+  }
+
   return options;
 }
 
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 0371d2fa8a9..1419f568874 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -3717,6 +3717,56 @@ test("derives a Merge Gateway reasoning toggle when the selected route supports
   expect(model).toMatchObject({ reasoning_options: [{ type: "toggle" }] });
 });
 
+test("syncs Merge Gateway explicitly advertised thinking budgets", () => {
+  const selected = mergeGatewayVendor();
+  selected.capabilities.supports_reasoning = true;
+  selected.capabilities.reasoning = {
+    configurable: true,
+    disable_supported: true,
+    default_enabled: false,
+    controls: ["thinking.budget_tokens"],
+    output_style: "reasoning_content",
+  };
+  const source = mergeGatewayModel({ vendors: { openai: selected } });
+  const translated = mergeGateway.translateModel(source, {
+    existing: () => ({ reasoning: true, reasoning_options: [] }),
+    authored: () => undefined,
+  });
+  expect(translated?.model.reasoning_options).toEqual([
+    { type: "toggle" },
+    { type: "budget_tokens" },
+  ]);
+  expect(translated?.header).toStartWith('# Toggle: thinking.type = "enabled"|"disabled"');
+
+  selected.capabilities.reasoning.disable_supported = false;
+  expect(buildMergeGatewayModel(source, { reasoning: true })?.reasoning_options).toEqual([
+    { type: "budget_tokens" },
+  ]);
+});
+
+test("does not infer Merge Gateway budgets from other controls or output limits", () => {
+  for (const controls of [undefined, [], ["thinking"], ["max_tokens"], ["reasoning.effort"]]) {
+    const selected = mergeGatewayVendor();
+    selected.capabilities.reasoning = { configurable: true, controls };
+    const model = buildMergeGatewayModel(mergeGatewayModel({ vendors: { openai: selected } }), {
+      reasoning: true,
+      reasoning_options: [],
+    });
+    expect(model?.reasoning_options).toEqual([]);
+  }
+});
+
+test("preserves curated Merge Gateway controls when a budget is advertised", () => {
+  const selected = mergeGatewayVendor();
+  selected.capabilities.reasoning = { controls: ["thinking.budget_tokens"] };
+  const reasoning_options = [{ type: "effort" as const, values: ["high"] }];
+  const model = buildMergeGatewayModel(mergeGatewayModel({ vendors: { openai: selected } }), {
+    reasoning: true,
+    reasoning_options,
+  });
+  expect(model?.reasoning_options).toEqual(reasoning_options);
+});
+
 // Effort control yields toggle + effort, not a bare toggle (claude-opus-5 regression).
 test("derives Merge Gateway toggle + effort from an effort control", () => {
   const selected = mergeGatewayVendor({

From 5437a9ff7ae7f99cc4dea173c57c4b622c95bc90 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:37:35 -0500
Subject: [PATCH 0013/2012] fix(crossmodel): omit redundant reasoning toggles
 (#5577)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/crossmodel.ts     |  3 +++
 packages/core/test/sync.test.ts                    | 14 ++++++++++----
 .../models/gemini/gemini-2.5-flash-lite.toml       |  3 ---
 .../crossmodel/models/gemini/gemini-2.5-flash.toml |  3 ---
 providers/crossmodel/models/tencent/hy3.toml       |  3 ---
 providers/crossmodel/models/x-ai/grok-4.3.toml     |  3 ---
 .../crossmodel/models/xiaomi/mimo-v2.5-pro.toml    |  3 ---
 providers/crossmodel/models/xiaomi/mimo-v2.5.toml  |  3 ---
 8 files changed, 13 insertions(+), 22 deletions(-)

diff --git a/packages/core/src/sync/providers/crossmodel.ts b/packages/core/src/sync/providers/crossmodel.ts
index 576cca122ea..e405a811a8f 100644
--- a/packages/core/src/sync/providers/crossmodel.ts
+++ b/packages/core/src/sync/providers/crossmodel.ts
@@ -180,6 +180,9 @@ function reasoningOptions(model: CrossModelModel): SyncedModel["reasoning_option
     if (reasoning.budget_tokens.max !== undefined) budget.max = reasoning.budget_tokens.max;
     options.push(budget);
   }
+  if (options.some((option) => option.type === "effort" && option.values.includes("none"))) {
+    return options.filter((option) => option.type !== "toggle");
+  }
   return options;
 }
 
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 1419f568874..6eca95b68a8 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -474,7 +474,13 @@ test("parses CrossModel's nullable reasoning controls", () => {
   });
 });
 
-test("syncs CrossModel's explicit reasoning controls", () => {
+test("preserves CrossModel's toggle-only reasoning control", () => {
+  const model = buildCrossModel(crossModelModel(), undefined);
+  expect(model?.reasoning_options).toEqual([{ type: "toggle" }]);
+});
+
+test.each([{ off: false }, { off: true }])("syncs CrossModel's reasoning controls (effort includes none: $off)", ({ off }) => {
+  const effort = off ? ["none", "low", "high", "max"] as const : ["low", "high", "max"] as const;
   const model = buildCrossModel(
     crossModelModel({
       capabilities: {
@@ -482,7 +488,7 @@ test("syncs CrossModel's explicit reasoning controls", () => {
         reasoning: {
           supported: true,
           toggle: true,
-          effort: ["low", "high", "max"],
+          effort: [...effort],
           budget_tokens: { min: 1_024, max: 32_000 },
         },
       },
@@ -492,8 +498,8 @@ test("syncs CrossModel's explicit reasoning controls", () => {
 
   expect(model).toMatchObject({
     reasoning_options: [
-      { type: "toggle" },
-      { type: "effort", values: ["low", "high", "max"] },
+      ...off ? [] : [{ type: "toggle" }],
+      { type: "effort", values: effort },
       { type: "budget_tokens", min: 1_024, max: 32_000 },
     ],
   });
diff --git a/providers/crossmodel/models/gemini/gemini-2.5-flash-lite.toml b/providers/crossmodel/models/gemini/gemini-2.5-flash-lite.toml
index 104e3400b16..4352596bb4e 100644
--- a/providers/crossmodel/models/gemini/gemini-2.5-flash-lite.toml
+++ b/providers/crossmodel/models/gemini/gemini-2.5-flash-lite.toml
@@ -1,8 +1,5 @@
 base_model = "google/gemini-2.5-flash-lite"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
diff --git a/providers/crossmodel/models/gemini/gemini-2.5-flash.toml b/providers/crossmodel/models/gemini/gemini-2.5-flash.toml
index 95d431f646c..bfcddbf5a98 100644
--- a/providers/crossmodel/models/gemini/gemini-2.5-flash.toml
+++ b/providers/crossmodel/models/gemini/gemini-2.5-flash.toml
@@ -1,8 +1,5 @@
 base_model = "google/gemini-2.5-flash"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
diff --git a/providers/crossmodel/models/tencent/hy3.toml b/providers/crossmodel/models/tencent/hy3.toml
index 7154c364956..0aca4d6ee70 100644
--- a/providers/crossmodel/models/tencent/hy3.toml
+++ b/providers/crossmodel/models/tencent/hy3.toml
@@ -1,9 +1,6 @@
 base_model = "tencent/hy3"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "high"]
diff --git a/providers/crossmodel/models/x-ai/grok-4.3.toml b/providers/crossmodel/models/x-ai/grok-4.3.toml
index 751982e3814..3a82998f376 100644
--- a/providers/crossmodel/models/x-ai/grok-4.3.toml
+++ b/providers/crossmodel/models/x-ai/grok-4.3.toml
@@ -1,8 +1,5 @@
 base_model = "xai/grok-4.3"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high"]
diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.5-pro.toml b/providers/crossmodel/models/xiaomi/mimo-v2.5-pro.toml
index f969d5cc0f1..3d46f73b352 100644
--- a/providers/crossmodel/models/xiaomi/mimo-v2.5-pro.toml
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.5-pro.toml
@@ -1,9 +1,6 @@
 base_model = "xiaomi/mimo-v2.5-pro"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high"]
diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.5.toml b/providers/crossmodel/models/xiaomi/mimo-v2.5.toml
index f609d4e60ba..2c31756f08a 100644
--- a/providers/crossmodel/models/xiaomi/mimo-v2.5.toml
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.5.toml
@@ -1,9 +1,6 @@
 base_model = "xiaomi/mimo-v2.5"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high"]

From 2c382ac47fbfce81ea636eab82c81d6e3373893a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:37:53 -0500
Subject: [PATCH 0014/2012] fix(empiriolabs): omit redundant reasoning toggles
 (#5576)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/empiriolabs.ts    |  3 ++
 packages/core/test/empiriolabs.test.ts        | 47 ++++++++++++++++++-
 packages/core/test/sync.test.ts               |  1 -
 .../empiriolabs/models/deepseek-v3-2.toml     |  3 --
 .../models/deepseek-v4-flash-0731.toml        |  3 --
 .../empiriolabs/models/deepseek-v4-flash.toml |  3 --
 .../models/deepseek-v4-pro-0813.toml          |  3 --
 .../empiriolabs/models/deepseek-v4-pro.toml   |  3 --
 .../empiriolabs/models/gemma-4-26b-a4b.toml   |  3 --
 providers/empiriolabs/models/glm-5-1.toml     |  3 --
 providers/empiriolabs/models/glm-5-2.toml     |  3 --
 providers/empiriolabs/models/kimi-k2-6.toml   |  3 --
 .../empiriolabs/models/qwen3-5-122b-a10b.toml |  3 --
 providers/empiriolabs/models/qwen3-5-27b.toml |  3 --
 .../empiriolabs/models/qwen3-5-35b-a3b.toml   |  3 --
 .../empiriolabs/models/qwen3-5-397b-a17b.toml |  3 --
 providers/empiriolabs/models/qwen3-5-4b.toml  |  3 --
 providers/empiriolabs/models/qwen3-5-9b.toml  |  3 --
 providers/empiriolabs/models/qwen3-6-27b.toml |  3 --
 .../empiriolabs/models/qwen3-6-35b-a3b.toml   |  3 --
 .../empiriolabs/models/qwen3-6-flash.toml     |  3 --
 .../models/qwen3-6-max-preview.toml           |  3 --
 .../empiriolabs/models/qwen3-7-flash.toml     |  3 --
 providers/empiriolabs/models/qwen3-7-max.toml |  3 --
 .../empiriolabs/models/qwen3-7-plus.toml      |  3 --
 providers/empiriolabs/models/qwen3-8-27b.toml |  3 --
 .../empiriolabs/models/qwen3-8-flash.toml     |  3 --
 providers/empiriolabs/models/qwen3-8-max.toml |  3 --
 28 files changed, 49 insertions(+), 77 deletions(-)

diff --git a/packages/core/src/sync/providers/empiriolabs.ts b/packages/core/src/sync/providers/empiriolabs.ts
index 992d1edd90c..3667cff64e6 100644
--- a/packages/core/src/sync/providers/empiriolabs.ts
+++ b/packages/core/src/sync/providers/empiriolabs.ts
@@ -191,6 +191,9 @@ function reasoningOptions(model: EmpiriolabsModel): SyncedModel["reasoning_optio
     if (budget.max !== undefined) option.max = budget.max;
     options.push(option);
   }
+  if (options.some((option) => option.type === "effort" && option.values.includes("none"))) {
+    return options.filter((option) => option.type !== "toggle");
+  }
   return options;
 }
 
diff --git a/packages/core/test/empiriolabs.test.ts b/packages/core/test/empiriolabs.test.ts
index e2c6aefa0b0..2bb17ad4130 100644
--- a/packages/core/test/empiriolabs.test.ts
+++ b/packages/core/test/empiriolabs.test.ts
@@ -1,6 +1,51 @@
 import { expect, test } from "bun:test";
 
-import { resolveEmpiriolabsBaseModel } from "../src/sync/providers/empiriolabs.js";
+import { buildEmpiriolabsModel, resolveEmpiriolabsBaseModel } from "../src/sync/providers/empiriolabs.js";
+
+test.each([
+  {
+    name: "toggle-only",
+    parameters: [{ name: "enable_thinking" }],
+    expected: [{ type: "toggle" }],
+  },
+  {
+    name: "toggle and effort without none",
+    parameters: [
+      { name: "enable_thinking" },
+      { name: "reasoning_effort", options: ["low", "high"] },
+    ],
+    expected: [{ type: "toggle" }, { type: "effort", values: ["low", "high"] }],
+  },
+  {
+    name: "effort with none instead of a redundant toggle",
+    parameters: [
+      { name: "enable_thinking" },
+      { name: "reasoning_effort", options: ["none", "low", "high"] },
+    ],
+    expected: [{ type: "effort", values: ["none", "low", "high"] }],
+  },
+  {
+    name: "effort with none preserves the reasoning budget",
+    parameters: [
+      { name: "enable_thinking" },
+      { name: "reasoning_effort", options: ["none", "low", "high"] },
+      { name: "thinking_budget", min: 1_024, max: 32_768 },
+    ],
+    expected: [
+      { type: "effort", values: ["none", "low", "high"] },
+      { type: "budget_tokens", min: 1_024, max: 32_768 },
+    ],
+  },
+])("syncs EmpirioLabs reasoning controls: $name", ({ parameters, expected }) => {
+  const model = buildEmpiriolabsModel({
+    id: "qwen3-5-9b",
+    context_length: 262_144,
+    capabilities: { reasoning: true },
+    supported_parameters: parameters,
+  }, undefined);
+
+  expect(model?.reasoning_options).toEqual(expected);
+});
 
 test("resolves existing lab metadata without a hardcoded map", () => {
   expect(resolveEmpiriolabsBaseModel("muse-glimmer-30b")).toBe("meta/muse-glimmer-30b");
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 6eca95b68a8..d74711fb722 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -4363,7 +4363,6 @@ test("syncs EmpirioLabs pricing tiers and reasoning controls", () => {
     base_model: "minimax/MiniMax-M3",
     structured_output: true,
     reasoning_options: [
-      { type: "toggle" },
       { type: "effort", values: ["none", "low", "medium", "high", "max"] },
       { type: "budget_tokens", min: 1_024, max: 32_768 },
     ],
diff --git a/providers/empiriolabs/models/deepseek-v3-2.toml b/providers/empiriolabs/models/deepseek-v3-2.toml
index 39659c40e1e..02394655514 100644
--- a/providers/empiriolabs/models/deepseek-v3-2.toml
+++ b/providers/empiriolabs/models/deepseek-v3-2.toml
@@ -1,8 +1,5 @@
 base_model = "deepseek/deepseek-v3.2"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/deepseek-v4-flash-0731.toml b/providers/empiriolabs/models/deepseek-v4-flash-0731.toml
index a1ea834cde8..f5aabea1560 100644
--- a/providers/empiriolabs/models/deepseek-v4-flash-0731.toml
+++ b/providers/empiriolabs/models/deepseek-v4-flash-0731.toml
@@ -1,8 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/deepseek-v4-flash.toml b/providers/empiriolabs/models/deepseek-v4-flash.toml
index 7bc75ba4c5c..16373cc6247 100644
--- a/providers/empiriolabs/models/deepseek-v4-flash.toml
+++ b/providers/empiriolabs/models/deepseek-v4-flash.toml
@@ -1,9 +1,6 @@
 base_model = "deepseek/deepseek-v4-flash"
 last_updated = "2026-06-12"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/deepseek-v4-pro-0813.toml b/providers/empiriolabs/models/deepseek-v4-pro-0813.toml
index 5872f8a2f05..5c90d606912 100644
--- a/providers/empiriolabs/models/deepseek-v4-pro-0813.toml
+++ b/providers/empiriolabs/models/deepseek-v4-pro-0813.toml
@@ -1,8 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/deepseek-v4-pro.toml b/providers/empiriolabs/models/deepseek-v4-pro.toml
index 2c83c4cb98e..79372250253 100644
--- a/providers/empiriolabs/models/deepseek-v4-pro.toml
+++ b/providers/empiriolabs/models/deepseek-v4-pro.toml
@@ -1,9 +1,6 @@
 base_model = "deepseek/deepseek-v4-pro"
 last_updated = "2026-06-12"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/gemma-4-26b-a4b.toml b/providers/empiriolabs/models/gemma-4-26b-a4b.toml
index 785a46f03ee..bb6dbc1a17e 100644
--- a/providers/empiriolabs/models/gemma-4-26b-a4b.toml
+++ b/providers/empiriolabs/models/gemma-4-26b-a4b.toml
@@ -2,9 +2,6 @@ base_model = "google/gemma-4-26b-a4b-it"
 name = "Gemma 4 26B-A4B"
 last_updated = "2026-06-12"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/glm-5-1.toml b/providers/empiriolabs/models/glm-5-1.toml
index 87b108a67ea..0224b7de115 100644
--- a/providers/empiriolabs/models/glm-5-1.toml
+++ b/providers/empiriolabs/models/glm-5-1.toml
@@ -2,9 +2,6 @@ base_model = "zhipuai/glm-5.1"
 name = "GLM 5.1"
 last_updated = "2026-06-12"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/glm-5-2.toml b/providers/empiriolabs/models/glm-5-2.toml
index cebceb91bb3..d0d3a97a50e 100644
--- a/providers/empiriolabs/models/glm-5-2.toml
+++ b/providers/empiriolabs/models/glm-5-2.toml
@@ -1,9 +1,6 @@
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
diff --git a/providers/empiriolabs/models/kimi-k2-6.toml b/providers/empiriolabs/models/kimi-k2-6.toml
index 5e61c74fd04..d1ee9f7a896 100644
--- a/providers/empiriolabs/models/kimi-k2-6.toml
+++ b/providers/empiriolabs/models/kimi-k2-6.toml
@@ -1,9 +1,6 @@
 base_model = "moonshotai/kimi-k2.6"
 last_updated = "2026-06-12"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-122b-a10b.toml b/providers/empiriolabs/models/qwen3-5-122b-a10b.toml
index 4e5a031a156..051fb2ed4fc 100644
--- a/providers/empiriolabs/models/qwen3-5-122b-a10b.toml
+++ b/providers/empiriolabs/models/qwen3-5-122b-a10b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.5-122b-a10b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-27b.toml b/providers/empiriolabs/models/qwen3-5-27b.toml
index 8ec24690b05..65e6d1423b0 100644
--- a/providers/empiriolabs/models/qwen3-5-27b.toml
+++ b/providers/empiriolabs/models/qwen3-5-27b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.5-27b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-35b-a3b.toml b/providers/empiriolabs/models/qwen3-5-35b-a3b.toml
index 79c7b132465..061c41b87d9 100644
--- a/providers/empiriolabs/models/qwen3-5-35b-a3b.toml
+++ b/providers/empiriolabs/models/qwen3-5-35b-a3b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.5-35b-a3b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-397b-a17b.toml b/providers/empiriolabs/models/qwen3-5-397b-a17b.toml
index 71c88ef9155..0063bf09354 100644
--- a/providers/empiriolabs/models/qwen3-5-397b-a17b.toml
+++ b/providers/empiriolabs/models/qwen3-5-397b-a17b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.5-397b-a17b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-4b.toml b/providers/empiriolabs/models/qwen3-5-4b.toml
index f7a003f1cef..af27e389880 100644
--- a/providers/empiriolabs/models/qwen3-5-4b.toml
+++ b/providers/empiriolabs/models/qwen3-5-4b.toml
@@ -9,9 +9,6 @@ tool_call = true
 structured_output = true
 open_weights = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-5-9b.toml b/providers/empiriolabs/models/qwen3-5-9b.toml
index f9c215d402c..23363844f21 100644
--- a/providers/empiriolabs/models/qwen3-5-9b.toml
+++ b/providers/empiriolabs/models/qwen3-5-9b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.5-9b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-6-27b.toml b/providers/empiriolabs/models/qwen3-6-27b.toml
index 67e1fc45fb6..80c72e5ecd1 100644
--- a/providers/empiriolabs/models/qwen3-6-27b.toml
+++ b/providers/empiriolabs/models/qwen3-6-27b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.6-27b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-6-35b-a3b.toml b/providers/empiriolabs/models/qwen3-6-35b-a3b.toml
index d7c0dce5058..5b6add630f9 100644
--- a/providers/empiriolabs/models/qwen3-6-35b-a3b.toml
+++ b/providers/empiriolabs/models/qwen3-6-35b-a3b.toml
@@ -2,9 +2,6 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B A3B"
 attachment = false
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-6-flash.toml b/providers/empiriolabs/models/qwen3-6-flash.toml
index 9ac389cedaf..0b1cbc497d6 100644
--- a/providers/empiriolabs/models/qwen3-6-flash.toml
+++ b/providers/empiriolabs/models/qwen3-6-flash.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.6-flash"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-6-max-preview.toml b/providers/empiriolabs/models/qwen3-6-max-preview.toml
index dc02cfe19ba..be4ad9c722e 100644
--- a/providers/empiriolabs/models/qwen3-6-max-preview.toml
+++ b/providers/empiriolabs/models/qwen3-6-max-preview.toml
@@ -1,9 +1,6 @@
 base_model = "alibaba/qwen3.6-max-preview"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-7-flash.toml b/providers/empiriolabs/models/qwen3-7-flash.toml
index 2f3d3d5a4d5..26b58b3e7b3 100644
--- a/providers/empiriolabs/models/qwen3-7-flash.toml
+++ b/providers/empiriolabs/models/qwen3-7-flash.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.7-flash"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-7-max.toml b/providers/empiriolabs/models/qwen3-7-max.toml
index b651d772d2f..52fda1a63a5 100644
--- a/providers/empiriolabs/models/qwen3-7-max.toml
+++ b/providers/empiriolabs/models/qwen3-7-max.toml
@@ -2,9 +2,6 @@ base_model = "alibaba/qwen3.7-max"
 last_updated = "2026-06-12"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-7-plus.toml b/providers/empiriolabs/models/qwen3-7-plus.toml
index e9d36d2cfea..6ee8b3e76c8 100644
--- a/providers/empiriolabs/models/qwen3-7-plus.toml
+++ b/providers/empiriolabs/models/qwen3-7-plus.toml
@@ -2,9 +2,6 @@ base_model = "alibaba/qwen3.7-plus"
 last_updated = "2026-06-12"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-8-27b.toml b/providers/empiriolabs/models/qwen3-8-27b.toml
index d0f170be85f..c5d6f2e57ac 100644
--- a/providers/empiriolabs/models/qwen3-8-27b.toml
+++ b/providers/empiriolabs/models/qwen3-8-27b.toml
@@ -1,8 +1,5 @@
 base_model = "alibaba/qwen3.8-27b"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "xhigh"]
diff --git a/providers/empiriolabs/models/qwen3-8-flash.toml b/providers/empiriolabs/models/qwen3-8-flash.toml
index b0a68f51f8c..de93d1d0e24 100644
--- a/providers/empiriolabs/models/qwen3-8-flash.toml
+++ b/providers/empiriolabs/models/qwen3-8-flash.toml
@@ -1,9 +1,6 @@
 base_model = "alibaba/qwen3.8-flash"
 temperature = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]
diff --git a/providers/empiriolabs/models/qwen3-8-max.toml b/providers/empiriolabs/models/qwen3-8-max.toml
index f9b405845ed..5d907ad22d4 100644
--- a/providers/empiriolabs/models/qwen3-8-max.toml
+++ b/providers/empiriolabs/models/qwen3-8-max.toml
@@ -1,9 +1,6 @@
 base_model = "alibaba/qwen3.8-max"
 structured_output = true
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "high", "max"]

From e54fb3b6c572db7873694a031431fae3e84d00bd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 14:38:11 -0500
Subject: [PATCH 0015/2012] fix(tinfoil): sync reasoning capability and correct
 effort levels (#5573)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/tinfoil.ts   |  5 ++++
 packages/core/test/sync.test.ts               | 25 +++++++++++++++++++
 .../tinfoil/models/deepseek-v4-flash.toml     | 18 ++++++-------
 providers/tinfoil/models/kimi-k3.toml         | 10 ++++----
 sync.md                                       |  3 ++-
 5 files changed, 44 insertions(+), 17 deletions(-)

diff --git a/packages/core/src/sync/providers/tinfoil.ts b/packages/core/src/sync/providers/tinfoil.ts
index 90dcd6fd9c3..9e49d184f7c 100644
--- a/packages/core/src/sync/providers/tinfoil.ts
+++ b/packages/core/src/sync/providers/tinfoil.ts
@@ -88,6 +88,9 @@ export function buildTinfoilModel(
   if (existing.cost === undefined || existing.limit?.context === undefined) {
     throw new Error(`Tinfoil model ${model.id} has incomplete local pricing or limits required for sync`);
   }
+  if (model.reasoning && existing.reasoning_options === undefined) {
+    throw new Error(`Tinfoil model ${model.id} requires hand-authored reasoning_options; the catalog exposes no reasoning controls`);
+  }
 
   const { base_model: baseModel, base_model_omit: baseModelOmit, ...current } = existing;
   const cost = {
@@ -102,6 +105,8 @@ export function buildTinfoilModel(
   };
   const values = {
     ...current,
+    reasoning: model.reasoning,
+    reasoning_options: model.reasoning ? existing.reasoning_options : undefined,
     cost,
     limit,
   } as SyncedFullModel;
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index d74711fb722..9266e2b120e 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -1230,6 +1230,31 @@ test("syncs Tinfoil cached-input pricing from the public model catalog", () => {
   });
 });
 
+test.each([undefined, "zhipuai/glm-5.2"])("syncs Tinfoil reasoning with base model %s", (base_model) => {
+  const existing = { ...existingTinfoilGLM, base_model };
+  const enabled = buildTinfoilModel(tinfoilModel(), { ...existing, reasoning: false });
+  expect(enabled.reasoning_options).toEqual(existing.reasoning_options);
+  // Factored models inherit true from the lab; standalone models must author it.
+  expect(enabled.reasoning).toBe(base_model === undefined ? true : undefined);
+
+  const disabled = buildTinfoilModel(tinfoilModel({ reasoning: false }), existing);
+  expect(disabled.reasoning).toBe(false);
+  expect(disabled.reasoning_options).toBeUndefined();
+});
+
+test("requires authored Tinfoil controls instead of inventing an empty set", () => {
+  expect(() => buildTinfoilModel(tinfoilModel(), {
+    ...existingTinfoilGLM,
+    reasoning_options: undefined,
+  })).toThrow("requires hand-authored reasoning_options");
+
+  const model = buildTinfoilModel(tinfoilModel(), {
+    ...existingTinfoilGLM,
+    reasoning_options: [],
+  });
+  expect(model.reasoning_options).toEqual([]);
+});
+
 test("removes stale Tinfoil cache pricing when the public catalog omits it", () => {
   const model = buildTinfoilModel(tinfoilModel({
     pricing: {
diff --git a/providers/tinfoil/models/deepseek-v4-flash.toml b/providers/tinfoil/models/deepseek-v4-flash.toml
index e7a7d694206..b2b050f3391 100644
--- a/providers/tinfoil/models/deepseek-v4-flash.toml
+++ b/providers/tinfoil/models/deepseek-v4-flash.toml
@@ -1,18 +1,14 @@
-# Sources (accessed 2026-08-12):
-# - https://inference.tinfoil.sh/v1/models (id deepseek-v4-flash; context 1000000;
-#   input $0.7/MTok; output $1.9/MTok; cachedInput $0.125/MTok; reasoning true)
-# - https://docs.tinfoil.sh/models/chat.md (weights deepseek-ai/DeepSeek-V4-Flash-0731; 1M context)
-# - https://docs.tinfoil.sh/guides/reasoning.md (OpenAI-compatible reasoning_effort; response message.reasoning)
-# - https://docs.tinfoil.sh/resources/changelog.md (August 2026: added deepseek-v4-flash)
-# Effort: lab DeepSeek V4 Flash OpenAI path uses reasoning_effort low|high|max
-# (providers/deepseek/models/deepseek-v4-flash.toml). Tinfoil documents only
-# reasoning_effort (no separate thinking.type toggle), matching other Tinfoil
-# reasoners that author effort-only options.
+# Sources (accessed 2026-08-26):
+# - https://inference.tinfoil.sh/v1/models
+# - https://docs.tinfoil.sh/models/chat.md (weights deepseek-ai/DeepSeek-V4-Flash-0731)
+# - https://docs.tinfoil.sh/guides/reasoning.md
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max on Tinfoil.
+# Off is reasoning_effort=none; no separate toggle.
 base_model = "deepseek/deepseek-v4-flash-0731"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "max"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.3
diff --git a/providers/tinfoil/models/kimi-k3.toml b/providers/tinfoil/models/kimi-k3.toml
index d15645dd9ab..208c3ee17e9 100644
--- a/providers/tinfoil/models/kimi-k3.toml
+++ b/providers/tinfoil/models/kimi-k3.toml
@@ -1,12 +1,12 @@
-# Sources (accessed 2026-08-08):
-# - https://inference.tinfoil.sh/v1/models (id kimi-k3; context 256000; input $2/MTok; output $6/MTok; reasoning true)
-# - https://docs.tinfoil.sh/guides/reasoning.md (reasoning_effort on OpenAI-compatible chat; standard scale none…max for Tinfoil chat reasoners)
-# Effort values mirror providers/tinfoil/models/kimi-k2-6.toml and glm-5-2/gemma4-31b peers on this host.
+# Sources (accessed 2026-08-26):
+# - https://inference.tinfoil.sh/v1/models
+# - https://docs.tinfoil.sh/guides/reasoning.md
+# Effort: reasoning_effort = low|high|max on Tinfoil; reasoning is always on.
 base_model = "moonshotai/kimi-k3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "high", "max"]
 
 [cost]
 input = 4
diff --git a/sync.md b/sync.md
index 8e11926ea23..9dae6607429 100644
--- a/sync.md
+++ b/sync.md
@@ -222,8 +222,9 @@ xAI is implemented in `packages/core/src/sync/providers/xai.ts`.
 - Tinfoil is implemented in `packages/core/src/sync/providers/tinfoil.ts`.
 - Source endpoint: `https://inference.tinfoil.sh/v1/models`.
 - No authentication is required; the catalog is public.
-- Existing Tinfoil models are updated from API-authoritative input, output, cached-input pricing, context windows, and catalog availability.
+- Existing Tinfoil models are updated from API-authoritative input, output, cached-input pricing, context windows, reasoning capability, and catalog availability.
 - Provider-specific metadata that the endpoint does not expose, including exact modalities, output limits, reasoning controls, and lifecycle status, remains hand-authored.
+- Reasoning controls are preserved for reasoners and removed when the API reports `reasoning: false`. A reasoner without authored controls fails sync for manual review rather than inventing an empty control set.
 - New token-priced chat, safety, and embedding models are not created automatically (`skipCreates`); each missing ID opens a deduped GitHub issue for hand-authored metadata.
 - Per-request tool, TTS, transcription, realtime, and document-processing services are ignored because their pricing cannot be represented by the token-cost schema.
 

From 4be2746b8d558d40952cc2dd77f1748ace2b185b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:55:09 +0000
Subject: [PATCH 0016/2012] chore(sync): update Kilo model catalog (#5578)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..731fb78b785 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 644d82deb70a89fe600fe2bb68fb508e273ef029 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:55:11 +0000
Subject: [PATCH 0017/2012] chore(sync): update Merge Gateway model catalog
 (#5579)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 0a7a2254ee8..b4bdd294444 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 35e52350151..f30ae74cdb3 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 1e294904efec6de5b0cef3a8b95f4393c2a51d57 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:55:59 +0000
Subject: [PATCH 0018/2012] chore(sync): update OpenRouter model catalog
 (#5582)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/amazon/nova-2-lite-v1.toml      |  6 +++++-
 .../openrouter/models/anthropic/claude-haiku-4.5.toml |  7 ++-----
 .../openrouter/models/anthropic/claude-opus-4.1.toml  |  7 ++-----
 .../openrouter/models/anthropic/claude-opus-4.5.toml  | 11 ++---------
 .../openrouter/models/anthropic/claude-opus-4.toml    |  6 +++++-
 .../models/anthropic/claude-sonnet-4.5.toml           |  7 ++-----
 .../openrouter/models/anthropic/claude-sonnet-4.toml  |  6 +++++-
 .../models/baidu/ernie-4.5-vl-424b-a47b.toml          |  6 +++++-
 .../models/bytedance-seed/seed-1.6-flash.toml         |  6 +++++-
 .../openrouter/models/bytedance-seed/seed-1.6.toml    |  6 +++++-
 .../models/bytedance-seed/seed-2-1-turbo.toml         |  6 +++++-
 .../models/cohere/north-mini-code:free.toml           |  6 +++++-
 .../deepseek/deepseek-r1-distill-llama-70b.toml       |  6 +++++-
 .../models/dots-studio/dots-3-note-preview:free.toml  |  6 +++++-
 .../models/google/gemini-2.5-flash-lite.toml          |  7 ++-----
 .../openrouter/models/google/gemini-2.5-flash.toml    |  7 ++-----
 .../openrouter/models/inclusionai/ling-3.0-flash.toml |  6 +++++-
 providers/openrouter/models/meituan/longcat-2.0.toml  |  5 +++++
 providers/openrouter/models/minimax/minimax-m1.toml   |  6 +++++-
 providers/openrouter/models/minimax/minimax-m3.toml   |  6 +++++-
 .../openrouter/models/minimax/minimax-m3:free.toml    |  6 +++++-
 providers/openrouter/models/moonshotai/kimi-k2.5.toml |  6 +++++-
 providers/openrouter/models/moonshotai/kimi-k2.6.toml |  6 +++++-
 providers/openrouter/models/nex-agi/nex-n2-mini.toml  |  6 +++++-
 providers/openrouter/models/nex-agi/nex-n2-pro.toml   |  6 +++++-
 .../models/nvidia/nemotron-3-nano-30b-a3b.toml        |  6 +++++-
 .../nemotron-3-nano-omni-30b-a3b-reasoning:free.toml  |  5 +++++
 .../nvidia/nemotron-3.5-content-safety:free.toml      |  6 +++++-
 providers/openrouter/models/openai/o1-pro.toml        |  6 +++++-
 providers/openrouter/models/openai/o1.toml            |  5 +++--
 providers/openrouter/models/openai/o3-mini.toml       |  5 +++--
 providers/openrouter/models/openai/o3-pro.toml        |  5 +++--
 providers/openrouter/models/openai/o3.toml            |  5 +++--
 providers/openrouter/models/openai/o4-mini.toml       |  5 +++--
 .../openrouter/models/perceptron/perceptron-mk1.toml  |  6 +++++-
 .../models/perplexity/sonar-deep-research.toml        |  6 +++++-
 .../models/perplexity/sonar-reasoning-pro.toml        |  6 +++++-
 .../openrouter/models/poolside/laguna-s-2.1.toml      |  6 +++++-
 .../openrouter/models/poolside/laguna-s-2.1:free.toml |  6 +++++-
 .../openrouter/models/poolside/laguna-xs-2.1.toml     |  6 +++++-
 .../models/poolside/laguna-xs-2.1:free.toml           |  6 +++++-
 providers/openrouter/models/qwen/qwen3-235b-a22b.toml |  7 ++-----
 .../openrouter/models/qwen/qwen3-max-thinking.toml    |  6 +++---
 .../openrouter/models/qwen/qwen3.5-flash-02-23.toml   |  6 +++++-
 .../openrouter/models/qwen/qwen3.5-plus-02-15.toml    |  7 ++-----
 .../openrouter/models/qwen/qwen3.5-plus-20260420.toml |  7 ++-----
 providers/openrouter/models/qwen/qwen3.6-flash.toml   |  7 ++-----
 .../openrouter/models/qwen/qwen3.6-max-preview.toml   |  7 ++-----
 providers/openrouter/models/qwen/qwen3.6-plus.toml    |  7 ++-----
 providers/openrouter/models/qwen/qwen3.7-flash.toml   |  5 +++++
 providers/openrouter/models/qwen/qwen3.7-max.toml     |  7 ++-----
 providers/openrouter/models/qwen/qwen3.7-plus.toml    |  7 ++-----
 .../models/tencent/hunyuan-a13b-instruct.toml         |  6 +++++-
 providers/openrouter/models/upstage/solar-pro-3.toml  |  6 +++++-
 providers/openrouter/models/upstage/solar-pro4.toml   |  6 +++++-
 providers/openrouter/models/x-ai/grok-4.20.toml       |  6 +++++-
 providers/openrouter/models/z-ai/glm-4.6.toml         |  6 +++++-
 providers/openrouter/models/z-ai/glm-4.6v.toml        |  6 +++++-
 providers/openrouter/models/z-ai/glm-4.7-flash.toml   |  6 +++++-
 providers/openrouter/models/z-ai/glm-4.7.toml         |  6 +++++-
 providers/openrouter/models/z-ai/glm-5-turbo.toml     |  6 +++++-
 providers/openrouter/models/z-ai/glm-5.1.toml         |  6 +++++-
 providers/openrouter/models/z-ai/glm-5.toml           |  6 +++++-
 providers/openrouter/models/z-ai/glm-5v-turbo.toml    |  6 +++++-
 .../models/~anthropic/claude-haiku-latest.toml        |  6 +++++-
 65 files changed, 271 insertions(+), 129 deletions(-)

diff --git a/providers/openrouter/models/amazon/nova-2-lite-v1.toml b/providers/openrouter/models/amazon/nova-2-lite-v1.toml
index 1b51e0c4f1a..ffb27f523af 100644
--- a/providers/openrouter/models/amazon/nova-2-lite-v1.toml
+++ b/providers/openrouter/models/amazon/nova-2-lite-v1.toml
@@ -1,4 +1,5 @@
-reasoning_options = []
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Nova 2 Lite"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "nova"
@@ -11,6 +12,9 @@ tool_call = true
 structured_output = false
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.3
 output = 2.5
diff --git a/providers/openrouter/models/anthropic/claude-haiku-4.5.toml b/providers/openrouter/models/anthropic/claude-haiku-4.5.toml
index af225d0ca15..187e02907a2 100644
--- a/providers/openrouter/models/anthropic/claude-haiku-4.5.toml
+++ b/providers/openrouter/models/anthropic/claude-haiku-4.5.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "anthropic/claude-haiku-4-5"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 63_999
-
 [cost]
 input = 1
 output = 5
diff --git a/providers/openrouter/models/anthropic/claude-opus-4.1.toml b/providers/openrouter/models/anthropic/claude-opus-4.1.toml
index e8fcf991080..e2af373bed2 100644
--- a/providers/openrouter/models/anthropic/claude-opus-4.1.toml
+++ b/providers/openrouter/models/anthropic/claude-opus-4.1.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "anthropic/claude-opus-4-1"
 structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 31_999
-
 [cost]
 input = 15
 output = 75
diff --git a/providers/openrouter/models/anthropic/claude-opus-4.5.toml b/providers/openrouter/models/anthropic/claude-opus-4.5.toml
index 5199bb685e2..28abffd9a66 100644
--- a/providers/openrouter/models/anthropic/claude-opus-4.5.toml
+++ b/providers/openrouter/models/anthropic/claude-opus-4.5.toml
@@ -1,18 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "anthropic/claude-opus-4-5"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 63_999
-
 [cost]
 input = 5
 output = 25
diff --git a/providers/openrouter/models/anthropic/claude-opus-4.toml b/providers/openrouter/models/anthropic/claude-opus-4.toml
index 53e62ce8f24..094f7f59203 100644
--- a/providers/openrouter/models/anthropic/claude-opus-4.toml
+++ b/providers/openrouter/models/anthropic/claude-opus-4.toml
@@ -1,4 +1,5 @@
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 31999 }]
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Claude Opus 4"
 description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
 family = "claude-opus"
@@ -12,6 +13,9 @@ structured_output = false
 knowledge = "2025-01-31"
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 15
 output = 75
diff --git a/providers/openrouter/models/anthropic/claude-sonnet-4.5.toml b/providers/openrouter/models/anthropic/claude-sonnet-4.5.toml
index b8f2ac6f412..e6f4db6ea47 100644
--- a/providers/openrouter/models/anthropic/claude-sonnet-4.5.toml
+++ b/providers/openrouter/models/anthropic/claude-sonnet-4.5.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "anthropic/claude-sonnet-4-5"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 63_999
-
 [cost]
 input = 3
 output = 15
diff --git a/providers/openrouter/models/anthropic/claude-sonnet-4.toml b/providers/openrouter/models/anthropic/claude-sonnet-4.toml
index 7dcce2aabf0..7baa5f4f635 100644
--- a/providers/openrouter/models/anthropic/claude-sonnet-4.toml
+++ b/providers/openrouter/models/anthropic/claude-sonnet-4.toml
@@ -1,4 +1,5 @@
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 63999 }]
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Claude Sonnet 4"
 description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
 family = "claude-sonnet"
@@ -12,6 +13,9 @@ structured_output = false
 knowledge = "2025-01-31"
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 3
 output = 15
diff --git a/providers/openrouter/models/baidu/ernie-4.5-vl-424b-a47b.toml b/providers/openrouter/models/baidu/ernie-4.5-vl-424b-a47b.toml
index 9613f948a58..ca55754cdf6 100644
--- a/providers/openrouter/models/baidu/ernie-4.5-vl-424b-a47b.toml
+++ b/providers/openrouter/models/baidu/ernie-4.5-vl-424b-a47b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "ERNIE 4.5 VL 424B A47B "
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "ernie"
@@ -10,7 +12,9 @@ tool_call = false
 structured_output = false
 knowledge = "2025-03-31"
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.42
diff --git a/providers/openrouter/models/bytedance-seed/seed-1.6-flash.toml b/providers/openrouter/models/bytedance-seed/seed-1.6-flash.toml
index 8b41db7edd3..fc1a19bfeb6 100644
--- a/providers/openrouter/models/bytedance-seed/seed-1.6-flash.toml
+++ b/providers/openrouter/models/bytedance-seed/seed-1.6-flash.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Seed 1.6 Flash"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "seed"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.075
diff --git a/providers/openrouter/models/bytedance-seed/seed-1.6.toml b/providers/openrouter/models/bytedance-seed/seed-1.6.toml
index a1f8f345f22..9790f2d6762 100644
--- a/providers/openrouter/models/bytedance-seed/seed-1.6.toml
+++ b/providers/openrouter/models/bytedance-seed/seed-1.6.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Seed 1.6"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "seed"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.25
diff --git a/providers/openrouter/models/bytedance-seed/seed-2-1-turbo.toml b/providers/openrouter/models/bytedance-seed/seed-2-1-turbo.toml
index 37e4e63fa34..d129903d8f1 100644
--- a/providers/openrouter/models/bytedance-seed/seed-2-1-turbo.toml
+++ b/providers/openrouter/models/bytedance-seed/seed-2-1-turbo.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Seed 2.1 Turbo"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "seed"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.5
diff --git a/providers/openrouter/models/cohere/north-mini-code:free.toml b/providers/openrouter/models/cohere/north-mini-code:free.toml
index 8616b407eae..48751d38e9d 100644
--- a/providers/openrouter/models/cohere/north-mini-code:free.toml
+++ b/providers/openrouter/models/cohere/north-mini-code:free.toml
@@ -1,6 +1,7 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "North Mini Code (free)"
 description = "Cohere coding model for practical software engineering and agentic edits"
-reasoning_options = []
 family = "north"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
@@ -11,6 +12,9 @@ tool_call = true
 structured_output = false
 open_weights = true
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0
 output = 0
diff --git a/providers/openrouter/models/deepseek/deepseek-r1-distill-llama-70b.toml b/providers/openrouter/models/deepseek/deepseek-r1-distill-llama-70b.toml
index a27a50348f9..9d94410007a 100644
--- a/providers/openrouter/models/deepseek/deepseek-r1-distill-llama-70b.toml
+++ b/providers/openrouter/models/deepseek/deepseek-r1-distill-llama-70b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "R1 Distill Llama 70B"
 description = "DeepSeek reasoning model for multi-step analysis, math, coding, and tools"
 family = "deepseek-thinking"
@@ -10,7 +12,9 @@ tool_call = false
 structured_output = false
 knowledge = "2024-07-31"
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.8
diff --git a/providers/openrouter/models/dots-studio/dots-3-note-preview:free.toml b/providers/openrouter/models/dots-studio/dots-3-note-preview:free.toml
index c4cfb09df88..2de0d76750a 100644
--- a/providers/openrouter/models/dots-studio/dots-3-note-preview:free.toml
+++ b/providers/openrouter/models/dots-studio/dots-3-note-preview:free.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Dots3-Note Preview (free)"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 release_date = "2026-08-14"
@@ -8,7 +10,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0
diff --git a/providers/openrouter/models/google/gemini-2.5-flash-lite.toml b/providers/openrouter/models/google/gemini-2.5-flash-lite.toml
index dfb707f9360..f5cc3f5c707 100644
--- a/providers/openrouter/models/google/gemini-2.5-flash-lite.toml
+++ b/providers/openrouter/models/google/gemini-2.5-flash-lite.toml
@@ -1,13 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "google/gemini-2.5-flash-lite"
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 512
-max = 24_576
-
 [cost]
 input = 0.1
 output = 0.4
diff --git a/providers/openrouter/models/google/gemini-2.5-flash.toml b/providers/openrouter/models/google/gemini-2.5-flash.toml
index b465dc01456..9a8ff7ace41 100644
--- a/providers/openrouter/models/google/gemini-2.5-flash.toml
+++ b/providers/openrouter/models/google/gemini-2.5-flash.toml
@@ -1,13 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "google/gemini-2.5-flash"
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 0
-max = 24_576
-
 [cost]
 input = 0.3
 output = 2.5
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
index b9a9c35a085..59e76bd4c20 100644
--- a/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Ling-3.0-flash"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 family = "ling"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = false
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.021
diff --git a/providers/openrouter/models/meituan/longcat-2.0.toml b/providers/openrouter/models/meituan/longcat-2.0.toml
index 29e85d5574e..14881c85746 100644
--- a/providers/openrouter/models/meituan/longcat-2.0.toml
+++ b/providers/openrouter/models/meituan/longcat-2.0.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "LongCat 2.0"
 description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
 family = "longcat"
@@ -10,6 +12,9 @@ tool_call = true
 structured_output = false
 open_weights = true
 
+[[reasoning_options]]
+type = "toggle"
+
 [[reasoning_options]]
 type = "budget_tokens"
 
diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 5663a97b096..178789497a8 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "MiniMax M1"
 description = "MiniMax model for chat, coding, office work, and agentic tasks"
 family = "minimax"
@@ -10,7 +12,9 @@ tool_call = true
 structured_output = false
 knowledge = "2024-06-30"
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.55
diff --git a/providers/openrouter/models/minimax/minimax-m3.toml b/providers/openrouter/models/minimax/minimax-m3.toml
index 4244d4862c3..bff429beeb6 100644
--- a/providers/openrouter/models/minimax/minimax-m3.toml
+++ b/providers/openrouter/models/minimax/minimax-m3.toml
@@ -1,6 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "minimax/MiniMax-M3"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.3
diff --git a/providers/openrouter/models/minimax/minimax-m3:free.toml b/providers/openrouter/models/minimax/minimax-m3:free.toml
index df4f09f7032..961bae5aa8e 100644
--- a/providers/openrouter/models/minimax/minimax-m3:free.toml
+++ b/providers/openrouter/models/minimax/minimax-m3:free.toml
@@ -1,8 +1,12 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "minimax/MiniMax-M3"
 name = "MiniMax M3 (free)"
 description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
 structured_output = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.5.toml b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
index 45ae7506d48..ebb15bee06b 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.5.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
@@ -1,10 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "moonshotai/kimi-k2.5"
 temperature = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_details"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.45
 output = 2.25
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.6.toml b/providers/openrouter/models/moonshotai/kimi-k2.6.toml
index d52ae05dcac..b0351ac363c 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.6.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.6.toml
@@ -1,9 +1,13 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "moonshotai/kimi-k2.6"
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_details"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.95
 output = 4
diff --git a/providers/openrouter/models/nex-agi/nex-n2-mini.toml b/providers/openrouter/models/nex-agi/nex-n2-mini.toml
index 47906f6afd0..ccad25ac0f1 100644
--- a/providers/openrouter/models/nex-agi/nex-n2-mini.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2-mini.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Nex-N2-Mini"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "agi"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.025
diff --git a/providers/openrouter/models/nex-agi/nex-n2-pro.toml b/providers/openrouter/models/nex-agi/nex-n2-pro.toml
index 7fb3dd88a59..df1064ba567 100644
--- a/providers/openrouter/models/nex-agi/nex-n2-pro.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2-pro.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Nex-N2-Pro"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "agi"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = false
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.25
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
index b4271827c2d..bf732029012 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -1,6 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3-nano-30b-a3b"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.05
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.toml
index f80a020a6f1..24b9785876b 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.toml
@@ -1,7 +1,12 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning"
 name = "Nemotron 3 Nano Omni (free)"
 structured_output = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [[reasoning_options]]
 type = "budget_tokens"
 
diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-content-safety:free.toml b/providers/openrouter/models/nvidia/nemotron-3.5-content-safety:free.toml
index 3c9543b28cb..ee73f2eee81 100644
--- a/providers/openrouter/models/nvidia/nemotron-3.5-content-safety:free.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-content-safety:free.toml
@@ -1,7 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3.5-content-safety"
 name = "Nemotron 3.5 Content Safety (free)"
 structured_output = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0
diff --git a/providers/openrouter/models/openai/o1-pro.toml b/providers/openrouter/models/openai/o1-pro.toml
index d9ca9966515..3f0fd40df4b 100644
--- a/providers/openrouter/models/openai/o1-pro.toml
+++ b/providers/openrouter/models/openai/o1-pro.toml
@@ -1,6 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o1-pro"
 tool_call = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 150
diff --git a/providers/openrouter/models/openai/o1.toml b/providers/openrouter/models/openai/o1.toml
index 40b6a2a4af8..0a0dcdc84ef 100644
--- a/providers/openrouter/models/openai/o1.toml
+++ b/providers/openrouter/models/openai/o1.toml
@@ -1,8 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o1"
 
 [[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+type = "toggle"
 
 [cost]
 input = 15
diff --git a/providers/openrouter/models/openai/o3-mini.toml b/providers/openrouter/models/openai/o3-mini.toml
index 59c55ee37a4..ff57052aaaa 100644
--- a/providers/openrouter/models/openai/o3-mini.toml
+++ b/providers/openrouter/models/openai/o3-mini.toml
@@ -1,9 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o3-mini"
 attachment = true
 
 [[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+type = "toggle"
 
 [cost]
 input = 1.1
diff --git a/providers/openrouter/models/openai/o3-pro.toml b/providers/openrouter/models/openai/o3-pro.toml
index b2ece450837..729c144d17a 100644
--- a/providers/openrouter/models/openai/o3-pro.toml
+++ b/providers/openrouter/models/openai/o3-pro.toml
@@ -1,8 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o3-pro"
 
 [[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+type = "toggle"
 
 [cost]
 input = 20
diff --git a/providers/openrouter/models/openai/o3.toml b/providers/openrouter/models/openai/o3.toml
index 977bf918256..b1933ecd752 100644
--- a/providers/openrouter/models/openai/o3.toml
+++ b/providers/openrouter/models/openai/o3.toml
@@ -1,8 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o3"
 
 [[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+type = "toggle"
 
 [cost]
 input = 2
diff --git a/providers/openrouter/models/openai/o4-mini.toml b/providers/openrouter/models/openai/o4-mini.toml
index f97e6e892b8..aed9393a8b0 100644
--- a/providers/openrouter/models/openai/o4-mini.toml
+++ b/providers/openrouter/models/openai/o4-mini.toml
@@ -1,8 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "openai/o4-mini"
 
 [[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+type = "toggle"
 
 [cost]
 input = 1.1
diff --git a/providers/openrouter/models/perceptron/perceptron-mk1.toml b/providers/openrouter/models/perceptron/perceptron-mk1.toml
index 97f315f941a..5f5abd16181 100644
--- a/providers/openrouter/models/perceptron/perceptron-mk1.toml
+++ b/providers/openrouter/models/perceptron/perceptron-mk1.toml
@@ -1,4 +1,5 @@
-reasoning_options = []
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Perceptron Mk1"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 release_date = "2026-05-12"
@@ -10,6 +11,9 @@ tool_call = false
 structured_output = true
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.15
 output = 1.5
diff --git a/providers/openrouter/models/perplexity/sonar-deep-research.toml b/providers/openrouter/models/perplexity/sonar-deep-research.toml
index 317389ef351..3ba392526b3 100644
--- a/providers/openrouter/models/perplexity/sonar-deep-research.toml
+++ b/providers/openrouter/models/perplexity/sonar-deep-research.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Sonar Deep Research"
 description = "Sonar search model for current answers, retrieval, and citation-backed chat"
 family = "sonar-deep-research"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = false
 structured_output = false
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 2
diff --git a/providers/openrouter/models/perplexity/sonar-reasoning-pro.toml b/providers/openrouter/models/perplexity/sonar-reasoning-pro.toml
index 216c17e90ed..aceaa472fd5 100644
--- a/providers/openrouter/models/perplexity/sonar-reasoning-pro.toml
+++ b/providers/openrouter/models/perplexity/sonar-reasoning-pro.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Sonar Reasoning Pro"
 description = "Web-grounded reasoning model for multi-step research and cited answers"
 family = "sonar-reasoning"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = false
 structured_output = false
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 2
diff --git a/providers/openrouter/models/poolside/laguna-s-2.1.toml b/providers/openrouter/models/poolside/laguna-s-2.1.toml
index 0357d84da69..75ddbd0b613 100644
--- a/providers/openrouter/models/poolside/laguna-s-2.1.toml
+++ b/providers/openrouter/models/poolside/laguna-s-2.1.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Laguna S 2.1"
 description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
 family = "laguna-s"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = false
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.09
diff --git a/providers/openrouter/models/poolside/laguna-s-2.1:free.toml b/providers/openrouter/models/poolside/laguna-s-2.1:free.toml
index 85050a7ee50..8e3c51877d8 100644
--- a/providers/openrouter/models/poolside/laguna-s-2.1:free.toml
+++ b/providers/openrouter/models/poolside/laguna-s-2.1:free.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Laguna S 2.1 (free)"
 description = "Free provider route for experiments, demos, and cost-sensitive chat workloads"
 family = "laguna-s"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = false
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0
diff --git a/providers/openrouter/models/poolside/laguna-xs-2.1.toml b/providers/openrouter/models/poolside/laguna-xs-2.1.toml
index f3a506b8121..f2dc3feeac9 100644
--- a/providers/openrouter/models/poolside/laguna-xs-2.1.toml
+++ b/providers/openrouter/models/poolside/laguna-xs-2.1.toml
@@ -1,5 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "poolside/laguna-xs-2.1"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.06
diff --git a/providers/openrouter/models/poolside/laguna-xs-2.1:free.toml b/providers/openrouter/models/poolside/laguna-xs-2.1:free.toml
index 9e6fbbab9d3..3f290d3dce8 100644
--- a/providers/openrouter/models/poolside/laguna-xs-2.1:free.toml
+++ b/providers/openrouter/models/poolside/laguna-xs-2.1:free.toml
@@ -1,7 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "poolside/laguna-xs-2.1"
 name = "Laguna XS 2.1 (free)"
 description = "Free provider route for experiments, demos, and cost-sensitive chat workloads"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b.toml
index c607f7353cc..3f3f59b0214 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3-235b-a22b"
 structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 38_912
-
 [cost]
 input = 0.455
 output = 1.82
diff --git a/providers/openrouter/models/qwen/qwen3-max-thinking.toml b/providers/openrouter/models/qwen/qwen3-max-thinking.toml
index 199f89f89ba..7eb62a6a784 100644
--- a/providers/openrouter/models/qwen/qwen3-max-thinking.toml
+++ b/providers/openrouter/models/qwen/qwen3-max-thinking.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Qwen3 Max Thinking"
 description = "Qwen reasoning model for deliberate problem solving, math, and coding"
 family = "qwen"
@@ -11,9 +13,7 @@ structured_output = true
 open_weights = false
 
 [[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 81_920
+type = "toggle"
 
 [cost]
 input = 0.78
diff --git a/providers/openrouter/models/qwen/qwen3.5-flash-02-23.toml b/providers/openrouter/models/qwen/qwen3.5-flash-02-23.toml
index d2d16963acc..b794840a4f0 100644
--- a/providers/openrouter/models/qwen/qwen3.5-flash-02-23.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-flash-02-23.toml
@@ -1,4 +1,5 @@
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1, max = 81920 }]
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Qwen3.5-Flash"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
@@ -11,6 +12,9 @@ tool_call = true
 structured_output = true
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.065
 output = 0.26
diff --git a/providers/openrouter/models/qwen/qwen3.5-plus-02-15.toml b/providers/openrouter/models/qwen/qwen3.5-plus-02-15.toml
index 42ace991588..72c8ade8b38 100644
--- a/providers/openrouter/models/qwen/qwen3.5-plus-02-15.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-plus-02-15.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Qwen3.5 Plus 2026-02-15"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
@@ -14,11 +16,6 @@ open_weights = false
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 81_920
-
 [cost]
 input = 0.26
 output = 1.56
diff --git a/providers/openrouter/models/qwen/qwen3.5-plus-20260420.toml b/providers/openrouter/models/qwen/qwen3.5-plus-20260420.toml
index 88f4ec4b376..8b453ba9902 100644
--- a/providers/openrouter/models/qwen/qwen3.5-plus-20260420.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-plus-20260420.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Qwen3.5 Plus 2026-04-20"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen3.5"
@@ -13,11 +15,6 @@ open_weights = false
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 81_920
-
 [cost]
 input = 0.3
 output = 1.8
diff --git a/providers/openrouter/models/qwen/qwen3.6-flash.toml b/providers/openrouter/models/qwen/qwen3.6-flash.toml
index 5f723cbff26..efbf5fea742 100644
--- a/providers/openrouter/models/qwen/qwen3.6-flash.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-flash.toml
@@ -1,13 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.6-flash"
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 81_920
-
 [cost]
 input = 0.1875
 output = 1.125
diff --git a/providers/openrouter/models/qwen/qwen3.6-max-preview.toml b/providers/openrouter/models/qwen/qwen3.6-max-preview.toml
index f6abb54fe40..635c24238b4 100644
--- a/providers/openrouter/models/qwen/qwen3.6-max-preview.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-max-preview.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.6-max-preview"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 131_072
-
 [cost]
 input = 1.027
 output = 6.162
diff --git a/providers/openrouter/models/qwen/qwen3.6-plus.toml b/providers/openrouter/models/qwen/qwen3.6-plus.toml
index 601d87077b1..34291103593 100644
--- a/providers/openrouter/models/qwen/qwen3.6-plus.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-plus.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.6-plus"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 81_920
-
 [cost]
 input = 0.325
 output = 1.95
diff --git a/providers/openrouter/models/qwen/qwen3.7-flash.toml b/providers/openrouter/models/qwen/qwen3.7-flash.toml
index 69e72585c8c..0b62b4dce7a 100644
--- a/providers/openrouter/models/qwen/qwen3.7-flash.toml
+++ b/providers/openrouter/models/qwen/qwen3.7-flash.toml
@@ -1,7 +1,12 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.7-flash"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 structured_output = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [[reasoning_options]]
 type = "budget_tokens"
 
diff --git a/providers/openrouter/models/qwen/qwen3.7-max.toml b/providers/openrouter/models/qwen/qwen3.7-max.toml
index 772a73bde77..08fbca07c1e 100644
--- a/providers/openrouter/models/qwen/qwen3.7-max.toml
+++ b/providers/openrouter/models/qwen/qwen3.7-max.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.7-max"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 262_144
-
 [cost]
 input = 1.475
 output = 4.425
diff --git a/providers/openrouter/models/qwen/qwen3.7-plus.toml b/providers/openrouter/models/qwen/qwen3.7-plus.toml
index 50868c3df1f..532e531a541 100644
--- a/providers/openrouter/models/qwen/qwen3.7-plus.toml
+++ b/providers/openrouter/models/qwen/qwen3.7-plus.toml
@@ -1,14 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.7-plus"
 structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
 
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1
-max = 262_144
-
 [cost]
 input = 0.32
 output = 1.28
diff --git a/providers/openrouter/models/tencent/hunyuan-a13b-instruct.toml b/providers/openrouter/models/tencent/hunyuan-a13b-instruct.toml
index 4af12259f08..d9d48787987 100644
--- a/providers/openrouter/models/tencent/hunyuan-a13b-instruct.toml
+++ b/providers/openrouter/models/tencent/hunyuan-a13b-instruct.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Hunyuan A13B Instruct"
 description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
 family = "hunyuan"
@@ -10,7 +12,9 @@ tool_call = false
 structured_output = true
 knowledge = "2025-03-31"
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.14
diff --git a/providers/openrouter/models/upstage/solar-pro-3.toml b/providers/openrouter/models/upstage/solar-pro-3.toml
index 7d71a472c5f..146505f4863 100644
--- a/providers/openrouter/models/upstage/solar-pro-3.toml
+++ b/providers/openrouter/models/upstage/solar-pro-3.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Solar Pro 3"
 description = "Flagship model for demanding analysis, coding, and production agent workflows"
 family = "solar-pro"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.15
diff --git a/providers/openrouter/models/upstage/solar-pro4.toml b/providers/openrouter/models/upstage/solar-pro4.toml
index 668c1cd157e..b88e5ba9776 100644
--- a/providers/openrouter/models/upstage/solar-pro4.toml
+++ b/providers/openrouter/models/upstage/solar-pro4.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Solar Pro 4"
 description = "Flagship model for demanding analysis, coding, and production agent workflows"
 family = "solar"
@@ -9,7 +11,9 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.03
diff --git a/providers/openrouter/models/x-ai/grok-4.20.toml b/providers/openrouter/models/x-ai/grok-4.20.toml
index ef19c8b3c7a..0f5920b9caf 100644
--- a/providers/openrouter/models/x-ai/grok-4.20.toml
+++ b/providers/openrouter/models/x-ai/grok-4.20.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Grok 4.20"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
 family = "grok"
@@ -10,7 +12,9 @@ tool_call = true
 structured_output = true
 knowledge = "2025-09-01"
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 1.25
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index cae4e80102b..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -1,6 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-4.6"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.5
diff --git a/providers/openrouter/models/z-ai/glm-4.6v.toml b/providers/openrouter/models/z-ai/glm-4.6v.toml
index f52538d7580..8a395d6a0c1 100644
--- a/providers/openrouter/models/z-ai/glm-4.6v.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6v.toml
@@ -1,6 +1,10 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-4.6v"
 structured_output = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.3
diff --git a/providers/openrouter/models/z-ai/glm-4.7-flash.toml b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
index 378e86c424d..d2109236f90 100644
--- a/providers/openrouter/models/z-ai/glm-4.7-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
@@ -1,10 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-4.7-flash"
 structured_output = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_details"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.06
 output = 0.4
diff --git a/providers/openrouter/models/z-ai/glm-4.7.toml b/providers/openrouter/models/z-ai/glm-4.7.toml
index 5055d1b819a..6471f3670cd 100644
--- a/providers/openrouter/models/z-ai/glm-4.7.toml
+++ b/providers/openrouter/models/z-ai/glm-4.7.toml
@@ -1,10 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-4.7"
 structured_output = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_details"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.4
 output = 1.75
diff --git a/providers/openrouter/models/z-ai/glm-5-turbo.toml b/providers/openrouter/models/z-ai/glm-5-turbo.toml
index beeb3705236..906c3480673 100644
--- a/providers/openrouter/models/z-ai/glm-5-turbo.toml
+++ b/providers/openrouter/models/z-ai/glm-5-turbo.toml
@@ -1,10 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-5-turbo"
 structured_output = false
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 1.2
 output = 4
diff --git a/providers/openrouter/models/z-ai/glm-5.1.toml b/providers/openrouter/models/z-ai/glm-5.1.toml
index ecc2b7c648d..23d665940d6 100644
--- a/providers/openrouter/models/z-ai/glm-5.1.toml
+++ b/providers/openrouter/models/z-ai/glm-5.1.toml
@@ -1,9 +1,13 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-5.1"
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 1.26
 output = 3.96
diff --git a/providers/openrouter/models/z-ai/glm-5.toml b/providers/openrouter/models/z-ai/glm-5.toml
index c7ad523fd4a..49f3f9e0744 100644
--- a/providers/openrouter/models/z-ai/glm-5.toml
+++ b/providers/openrouter/models/z-ai/glm-5.toml
@@ -1,10 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-5"
 structured_output = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.6
 output = 1.92
diff --git a/providers/openrouter/models/z-ai/glm-5v-turbo.toml b/providers/openrouter/models/z-ai/glm-5v-turbo.toml
index 94f84f0e662..d5df6f56d83 100644
--- a/providers/openrouter/models/z-ai/glm-5v-turbo.toml
+++ b/providers/openrouter/models/z-ai/glm-5v-turbo.toml
@@ -1,7 +1,11 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-5v-turbo"
-reasoning_options = []
 structured_output = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 1.2
 output = 4
diff --git a/providers/openrouter/models/~anthropic/claude-haiku-latest.toml b/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
index 85dc865f562..6aae1a4db8d 100644
--- a/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
+++ b/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
@@ -1,4 +1,5 @@
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 63999 }]
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Anthropic Claude Haiku Latest"
 description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
 family = "claude-haiku"
@@ -11,6 +12,9 @@ tool_call = true
 structured_output = true
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 1
 output = 5

From f2737bd9874e38fdacef7133843a82dd936abbea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:56:39 +0000
Subject: [PATCH 0019/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5580)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/meta/muse-image-1.0.toml      | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/vercel/models/meta/muse-image-1.0.toml

diff --git a/providers/vercel/models/meta/muse-image-1.0.toml b/providers/vercel/models/meta/muse-image-1.0.toml
new file mode 100644
index 00000000000..723480c3b26
--- /dev/null
+++ b/providers/vercel/models/meta/muse-image-1.0.toml
@@ -0,0 +1,17 @@
+name = "Muse Image 1.0"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "muse"
+release_date = "2026-08-26"
+last_updated = "2026-08-26"
+attachment = true
+reasoning = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]

From 5c8e252f34fff156324daae8ec6f0c4a814aeb15 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:56:44 +0000
Subject: [PATCH 0020/2012] chore(sync): update Inceptron model catalog (#5581)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/inceptron/models/moonshotai/Kimi-K2.6.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
index d0ba64678f6..e47b97eabb0 100644
--- a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.56
+input = 0.54
 output = 3.39
-cache_read = 0.17
+cache_read = 0.13
 cache_write = 0
 
 [modalities]

From e839f82a014cb7dc0cc509d9fb7ca08a915db16f Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Wed, 26 Aug 2026 13:11:13 -0700
Subject: [PATCH 0021/2012] fix(sync): map spacexai to canonical xai metadata
 and treat empty reasoning_options as unset

The gateway renamed its xai provider to spacexai, but
CANONICAL_PROVIDER_PREFIXES had no spacexai entry, so the Vercel sync
created the renamed files without base_model resolution. With no base
to inherit from, preserveReasoningOptions stamped reasoning_options = []
on every new spacexai file, and the existing ?? base preservation kept
that empty array forever, erasing the effort menus the old xai files
carried.

- Add spacexai to CANONICAL_PROVIDER_PREFIXES so the renamed models
  factor onto the canonical xai metadata again (verified via
  models:sync vercel: base_model restored, duplicated fields factored
  out, no other output changes).
- In buildVercelModel, treat an empty existing reasoning_options array
  as unset so it no longer shadows the route base's menu.
- Re-author the lost effort menus, mirroring the first-party xAI
  entries: grok-4.3 (none/low/medium/high), grok-4.5 (low/medium/high),
  grok-4.6 and grok-4.20-multi-agent(+beta) (low/medium/high/xhigh),
  and gpt-5-codex (low/medium/high, matching what the gateway
  advertises today). Models whose first-party entry also declares no
  controls (grok-4.20-reasoning, grok-build-0.1, grok-4.1-fast-*) keep
  the empty array.
---
 .../core/src/sync/providers/openrouter.ts     |  1 +
 packages/core/src/sync/providers/vercel.ts    |  4 +-
 packages/core/test/sync.test.ts               | 73 +++++++++++++++++++
 .../vercel/models/openai/gpt-5-codex.toml     |  4 +-
 .../spacexai/grok-4.20-multi-agent-beta.toml  |  4 +-
 .../spacexai/grok-4.20-multi-agent.toml       |  4 +-
 .../vercel/models/spacexai/grok-4.3.toml      | 18 +----
 .../vercel/models/spacexai/grok-4.5.toml      | 18 +----
 .../vercel/models/spacexai/grok-4.6.toml      | 21 +-----
 .../models/spacexai/grok-build-0.1.toml       | 14 +---
 .../spacexai/grok-imagine-image-2.0.toml      | 10 +--
 .../spacexai/grok-imagine-video-1.5.toml      | 10 +--
 12 files changed, 101 insertions(+), 80 deletions(-)

diff --git a/packages/core/src/sync/providers/openrouter.ts b/packages/core/src/sync/providers/openrouter.ts
index 58764e308a9..82a9cfc76b5 100644
--- a/packages/core/src/sync/providers/openrouter.ts
+++ b/packages/core/src/sync/providers/openrouter.ts
@@ -44,6 +44,7 @@ const CANONICAL_PROVIDER_PREFIXES = {
   thinkingmachines: { provider: "thinkingmachines", metadata: "thinkingmachines" },
   "x-ai": { provider: "xai", metadata: "xai" },
   xai: { provider: "xai", metadata: "xai" },
+  spacexai: { provider: "xai", metadata: "xai" },
   xiaomi: { provider: "xiaomi", metadata: "xiaomi" },
   zai: { provider: "zai", metadata: "zhipuai" },
   "z-ai": { provider: "zai", metadata: "zhipuai" },
diff --git a/packages/core/src/sync/providers/vercel.ts b/packages/core/src/sync/providers/vercel.ts
index 621b9ca3b01..714f0e9bf4a 100644
--- a/packages/core/src/sync/providers/vercel.ts
+++ b/packages/core/src/sync/providers/vercel.ts
@@ -145,7 +145,9 @@ export function buildVercelModel(
     last_updated: existing?.last_updated ?? releaseDate,
     attachment: existing?.attachment ?? (tags.has("vision") || tags.has("file-input")),
     reasoning: existing?.reasoning ?? tags.has("reasoning"),
-    reasoning_options: existing?.reasoning_options ?? base?.reasoning_options,
+    reasoning_options: existing?.reasoning_options?.length
+      ? existing.reasoning_options
+      : base?.reasoning_options,
     temperature: existing?.temperature,
     tool_call: model.type === "language"
       ? existing?.tool_call ?? tags.has("tool-use")
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 9266e2b120e..b78a69d92e0 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2910,6 +2910,11 @@ test("factors OpenRouter Pro routes against canonical OpenAI metadata", () => {
   expect("release_date" in model).toBe(false);
 });
 
+// The gateway renamed its xAI provider to spacexai; canonical metadata stays xai.
+test("resolves SpaceXAI provider IDs to canonical xAI metadata", () => {
+  expect(resolveCanonicalBaseModel("spacexai/grok-4.5")).toBe("xai/grok-4.5");
+});
+
 // Ensures Merge Gateway namespaces reuse the matching canonical model metadata.
 test("resolves Merge Gateway provider aliases to canonical metadata", () => {
   expect([
@@ -4213,6 +4218,74 @@ test("Vercel Claude Opus fast variants factor onto base opus metadata", () => {
   expect(synced).not.toHaveProperty("family");
 });
 
+// A sync-stamped empty array means "unknown", not "no controls", so it must
+// not shadow the route base's menu forever (spacexai rename fallout).
+test("Vercel empty existing reasoning_options falls back to the route base menu", () => {
+  const [model] = vercel.parseModels({
+    data: [{
+      id: "minimax/minimax-m2.7-free",
+      name: "MiniMax M2.7 (Free)",
+      created: 1_784_160_000,
+      context_window: 200_000,
+      max_tokens: 128_000,
+      type: "language",
+      tags: ["reasoning", "tool-use"],
+      pricing: { input: "0", output: "0" },
+    }],
+  });
+
+  const translated = vercel.translateModel(model!, {
+    existing(id) {
+      if (id === "minimax/minimax-m2.7-free") return { reasoning_options: [] };
+      if (id === "minimax/minimax-m2.7") {
+        return { reasoning_options: [{ type: "effort", values: ["low", "high"] }] };
+      }
+      return undefined;
+    },
+    authored() {
+      return undefined;
+    },
+  });
+
+  expect(translated?.model).toMatchObject({
+    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
+  });
+});
+
+test("Vercel preserves a non-empty existing reasoning_options over the base menu", () => {
+  const [model] = vercel.parseModels({
+    data: [{
+      id: "minimax/minimax-m2.7-free",
+      name: "MiniMax M2.7 (Free)",
+      created: 1_784_160_000,
+      context_window: 200_000,
+      max_tokens: 128_000,
+      type: "language",
+      tags: ["reasoning", "tool-use"],
+      pricing: { input: "0", output: "0" },
+    }],
+  });
+
+  const translated = vercel.translateModel(model!, {
+    existing(id) {
+      if (id === "minimax/minimax-m2.7-free") {
+        return { reasoning_options: [{ type: "toggle" }] };
+      }
+      if (id === "minimax/minimax-m2.7") {
+        return { reasoning_options: [{ type: "effort", values: ["low", "high"] }] };
+      }
+      return undefined;
+    },
+    authored() {
+      return undefined;
+    },
+  });
+
+  expect(translated?.model).toMatchObject({
+    reasoning_options: [{ type: "toggle" }],
+  });
+});
+
 test("OpenRouter Claude Opus fast variants factor onto base opus metadata", () => {
   const model = buildOpenRouterModel(openRouterModel({
     id: "anthropic/claude-opus-5-fast",
diff --git a/providers/vercel/models/openai/gpt-5-codex.toml b/providers/vercel/models/openai/gpt-5-codex.toml
index a8bf02f0ba2..8d331a002fe 100644
--- a/providers/vercel/models/openai/gpt-5-codex.toml
+++ b/providers/vercel/models/openai/gpt-5-codex.toml
@@ -1,6 +1,8 @@
 base_model = "openai/gpt-5-codex"
 attachment = true
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
index ed37aa69ee5..b9d8c9c3eba 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
@@ -7,7 +7,9 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
index 37a06048cc0..fe33d6ebf44 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
@@ -7,7 +7,9 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.3.toml b/providers/vercel/models/spacexai/grok-4.3.toml
index 665e79a33d9..32e4d4f5357 100644
--- a/providers/vercel/models/spacexai/grok-4.3.toml
+++ b/providers/vercel/models/spacexai/grok-4.3.toml
@@ -1,13 +1,8 @@
-name = "Grok 4.3"
+base_model = "xai/grok-4.3"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-family = "grok"
-release_date = "2026-04-30"
-last_updated = "2026-04-30"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
 
 [cost]
 input = 1.25
@@ -15,9 +10,4 @@ output = 2.5
 cache_read = 0.2
 
 [limit]
-context = 1_000_000
 output = 1_000_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/vercel/models/spacexai/grok-4.5.toml b/providers/vercel/models/spacexai/grok-4.5.toml
index a995203dacc..2eb4ebf0d6e 100644
--- a/providers/vercel/models/spacexai/grok-4.5.toml
+++ b/providers/vercel/models/spacexai/grok-4.5.toml
@@ -1,23 +1,13 @@
-name = "Grok 4.5"
+base_model = "xai/grok-4.5"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-family = "grok"
-release_date = "2026-07-08"
-last_updated = "2026-07-08"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 2
 output = 6
 cache_read = 0.3
 
-[limit]
-context = 500_000
-output = 500_000
-
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/vercel/models/spacexai/grok-4.6.toml b/providers/vercel/models/spacexai/grok-4.6.toml
index 4ecff4de95c..f44f1b0a0b5 100644
--- a/providers/vercel/models/spacexai/grok-4.6.toml
+++ b/providers/vercel/models/spacexai/grok-4.6.toml
@@ -1,23 +1,10 @@
-name = "Grok 4.6"
+base_model = "xai/grok-4.6"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-family = "grok"
-release_date = "2026-08-12"
-last_updated = "2026-08-12"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
 
 [cost]
 input = 2
 output = 6
 cache_read = 0.5
-
-[limit]
-context = 500_000
-output = 500_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/spacexai/grok-build-0.1.toml b/providers/vercel/models/spacexai/grok-build-0.1.toml
index 70d07cae413..dfcdb207ca7 100644
--- a/providers/vercel/models/spacexai/grok-build-0.1.toml
+++ b/providers/vercel/models/spacexai/grok-build-0.1.toml
@@ -1,12 +1,5 @@
-name = "Grok Build 0.1"
+base_model = "xai/grok-build-0.1"
 description = "Grok coding model for agentic engineering, edits, and codebase workflows"
-family = "grok-build"
-release_date = "2026-05-20"
-last_updated = "2026-05-20"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
 reasoning_options = []
 
 [cost]
@@ -14,10 +7,5 @@ input = 1
 output = 2
 cache_read = 0.2
 
-[limit]
-context = 256_000
-output = 256_000
-
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/spacexai/grok-imagine-image-2.0.toml b/providers/vercel/models/spacexai/grok-imagine-image-2.0.toml
index 35dbfb81f23..64ba6e53db2 100644
--- a/providers/vercel/models/spacexai/grok-imagine-image-2.0.toml
+++ b/providers/vercel/models/spacexai/grok-imagine-image-2.0.toml
@@ -1,17 +1,9 @@
-name = "Grok Imagine Image 2.0"
+base_model = "xai/grok-imagine-image-2.0"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "grok"
-release_date = "2026-08-07"
-last_updated = "2026-08-07"
 attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
 
 [limit]
 context = 0
-output = 0
 
 [modalities]
 input = ["text"]
-output = ["image"]
diff --git a/providers/vercel/models/spacexai/grok-imagine-video-1.5.toml b/providers/vercel/models/spacexai/grok-imagine-video-1.5.toml
index 9ad28a47401..567198f09b3 100644
--- a/providers/vercel/models/spacexai/grok-imagine-video-1.5.toml
+++ b/providers/vercel/models/spacexai/grok-imagine-video-1.5.toml
@@ -1,17 +1,9 @@
-name = "Grok Imagine Video 1.5"
+base_model = "xai/grok-imagine-video-1.5"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "grok"
-release_date = "2026-06-22"
-last_updated = "2026-06-22"
 attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
 
 [limit]
 context = 0
-output = 0
 
 [modalities]
 input = ["text"]
-output = ["video"]

From 907c0f535e56f8e8a3b7acf4977a55ad4876b564 Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Wed, 26 Aug 2026 13:40:59 -0700
Subject: [PATCH 0022/2012] test(sync): drop explanatory comments from new
 reasoning_options tests

---
 packages/core/test/sync.test.ts | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index b78a69d92e0..84266188070 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2910,7 +2910,6 @@ test("factors OpenRouter Pro routes against canonical OpenAI metadata", () => {
   expect("release_date" in model).toBe(false);
 });
 
-// The gateway renamed its xAI provider to spacexai; canonical metadata stays xai.
 test("resolves SpaceXAI provider IDs to canonical xAI metadata", () => {
   expect(resolveCanonicalBaseModel("spacexai/grok-4.5")).toBe("xai/grok-4.5");
 });
@@ -4218,8 +4217,6 @@ test("Vercel Claude Opus fast variants factor onto base opus metadata", () => {
   expect(synced).not.toHaveProperty("family");
 });
 
-// A sync-stamped empty array means "unknown", not "no controls", so it must
-// not shadow the route base's menu forever (spacexai rename fallout).
 test("Vercel empty existing reasoning_options falls back to the route base menu", () => {
   const [model] = vercel.parseModels({
     data: [{

From 734ac27f6ba5e6abde64cedda45680b9258c0586 Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Wed, 26 Aug 2026 13:52:56 -0700
Subject: [PATCH 0023/2012] fix(sync): scope vercel reasoning menus to what the
 gateway bridge honors

Empirical probes against the xAI API plus the gateway reasoning bridge
(materializeXai clamps none/minimal->low and xhigh/max->high, and only
grok-3-mini/4.3/4.5/4.6 families support effort at all) show the
previously restored menus over-advertised:

- grok-4.3: drop none (gateway clamps it to low; none is a real off
  switch only on xAI direct, verified 0 reasoning tokens)
- grok-4.6: drop xhigh (gateway clamps it to high)
- grok-4.20-multi-agent(+beta): revert to empty, the bridge ignores
  effort entirely for this family, so a menu would be decorative

gpt-5-codex keeps low/medium/high (the gateway OpenAI dialect passes
effort through; direct verification blocked, the model 404s on our
OpenAI key).
---
 .../vercel/models/spacexai/grok-4.20-multi-agent-beta.toml    | 4 +---
 providers/vercel/models/spacexai/grok-4.20-multi-agent.toml   | 4 +---
 providers/vercel/models/spacexai/grok-4.3.toml                | 2 +-
 providers/vercel/models/spacexai/grok-4.6.toml                | 2 +-
 4 files changed, 4 insertions(+), 8 deletions(-)

diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
index b9d8c9c3eba..ed37aa69ee5 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
@@ -7,9 +7,7 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh"]
+reasoning_options = []
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
index fe33d6ebf44..37a06048cc0 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
@@ -7,9 +7,7 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh"]
+reasoning_options = []
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.3.toml b/providers/vercel/models/spacexai/grok-4.3.toml
index 32e4d4f5357..d379c871d7f 100644
--- a/providers/vercel/models/spacexai/grok-4.3.toml
+++ b/providers/vercel/models/spacexai/grok-4.3.toml
@@ -2,7 +2,7 @@ base_model = "xai/grok-4.3"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high"]
+values = ["low", "medium", "high"]
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/spacexai/grok-4.6.toml b/providers/vercel/models/spacexai/grok-4.6.toml
index f44f1b0a0b5..713ac05e4b3 100644
--- a/providers/vercel/models/spacexai/grok-4.6.toml
+++ b/providers/vercel/models/spacexai/grok-4.6.toml
@@ -2,7 +2,7 @@ base_model = "xai/grok-4.6"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high", "xhigh"]
+values = ["low", "medium", "high"]
 
 [cost]
 input = 2

From 2254052d7cabc69a80b99b820656c59e0ee90d4d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:07:43 +0000
Subject: [PATCH 0024/2012] chore(sync): update Venice model catalog (#5587)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/z-ai-glm-5-3-flash.toml | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/providers/venice/models/z-ai-glm-5-3-flash.toml b/providers/venice/models/z-ai-glm-5-3-flash.toml
index 04f19975d31..dc4c775c583 100644
--- a/providers/venice/models/z-ai-glm-5-3-flash.toml
+++ b/providers/venice/models/z-ai-glm-5-3-flash.toml
@@ -2,10 +2,8 @@ base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
 release_date = "2026-08-21"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
+open_weights = true
+reasoning_options = []
 
 [cost]
 input = 0.09375

From 22d169a9aed3bc257978229d69f3a9167d751d94 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:07:48 +0000
Subject: [PATCH 0025/2012] chore(sync): update Merge Gateway model catalog
 (#5586)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 16d15cbb411..c680bd0f847 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +13,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 3a9871738b8c7adb4575ee8e21d7016bdb136db8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:07:53 +0000
Subject: [PATCH 0026/2012] chore(sync): update Weights & Biases model catalog
 (#5589)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml | 4 ++--
 providers/wandb/models/meta-llama/Llama-3.1-8B-Instruct.toml  | 3 ++-
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml b/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
index 2b70835acf0..7f9934ecc96 100644
--- a/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
+++ b/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
@@ -16,8 +16,8 @@ output = 0.8
 cache_read = 0.8
 
 [limit]
-context = 128_000
-output = 128_000
+context = 131_072
+output = 131_072
 
 [modalities]
 input = ["text"]
diff --git a/providers/wandb/models/meta-llama/Llama-3.1-8B-Instruct.toml b/providers/wandb/models/meta-llama/Llama-3.1-8B-Instruct.toml
index 08af964b412..824e662b391 100644
--- a/providers/wandb/models/meta-llama/Llama-3.1-8B-Instruct.toml
+++ b/providers/wandb/models/meta-llama/Llama-3.1-8B-Instruct.toml
@@ -11,4 +11,5 @@ output = 0.22
 cache_read = 0.22
 
 [limit]
-output = 128_000
+context = 131_072
+output = 131_072

From b7b836acc305f337115a3d48154cf2493fb184db Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:08:08 +0000
Subject: [PATCH 0027/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5591)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/zai/glm-5.3-flash.toml | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/providers/vercel/models/zai/glm-5.3-flash.toml b/providers/vercel/models/zai/glm-5.3-flash.toml
index 91deb939f46..83f4bcc9383 100644
--- a/providers/vercel/models/zai/glm-5.3-flash.toml
+++ b/providers/vercel/models/zai/glm-5.3-flash.toml
@@ -3,12 +3,15 @@
 # Thinking is always enabled; no toggle or reasoning-token budget.
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 output = 131_000

From ca6dc8eec924f89816baa8d4cccf50fc425fb69c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:08:29 +0000
Subject: [PATCH 0028/2012] chore(sync): update OpenRouter model catalog
 (#5590)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k2.5.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k2.5.toml b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
index ebb15bee06b..7943f5c3aaa 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.5.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
@@ -10,9 +10,9 @@ field = "reasoning_details"
 type = "toggle"
 
 [cost]
-input = 0.45
-output = 2.25
-cache_read = 0.07
+input = 0.6
+output = 3
+cache_read = 0.1
 
 [limit]
 output = 235_929

From 797ec04841cd75f55e03f9bc87d6273472fde6df Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 16:25:05 -0500
Subject: [PATCH 0029/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#5539)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3-flash.toml     | 25 +++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
new file mode 100644
index 00000000000..c242092c4a7
--- /dev/null
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -0,0 +1,25 @@
+name = "Glm 5.3 Flash"
+description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+family = "glm"
+release_date = "2026-08-26"
+last_updated = "2026-08-26"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_310_720
+output = 1_310_720
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 9da1ee01a77fe259c7c3034d6c347f930391fc9d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 22:16:33 +0000
Subject: [PATCH 0030/2012] chore(sync): update Kilo model catalog (#5592)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-flash.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/kilo/models/qwen/qwen3.8-flash.toml

diff --git a/providers/kilo/models/qwen/qwen3.8-flash.toml b/providers/kilo/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..16ecba16400
--- /dev/null
+++ b/providers/kilo/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,13 @@
+base_model = "alibaba/qwen3.8-flash"
+description = "Qwen3.8 Flash is a multimodal reasoning model from Alibaba. It is suited for coding assistance, agentic workflows, visual understanding, document and codebase analysis, desktop interaction, chart analysis, and long-video analysis."
+temperature = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From 3f93d08ccf0a63cdf028fcdafa661eb8f7f2f657 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 22:16:39 +0000
Subject: [PATCH 0031/2012] chore(sync): update OpenRouter model catalog
 (#5594)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml       |  8 +++++---
 .../openrouter/models/qwen/qwen3.8-flash.toml    | 16 ++++++++++++++++
 2 files changed, 21 insertions(+), 3 deletions(-)
 create mode 100644 providers/openrouter/models/qwen/qwen3.8-flash.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b899aaefaad..fcb4e40e521 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-flash"
 
 [interleaved]
@@ -11,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07952
-output = 0.15904
-cache_read = 0.015904
+input = 0.07798
+output = 0.15596
+cache_read = 0.015596
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-flash.toml b/providers/openrouter/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..8296ed361c2
--- /dev/null
+++ b/providers/openrouter/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,16 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "alibaba/qwen3.8-flash"
+temperature = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From bfc2f0e440c0bfa651144aa5192613d0cd6d9ded Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 22:16:53 +0000
Subject: [PATCH 0032/2012] chore(sync): update LLM Gateway model catalog
 (#5593)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../consensusprotocol/deepseek-v4-flash.toml   | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..e5c7e364851
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4-flash"
+name = "DeepSeek V4 Flash (Consensus Protocol)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.13
+output = 0.27
+cache_read = 0.02
+
+[limit]
+context = 524_288
+output = 393_216

From 3f87a632abb314bef2e0d48a339e389c5542bc3d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 23:26:11 +0000
Subject: [PATCH 0033/2012] chore(sync): update Merge Gateway model catalog
 (#5595)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index b4bdd294444..0a7a2254ee8 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index f30ae74cdb3..35e52350151 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 7ce5ab1fb6438543caf0707e20dccb609a8baceb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 23:26:15 +0000
Subject: [PATCH 0034/2012] chore(sync): update OpenRouter model catalog
 (#5596)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index fcb4e40e521..f56ac3460d0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07798
-output = 0.15596
-cache_read = 0.015596
+input = 0.07952
+output = 0.15904
+cache_read = 0.015904
 
 [limit]
 context = 1_048_576

From c7da9f9f932f9c1671a75d0b5678cd5b11f81e1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 02:00:38 +0000
Subject: [PATCH 0035/2012] chore(sync): update Merge Gateway model catalog
 (#5598)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 3 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index faa70b7c1e0..3ef83d21d9d 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 0a7a2254ee8..b4bdd294444 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 35e52350151..f30ae74cdb3 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 9e4a5498e2703fbafe2b23c0e03cd5065c30f5fe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 02:00:44 +0000
Subject: [PATCH 0036/2012] chore(sync): update DigitalOcean model catalog
 (#5600)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../digitalocean/models/openai-gpt-5.6-sol.toml      | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/digitalocean/models/openai-gpt-5.6-sol.toml b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
index 3c9af7df850..e14ff09a4ee 100644
--- a/providers/digitalocean/models/openai-gpt-5.6-sol.toml
+++ b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
@@ -6,15 +6,15 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
+input = 4
+output = 20
+cache_read = 0.4
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 10
-output = 45
-cache_read = 1
+input = 8
+output = 30
+cache_read = 0.8
 
 [modalities]
 input = ["text", "image"]

From ace365e6e299d9173a563f9504c07bb4a28850b2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 02:00:47 +0000
Subject: [PATCH 0037/2012] chore(sync): update Kilo model catalog (#5601)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 6 +++---
 .../kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml      | 2 +-
 providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml  | 9 +++++----
 providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml     | 2 +-
 providers/kilo/models/qwen/qwen3.5-35b-a3b.toml          | 3 ---
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                  | 9 +++++----
 7 files changed, 18 insertions(+), 19 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..e79a3173757 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
index dbe7bd30394..526a9ca0646 100644
--- a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -11,4 +11,4 @@ output = 0.2
 cache_read = 0.025
 
 [limit]
-output = 235_929
+output = 228_000
diff --git a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
index 5441601c6e9..02ea9134945 100644
--- a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -11,12 +11,13 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.25
-output = 0.75
+input = 0.8
+output = 1
+cache_read = 0.4
 
 [limit]
-context = 32_000
-output = 28_800
+context = 128_000
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
index 3f37ded3629..b356342358d 100644
--- a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
@@ -16,7 +16,7 @@ output = 0.598
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index 43f7cfc71e4..96076003f63 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -9,8 +9,5 @@ values = ["none", "high"]
 input = 0.1625
 output = 1.3
 
-[limit]
-output = 235_929
-
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 731fb78b785..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
 
 [limit]
-context = 202_752
+context = 198_000
+output = 16_384

From 95597c9afcaabfb2bec66ed72bcd0d708f23148b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 02:01:54 +0000
Subject: [PATCH 0038/2012] chore(sync): update OpenRouter model catalog
 (#5599)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml |  8 +++++---
 providers/openrouter/models/moonshotai/kimi-k2.5.toml |  6 +++---
 .../models/nvidia/nemotron-3-nano-30b-a3b.toml        |  4 ++--
 .../models/qwen/qwen2.5-vl-72b-instruct.toml          |  7 ++++---
 .../openrouter/models/qwen/qwen3-235b-a22b-2507.toml  |  7 ++++---
 providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml | 11 +++++------
 providers/openrouter/models/tencent/hy3.toml          |  6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml         |  9 ++++++---
 8 files changed, 32 insertions(+), 26 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 55662d5ab50..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-flash-vision-exp"
 description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
 structured_output = false
@@ -10,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.5.toml b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
index 7943f5c3aaa..ebb15bee06b 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.5.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
@@ -10,9 +10,9 @@ field = "reasoning_details"
 type = "toggle"
 
 [cost]
-input = 0.6
-output = 3
-cache_read = 0.1
+input = 0.45
+output = 2.25
+cache_read = 0.07
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
index bf732029012..6469684457d 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -9,7 +9,7 @@ type = "toggle"
 [cost]
 input = 0.05
 output = 0.2
-cache_read = 0.03
+cache_read = 0.025
 
 [limit]
-output = 235_929
+output = 228_000
diff --git a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
index c0bba4820f6..b32e69d51d6 100644
--- a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -12,12 +12,13 @@ knowledge = "2024-06-30"
 open_weights = true
 
 [cost]
-input = 0.25
-output = 0.75
+input = 0.8
+output = 1
+cache_read = 0.4
 
 [limit]
 context = 128_000
-output = 28_800
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index aa706cc488b..0d691163001 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,12 +12,13 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.55
+input = 0.0875
+output = 0.35
+cache_read = 0.0175
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index a1766f7f538..451e8493713 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -1,15 +1,14 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.5-35b-a3b"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.25
-output = 1.25
-cache_read = 0.25
-
-[limit]
-output = 235_929
+input = 0.225
+output = 1.8
+cache_read = 0.225
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384

From b1c44a8a9d6697d34b3be4c6c9ec786426364dc4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 02:01:58 +0000
Subject: [PATCH 0039/2012] chore(sync): update NanoGPT model catalog (#5602)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From a05af20b3529120359803d3a87062dd130c8c663 Mon Sep 17 00:00:00 2001
From: Kendell <KTibow@users.noreply.github.com>
Date: Wed, 26 Aug 2026 19:03:48 -0700
Subject: [PATCH 0040/2012] fix: author Alibaba context pricing tiers (#5597)

* fix: author Qwen3.5 context pricing tiers

Alibaba Cloud Model Studio prices the Qwen3.5 models on a context tier
keyed to input tokens per request, and every token in a request bills at
its tier's unit price. Three catalog entries published a single flat
rate for models that have two bands.

- alibaba-cn qwen3.5-397b-a17b published 0.430/2.58, which is the
  128K<Token<=256K band, as the model's only price. Requests at or below
  128K actually bill 0.172/1.032, so the entry overstated the common
  case by 2.5x. Corrected the base to the lower band and authored the
  upper band as a tier.
- merge-gateway qwen3.5-397b-a17b and qwen3.5-35b-a3b published only the
  lower band. The Gateway API's pricing object carries one input/output
  pair per vendor route and exposes no tier, so the sync cannot derive
  the second band; cost.tiers is preserved from disk across resyncs.

Thinking-mode and non-thinking-mode output are priced identically in
both bands, so alibaba-cn's cost.reasoning tiers alongside cost.output.
No cache-read price is published for the upper band, so the tiers do not
restate one.

Source: https://www.alibabacloud.com/help/en/model-studio/model-pricing
(accessed 2026-08-26)

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* revert: drop merge-gateway Qwen3.5 tiers

The merge-gateway entries for qwen3.5-397b-a17b and qwen3.5-35b-a3b are
not priced off Alibaba Cloud's own route, so Alibaba's context tier
schedule does not apply to them.

Both files carry limit.context = 256_000 and limit.output = 64_000. On
the Gateway's live catalog those are EmpirioLabs' figures; the Alibaba
Cloud route reports a 131K context and a 33K max output. The cache_read
values match the Alibaba Cloud route only because
mergeGatewayCachePricing falls back to existing?.cost?.cache_read when
the selected route exposes none, so they are preserved curated values
rather than evidence of which route was selected.

That leaves the selected route as a reseller mirroring Alibaba's lower
band at a wider context window, with no published tier of its own.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix: author remaining Alibaba context pricing tiers

Sweeping the vendor pricing page for every model whose published rate is
banded turns up six more catalog entries that publish a single flat
price. Same defect as qwen3.5-397b-a17b, same source.

Correct low band, upper bands simply missing:

- alibaba-cn qwen3-coder-480b-a35b-instruct  32K -> 1.291/5.161, 128K -> 2.151/8.602
- alibaba-cn qwen3-coder-30b-a3b-instruct    32K -> 0.323/1.291, 128K -> 0.538/2.151
- alibaba    qwen3-coder-480b-a35b-instruct  32K -> 2.7/13.5,    128K -> 4.5/22.5
- alibaba    qwen3-coder-30b-a3b-instruct    32K -> 0.75/3.75,   128K -> 1.2/6

Wrong band published as the flat rate:

- alibaba-cn glm-5 published 0.86/3.15, which is the 32K<Token<=166K
  band. The entry rate is 0.573/2.58.
- alibaba-cn glm-5.1 published 0.87/3.48, which matches neither band.
  Both bands are taken from the vendor page: 0.825/3.301 up to 32K and
  1.100/3.851 above it. The page publishes no cache-read price for
  either band, so the existing curated cache_read is left on the base
  and not restated on the tier.

glm-4.6 and glm-4.7 are banded on the same page but have no alibaba-cn
entry. glm-5.2 is flat-rate and already correct.

Source: https://www.alibabacloud.com/help/en/model-studio/model-pricing
(accessed 2026-08-26)

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 providers/alibaba-cn/models/glm-5.1.toml          | 15 +++++++++++++--
 providers/alibaba-cn/models/glm-5.toml            | 11 ++++++++++-
 .../models/qwen3-coder-30b-a3b-instruct.toml      | 13 +++++++++++++
 .../models/qwen3-coder-480b-a35b-instruct.toml    | 13 +++++++++++++
 .../alibaba-cn/models/qwen3.5-397b-a17b.toml      | 13 ++++++++++++-
 .../models/qwen3-coder-30b-a3b-instruct.toml      | 13 +++++++++++++
 .../models/qwen3-coder-480b-a35b-instruct.toml    | 13 +++++++++++++
 7 files changed, 87 insertions(+), 4 deletions(-)

diff --git a/providers/alibaba-cn/models/glm-5.1.toml b/providers/alibaba-cn/models/glm-5.1.toml
index 90cb5b7dff2..e19b35f1e84 100644
--- a/providers/alibaba-cn/models/glm-5.1.toml
+++ b/providers/alibaba-cn/models/glm-5.1.toml
@@ -1,3 +1,9 @@
+# Chinese mainland deployment is tiered on input tokens per request; every
+# token in a request bills at its tier's unit price. The 0.87/3.48 previously
+# published here matches neither band; both bands below are taken from the
+# vendor page. No cache-read price is published for either band, so the
+# existing curated cache_read is retained on the base and not restated.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "GLM-5.1"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 family = "glm"
@@ -21,10 +27,15 @@ max = 131_072
 field = "reasoning_content"
 
 [cost]
-input = 0.87
-output = 3.48
+input = 0.825
+output = 3.301
 cache_read = 0.17
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.1
+output = 3.851
+
 [limit]
 context = 202_752
 output = 128_000
diff --git a/providers/alibaba-cn/models/glm-5.toml b/providers/alibaba-cn/models/glm-5.toml
index c2c6aa7f5aa..5d48f4a4ce4 100644
--- a/providers/alibaba-cn/models/glm-5.toml
+++ b/providers/alibaba-cn/models/glm-5.toml
@@ -1,3 +1,7 @@
+# Chinese mainland deployment is tiered on input tokens per request; every
+# token in a request bills at its tier's unit price. The 0.86/3.15 previously
+# published here is the 32K<Token<=166K band, not the model's entry rate.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "GLM-5"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 family = "glm"
@@ -20,8 +24,13 @@ max = 32_768
 field = "reasoning_content"
 
 [cost]
+input = 0.573
+output = 2.58
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
 input = 0.86
-output = 3.15
+output = 3.154
 
 [limit]
 context = 202_752
diff --git a/providers/alibaba-cn/models/qwen3-coder-30b-a3b-instruct.toml b/providers/alibaba-cn/models/qwen3-coder-30b-a3b-instruct.toml
index 0bc3bb15439..682baf133a2 100644
--- a/providers/alibaba-cn/models/qwen3-coder-30b-a3b-instruct.toml
+++ b/providers/alibaba-cn/models/qwen3-coder-30b-a3b-instruct.toml
@@ -1,3 +1,6 @@
+# Chinese mainland deployment is tiered on input tokens per request; every
+# token in a request bills at its tier's unit price.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "Qwen3-Coder 30B-A3B Instruct"
 description = "Qwen coding model for software agents, repository edits, and code reasoning"
 family = "qwen"
@@ -14,6 +17,16 @@ open_weights = true
 input = 0.216
 output = 0.861
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.323
+output = 1.291
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 0.538
+output = 2.151
+
 [limit]
 context = 262_144
 output = 65_536
diff --git a/providers/alibaba-cn/models/qwen3-coder-480b-a35b-instruct.toml b/providers/alibaba-cn/models/qwen3-coder-480b-a35b-instruct.toml
index dd7e726d48f..e216e83d3ca 100644
--- a/providers/alibaba-cn/models/qwen3-coder-480b-a35b-instruct.toml
+++ b/providers/alibaba-cn/models/qwen3-coder-480b-a35b-instruct.toml
@@ -1,3 +1,6 @@
+# Chinese mainland deployment is tiered on input tokens per request; every
+# token in a request bills at its tier's unit price.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "Qwen3-Coder 480B-A35B Instruct"
 description = "Qwen coding model for software agents, repository edits, and code reasoning"
 family = "qwen"
@@ -14,6 +17,16 @@ open_weights = true
 input = 0.861
 output = 3.441
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.291
+output = 5.161
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 2.151
+output = 8.602
+
 [limit]
 context = 262_144
 output = 65_536
diff --git a/providers/alibaba-cn/models/qwen3.5-397b-a17b.toml b/providers/alibaba-cn/models/qwen3.5-397b-a17b.toml
index e93e8e59ec0..8bafd8221fa 100644
--- a/providers/alibaba-cn/models/qwen3.5-397b-a17b.toml
+++ b/providers/alibaba-cn/models/qwen3.5-397b-a17b.toml
@@ -1,3 +1,8 @@
+# Chinese mainland deployment is tiered on input tokens per request:
+# 0<Token<=128K bills at 0.172/1.032 and 128K<Token<=256K at 0.43/2.58, with
+# every token in a request billed at its tier's unit price. Thinking-mode and
+# non-thinking-mode output are priced the same in both bands.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "Qwen3.5 397B-A17B"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
@@ -18,7 +23,13 @@ type = "budget_tokens"
 max = 81_920
 
 [cost]
-input = 0.430
+input = 0.172
+output = 1.032
+reasoning = 1.032
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 0.43
 output = 2.58
 reasoning = 2.58
 
diff --git a/providers/alibaba/models/qwen3-coder-30b-a3b-instruct.toml b/providers/alibaba/models/qwen3-coder-30b-a3b-instruct.toml
index 0f95a1a5c44..4c5078746ba 100644
--- a/providers/alibaba/models/qwen3-coder-30b-a3b-instruct.toml
+++ b/providers/alibaba/models/qwen3-coder-30b-a3b-instruct.toml
@@ -1,3 +1,6 @@
+# International (Singapore) deployment is tiered on input tokens per request;
+# every token in a request bills at its tier's unit price.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "Qwen3-Coder 30B-A3B Instruct"
 description = "Smaller Qwen coder for efficient local agents and repo-level fixes"
 family = "qwen"
@@ -14,6 +17,16 @@ open_weights = true
 input = 0.45
 output = 2.25
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.75
+output = 3.75
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 1.2
+output = 6
+
 [limit]
 context = 262_144
 output = 65_536
diff --git a/providers/alibaba/models/qwen3-coder-480b-a35b-instruct.toml b/providers/alibaba/models/qwen3-coder-480b-a35b-instruct.toml
index 4b2f97ca39d..440ea63f022 100644
--- a/providers/alibaba/models/qwen3-coder-480b-a35b-instruct.toml
+++ b/providers/alibaba/models/qwen3-coder-480b-a35b-instruct.toml
@@ -1,3 +1,6 @@
+# International (Singapore) deployment is tiered on input tokens per request;
+# every token in a request bills at its tier's unit price.
+# https://www.alibabacloud.com/help/en/model-studio/model-pricing (accessed 2026-08-26)
 name = "Qwen3-Coder 480B-A35B Instruct"
 description = "Open Qwen coding heavyweight for repository reasoning and agentic engineering"
 family = "qwen"
@@ -14,6 +17,16 @@ open_weights = true
 input = 1.50
 output = 7.50
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 2.7
+output = 13.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 4.5
+output = 22.5
+
 [limit]
 context = 262_144
 output = 65_536

From b65709869e34b418d6e2362e20847d048154d4fe Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Thu, 27 Aug 2026 10:04:43 +0800
Subject: [PATCH 0041/2012] fix(volcengine): correct the reasoning_effort
 levels and drop redundant toggles (#5564)

Ark accepts none|minimal|low|medium|high|xhigh|max; these entries only
declared four, so max was never reachable. Invalid values like 'banana'
return InvalidParameter, so this is a real accept-set rather than a field
that swallows anything.

Per AGENTS.md an effort list containing 'none' must not be paired with a
toggle, and measurement agrees: sending an effort alone already enables
thinking (reasoning_tokens > 0 with no thinking.type sent), enabled+none
yields 0, and disabled + a graded effort is rejected outright. The two
controls conflict rather than compose.

doubao-seed-character-260628 keeps its toggle: it defaults thinking off
and rejects a bare effort, so thinking.type is the real switch there. Its
effort list omits 'none' accordingly.

Change-Id: I86485c5fd3ed1697fc620f54398d634f3453b188

Co-authored-by: chenyunpeng.1024 <chenyunpeng.1024@bytedance.com>
---
 .../volcengine/models/doubao-seed-1-8-251228.toml      | 10 ++++------
 .../models/doubao-seed-2-0-code-preview-260215.toml    | 10 ++++------
 .../volcengine/models/doubao-seed-2-0-lite-260428.toml | 10 ++++------
 .../volcengine/models/doubao-seed-2-0-mini-260428.toml | 10 ++++------
 .../volcengine/models/doubao-seed-2-0-pro-260215.toml  | 10 ++++------
 .../volcengine/models/doubao-seed-2-1-pro-260628.toml  | 10 ++++------
 .../models/doubao-seed-2-1-turbo-260628.toml           | 10 ++++------
 .../models/doubao-seed-character-260628.toml           |  5 +++--
 providers/volcengine/models/doubao-seed-evolving.toml  | 10 ++++------
 providers/volcengine/provider.toml                     |  4 +++-
 10 files changed, 38 insertions(+), 51 deletions(-)

diff --git a/providers/volcengine/models/doubao-seed-1-8-251228.toml b/providers/volcengine/models/doubao-seed-1-8-251228.toml
index 91169f4d431..94d9546206f 100644
--- a/providers/volcengine/models/doubao-seed-1-8-251228.toml
+++ b/providers/volcengine/models/doubao-seed-1-8-251228.toml
@@ -5,16 +5,14 @@
 # here since the catalog has no output-length dimension.
 # Output tiering note applies to the base [cost] block only.
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-1-8"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-0-code-preview-260215.toml b/providers/volcengine/models/doubao-seed-2-0-code-preview-260215.toml
index d007a3a66d2..75f52056bd8 100644
--- a/providers/volcengine/models/doubao-seed-2-0-code-preview-260215.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-code-preview-260215.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-0-code-preview-260215`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.0-code"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
index 887dffe3576..effe10d6c07 100644
--- a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-0-lite-260428`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.0-lite"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
index 6e198f3f5cb..a2774f2ddd9 100644
--- a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-0-mini-260428`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.0-mini"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-0-pro-260215.toml b/providers/volcengine/models/doubao-seed-2-0-pro-260215.toml
index 61bc3e87899..074981876ac 100644
--- a/providers/volcengine/models/doubao-seed-2-0-pro-260215.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-pro-260215.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-0-pro-260215`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.0-pro"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-1-pro-260628.toml b/providers/volcengine/models/doubao-seed-2-1-pro-260628.toml
index e084bfdaea9..2fd899c3e71 100644
--- a/providers/volcengine/models/doubao-seed-2-1-pro-260628.toml
+++ b/providers/volcengine/models/doubao-seed-2-1-pro-260628.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-1-pro-260628`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.1-pro"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-2-1-turbo-260628.toml b/providers/volcengine/models/doubao-seed-2-1-turbo-260628.toml
index db066348e2b..4480a293892 100644
--- a/providers/volcengine/models/doubao-seed-2-1-turbo-260628.toml
+++ b/providers/volcengine/models/doubao-seed-2-1-turbo-260628.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-2-1-turbo-260628`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-2.1-turbo"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-character-260628.toml b/providers/volcengine/models/doubao-seed-character-260628.toml
index c81590e9420..2bebe7a5463 100644
--- a/providers/volcengine/models/doubao-seed-character-260628.toml
+++ b/providers/volcengine/models/doubao-seed-character-260628.toml
@@ -2,7 +2,8 @@
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
 # Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = minimal|low|medium|high|xhigh|max (no none: this
+# model defaults thinking off, so thinking.type is the real switch here)
 base_model = "bytedance-seed/seed-character"
 
 [[reasoning_options]]
@@ -10,7 +11,7 @@ type = "toggle"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/models/doubao-seed-evolving.toml b/providers/volcengine/models/doubao-seed-evolving.toml
index b86e6b212f4..6b6ba236900 100644
--- a/providers/volcengine/models/doubao-seed-evolving.toml
+++ b/providers/volcengine/models/doubao-seed-evolving.toml
@@ -1,16 +1,14 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-evolving`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# none/minimal disable thinking, so no separate toggle: sending effort alone
+# already enables it, and thinking.type=disabled + a graded effort is rejected.
 base_model = "bytedance-seed/seed-evolving"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/volcengine/provider.toml b/providers/volcengine/provider.toml
index 5879eda36d0..6ca805b6ab9 100644
--- a/providers/volcengine/provider.toml
+++ b/providers/volcengine/provider.toml
@@ -5,7 +5,9 @@ npm = "@ai-sdk/openai-compatible"
 # OpenAI Chat: POST /api/v3/chat/completions; `thinking.type` = enabled|disabled
 # (`auto` is documented but rejected with "Unsupported thinking type for the
 # current model" on every model here except doubao-seed-1-6-flash),
-# `reasoning_effort` = minimal|low|medium|high. Plaintext reasoning is in
+# `reasoning_effort` = none|minimal|low|medium|high|xhigh|max (none and minimal
+# disable thinking; doubao-seed-1-6-251015 accepts only the middle four).
+# Plaintext reasoning is in
 # `choices[].message.reasoning_content`, encrypted block is in
 # `choices[].message.encrypted_content` (returned by default).
 # Responses: POST /api/v3/responses; `reasoning.effort`; encrypted block requires

From 0ee9261efd10c143c9fc3058d78d3efbb816513d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:06:10 -0500
Subject: [PATCH 0042/2012] chore(sync): update Requesty model catalog (#5571)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/glm-5.3-flash.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/requesty/models/glm-5.3-flash.toml

diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..93e48f11df3
--- /dev/null
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.135
+output = 0.45
+cache_read = 0.027
+
+[limit]
+output = 128_000

From e38d4eb96eb7ffe3d29c883a66fd94fa97469c3a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:17:26 -0500
Subject: [PATCH 0043/2012] fix(baseten): track GLM 5.3 Flash reasoning effort
 (#5603)

* fix(baseten): track GLM 5.3 Flash reasoning effort

* fix(baseten): keep reasoning correction catalog-only

---------

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../baseten/models/zai-org/GLM-5.3-Flash.toml | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/baseten/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/baseten/models/zai-org/GLM-5.3-Flash.toml b/providers/baseten/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..55e13e50ab6
--- /dev/null
+++ b/providers/baseten/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,22 @@
+# Effort: reasoning_effort = low|high|max; reasoning cannot be disabled.
+# https://docs.baseten.co/inference/model-apis/reasoning
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text", "image"]

From ef1f8f6a1b5e101d1aa41d414a538fbac320a20e Mon Sep 17 00:00:00 2001
From: sk0x0y <35445665+sk0x0y@users.noreply.github.com>
Date: Thu, 27 Aug 2026 11:18:17 +0900
Subject: [PATCH 0044/2012] Add DeepSeek V4 Pro, Qwen 3.8 27B, and DeepSeek V4
 Flash Flex to neuralwatt (#5561)

* Add DeepSeek V4 Pro to neuralwatt

* Add Qwen 3.8 27B and DeepSeek V4 Flash Flex to neuralwatt, update v4-flash pricing

* Fix cache-read rates to published values, inherit qwen-3.8 output cap

* Use /v1/models advertised cache-read rate; pricing page rounds display

* Verify preview models against authenticated /v1/models; add cache-read and output caps

* Add models API doc link to v4-flash sources

* Update stale 0.5 flex multiplier note in provider.toml
---
 .../models/deepseek-v4-flash-flex.toml        | 28 +++++++++++++
 .../neuralwatt/models/deepseek-v4-flash.toml  | 22 +++++++---
 .../neuralwatt/models/deepseek-v4-pro.toml    | 37 +++++++++++++++++
 providers/neuralwatt/models/qwen-3.8-27b.toml | 40 +++++++++++++++++++
 providers/neuralwatt/provider.toml            |  7 ++--
 5 files changed, 125 insertions(+), 9 deletions(-)
 create mode 100644 providers/neuralwatt/models/deepseek-v4-flash-flex.toml
 create mode 100644 providers/neuralwatt/models/deepseek-v4-pro.toml
 create mode 100644 providers/neuralwatt/models/qwen-3.8-27b.toml

diff --git a/providers/neuralwatt/models/deepseek-v4-flash-flex.toml b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
new file mode 100644
index 00000000000..a69bb2ad43d
--- /dev/null
+++ b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
@@ -0,0 +1,28 @@
+# Flex-tier variant of deepseek-v4-flash: same model, context window
+# (1,048,560), and output cap (65,536) as the standard tier; requested via the
+# `-flex` model ID or service_tier="flex" and requires stream=true
+# (non-streaming requests fall through to the standard tier).
+# Cost is the flex rate: 0.65x of standard pricing (35% off) per the flex-tier
+# guide, applied to the standard rates advertised by GET /v1/models
+# ($0.14/$0.28/$0.028, checked 2026-08-27).
+# Reasoning matches the standard tier: reasoning_effort = none|high|max
+# (default none, reasoning off by default; aliases low/minimal/medium -> high,
+# xhigh -> max). thinking_token_budget is rejected with a 400 on V4-Flash
+# (and -flex), so it is not declared.
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "deepseek/deepseek-v4-flash"
+name = "DeepSeek V4 Flash Flex"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[cost]
+input = 0.091
+output = 0.182
+cache_read = 0.0182
+
+[limit]
+context = 1_048_560
+output = 65_536
diff --git a/providers/neuralwatt/models/deepseek-v4-flash.toml b/providers/neuralwatt/models/deepseek-v4-flash.toml
index 0f6b1623651..dfd88aa481b 100644
--- a/providers/neuralwatt/models/deepseek-v4-flash.toml
+++ b/providers/neuralwatt/models/deepseek-v4-flash.toml
@@ -1,15 +1,25 @@
+# Pricing from GET /v1/models metadata.pricing, the effective billed rate
+# (checked 2026-08-27); the pricing page shows the same rates with cached
+# input rounded to $0.03. Previously listed here at $0.104/$0.207/$0.026.
+# Effort: reasoning_effort = none|high|max (default none, reasoning off by
+# default; aliases low/minimal/medium -> high, xhigh -> max) per the
+# chat-completions per-model table, verified against the live API on
+# 2026-08-26. Off is effort=none, so no separate toggle. thinking_token_budget
+# is rejected with a 400 on this model, so it is not declared.
+# https://portal.neuralwatt.com/docs/api/models
+# https://portal.neuralwatt.com/pricing
+# https://portal.neuralwatt.com/docs/api/chat-completions
 base_model = "deepseek/deepseek-v4-flash"
-name = "DeepSeek V4 Flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["none", "high", "max"]
 
 [cost]
-input = 0.104
-output = 0.207
-cache_read = 0.026
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_048_560
-output = 65_536
\ No newline at end of file
+output = 65_536
diff --git a/providers/neuralwatt/models/deepseek-v4-pro.toml b/providers/neuralwatt/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..a4ec5cb806f
--- /dev/null
+++ b/providers/neuralwatt/models/deepseek-v4-pro.toml
@@ -0,0 +1,37 @@
+# Preview rollout: deepseek-v4-pro is in private preview on Neuralwatt —
+# visible to paid users after requesting access per model card, per the
+# Preview Models guide. Pricing, output cap, and cache-read from the
+# authenticated GET /v1/models metadata (checked 2026-08-27); V4-Pro is not
+# yet in the public catalog or token-pricing table.
+# Reasoning per the chat-completions per-model table (docs verified against
+# the live API on 2026-08-26): reasoning_effort accepts none|low|high|max as
+# distinct levels with low as the default (reasoning on by default); aliases
+# medium/minimal map to low and xhigh maps to high. thinking_token_budget is
+# accepted on this model (unlike deepseek-v4-flash, which rejects it); no
+# bounds or disable sentinel documented.
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/docs/guides/preview-models
+base_model = "deepseek/deepseek-v4-pro"
+status = "beta"
+
+# Effort: reasoning_effort = none|low|high|max (off is effort=none; aliases
+# medium/minimal -> low, xhigh -> high)
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+# Budget: thinking_token_budget (integer reasoning tokens)
+[[reasoning_options]]
+type = "budget_tokens"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.0
+output = 3.0
+cache_read = 0.1
+
+[limit]
+context = 1_048_560
+output = 393_216
diff --git a/providers/neuralwatt/models/qwen-3.8-27b.toml b/providers/neuralwatt/models/qwen-3.8-27b.toml
new file mode 100644
index 00000000000..16289280aec
--- /dev/null
+++ b/providers/neuralwatt/models/qwen-3.8-27b.toml
@@ -0,0 +1,40 @@
+# Preview rollout: qwen-3.8-27b is in preview on Neuralwatt — early-access
+# program, granted via the portal enroll page; absent from the public
+# /v1/models catalog until access is granted.
+# Effort: reasoning_effort = none|low|medium|xhigh (default xhigh, reasoning
+# on by default); aliases max/xhigh/high -> xhigh and minimal -> low, per the
+# chat-completions per-model table (docs verified against the live API on
+# 2026-08-26). Off is effort=none, so no separate toggle.
+# Budget: thinking_token_budget (integer reasoning tokens) is accepted on this
+# model.
+# Neuralwatt serves text+image input only (lab metadata also lists video).
+# Pricing, context, output cap, and effort ladder from the model page and the
+# authenticated GET /v1/models metadata (checked 2026-08-27):
+# https://portal.neuralwatt.com/models/qwen-3.8-27b
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/docs/guides/preview-models
+base_model = "alibaba/qwen3.8-27b"
+status = "beta"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.45
+output = 3.2
+cache_read = 0.25
+
+[limit]
+context = 262_128
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/provider.toml b/providers/neuralwatt/provider.toml
index c7784906cea..c5916a38564 100644
--- a/providers/neuralwatt/provider.toml
+++ b/providers/neuralwatt/provider.toml
@@ -7,10 +7,11 @@ npm = "@ai-sdk/openai-compatible"
 # accepts `reasoning_effort`, with the model-specific normalization documented.
 # Flex tier uses either a `-flex` model ID or `service_tier = "flex"` and
 # requires `stream = true`; non-streaming requests fall through to the standard
-# tier. The official Energy Methodology docs currently state a 0.5 flex pricing
-# multiplier.
+# tier. The flex-tier guide states a 0.65 pricing multiplier (35% off), applied
+# to token billing and energy billing alike (accessed 2026-08-27); earlier
+# flex entries predate the move from 0.5.
 # https://portal.neuralwatt.com/docs/api/chat-completions (accessed 2026-06-25)
-# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/guides/flex-tier (accessed 2026-08-27)
 # https://portal.neuralwatt.com/docs/energy-methodology (accessed 2026-06-28)
 api = "https://api.neuralwatt.com/v1"
 doc = "https://portal.neuralwatt.com/docs"

From a994478f9a7a6d55086966ac72240587e1273093 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 26 Aug 2026 21:44:55 -0500
Subject: [PATCH 0045/2012] fix(edenai): preserve authored reasoning controls
 during sync (#5604)

* fix(edenai): fail before deleting models with unresolved controls

* fix(edenai): retain unresolved controls for manual review

* fix(edenai): preserve authored reasoning controls when unresolved

---------

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/edenai.ts |  17 ++--
 packages/core/test/sync.test.ts            | 109 +++++++++++++++++++--
 sync.md                                    |   8 ++
 3 files changed, 119 insertions(+), 15 deletions(-)

diff --git a/packages/core/src/sync/providers/edenai.ts b/packages/core/src/sync/providers/edenai.ts
index f3aa1566cf2..bedd8be7743 100644
--- a/packages/core/src/sync/providers/edenai.ts
+++ b/packages/core/src/sync/providers/edenai.ts
@@ -3,7 +3,7 @@ import path from "node:path";
 
 import { z } from "zod";
 
-import type { SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
+import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
 import {
   factorBaseModel,
   modelMetadata,
@@ -230,10 +230,10 @@ function regionVariantName(model: EdenAIModel, baseModel: string) {
 // Reasoning options
 // ========================================
 
-// Eden AI's only reasoning control is `reasoning_effort`, so a model is
-// published with the effort list its lab entry (or an established relay peer)
-// already documents. Peers exposing only `toggle` / `budget_tokens` have no
-// equivalent here, and those models are skipped rather than given a guess.
+// This sync currently maps only `reasoning_effort`, using the effort list the
+// lab entry (or an established relay peer) documents. Toggle / budget controls
+// need route-specific mappings. Preserve authored controls when unresolved;
+// skip new models rather than inventing an empty control set.
 function effortValues(options: unknown): string[] | "always-on" | undefined {
   if (!Array.isArray(options)) return undefined;
   if (options.length === 0) return "always-on";
@@ -425,6 +425,7 @@ function mapModalities(values: readonly string[] | null | undefined) {
 
 export function buildEdenAIModel(
   model: EdenAIModel,
+  existing?: ExistingModel,
   firstParty: ReadonlySet<string> = firstPartyBaseModels,
 ): SyncedModel | undefined {
   const baseModel = resolveEdenAIBaseModel(model);
@@ -449,7 +450,7 @@ export function buildEdenAIModel(
   // the lab entry owns it and only the effort controls are authored here.
   const reasoning = modelMetadata(baseModel).reasoning === true;
   const reasoningOptions = reasoning
-    ? reasoningOptionsFor(baseModel)
+    ? reasoningOptionsFor(baseModel) ?? existing?.reasoning_options
     : undefined;
   if (reasoning && reasoningOptions === undefined) return undefined;
 
@@ -510,8 +511,8 @@ export const edenai = {
     firstPartyBaseModels = collectFirstPartyBaseModels(models);
     return models;
   },
-  translateModel(model) {
-    const built = buildEdenAIModel(model);
+  translateModel(model, context) {
+    const built = buildEdenAIModel(model, context.existing(model.id));
     if (built === undefined) return undefined;
     return { id: model.id, model: built };
   },
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 84266188070..692aea87134 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2367,7 +2367,7 @@ test("deduplicates Eden AI case-only IDs without losing context metadata", () =>
   for (const data of [[lowercase, uppercase], [uppercase, lowercase]]) {
     const models = edenai.parseModels({ object: "list", data });
     expect(models).toEqual([{ ...lowercase, context_length: 786_432 }]);
-    expect(edenai.translateModel(models[0]!)).toMatchObject({
+    expect(edenai.translateModel(models[0]!, { existing: () => undefined, authored: () => undefined })).toMatchObject({
       id: lowercase.id,
       model: { base_model: "deepseek/deepseek-v4-flash-0731", limit: { context: 786_432 } },
     });
@@ -2408,9 +2408,8 @@ test("takes Eden AI reasoning options from the model's own lab entry", () => {
   ]);
 });
 
-test("skips Eden AI models whose reasoning control has no effort equivalent", () => {
-  // Lab and OpenRouter both expose these through budget_tokens, which Eden AI
-  // has no request field for.
+test("skips new Eden AI models whose reasoning control has no effort equivalent", () => {
+  // The sync does not yet map this route's budget control to Eden AI's API.
   expect(reasoningOptionsFor("google/gemini-2.5-pro")).toBeUndefined();
   expect(
     buildEdenAIModel(
@@ -2423,6 +2422,102 @@ test("skips Eden AI models whose reasoning control has no effort equivalent", ()
   ).toBeUndefined();
 });
 
+test("Eden AI preserves authored controls when reasoning mapping is unresolved", () => {
+  const authored: NonNullable<ExistingModel["reasoning_options"]>[] = [
+    [],
+    [{ type: "toggle" }],
+    [{ type: "effort", values: ["high"] }],
+    [{ type: "toggle" }, { type: "budget_tokens" }],
+  ];
+  for (const [id, base] of [
+    ["zai/glm-5", "zhipuai/glm-5"],
+    ["moonshot/kimi-k2.6", "moonshotai/kimi-k2.6"],
+    ["minimax/MiniMax-M3", "minimax/MiniMax-M3"],
+    ["deepinfra/nvidia/Nemotron-3-Nano-30B-A3B", "nvidia/nemotron-3-nano-30b-a3b"],
+    ["google/gemini-2.5-pro", "google/gemini-2.5-pro"],
+  ] as const) {
+    const model = edenAIModel({
+      id,
+      owned_by: id.slice(0, id.indexOf("/")),
+      model_name: id.slice(id.indexOf("/") + 1),
+    });
+    expect(buildEdenAIModel(model)).toBeUndefined();
+    for (const reasoning_options of authored) {
+      expect(buildEdenAIModel(model, { base_model: base, reasoning_options })).toMatchObject({
+        base_model: base,
+        reasoning_options,
+      });
+    }
+  }
+});
+
+test("Eden AI sync keeps listed models with unresolved reasoning controls", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "sync-edenai-"));
+  const modelsDir = path.join(root, "providers", "edenai", "models");
+  const repo = path.join(import.meta.dirname, "..", "..", "..");
+  const files = [
+    ["openai/gpt-4o-mini", "openai/gpt-4o-mini"],
+    ["zai/glm-5", "zhipuai/glm-5"],
+    ["retired/model", "openai/gpt-4o-mini"],
+  ] as const;
+
+  try {
+    for (const [id, base] of files) {
+      const destination = path.join(modelsDir, `${id}.toml`);
+      const metadata = path.join(root, "models", `${base}.toml`);
+      await mkdir(path.dirname(destination), { recursive: true });
+      await mkdir(path.dirname(metadata), { recursive: true });
+      await copyFile(path.join(repo, "models", `${base}.toml`), metadata);
+      await copyFile(
+        path.join(repo, "providers", "edenai", "models", `${id === "retired/model" ? "openai/gpt-4o-mini" : id}.toml`),
+        destination,
+      );
+    }
+    const glmPath = path.join(modelsDir, "zai/glm-5.toml");
+    const authored = (await readFile(glmPath, "utf8")).replace(
+      "reasoning_options = []",
+      'reasoning_options = [{ type = "toggle" }]',
+    );
+    const header = "# Toggle: extra_body.thinking.type = enabled|disabled\n";
+    await Bun.write(glmPath, header + authored);
+    const supported = edenAIModel({
+      id: "openai/gpt-4o-mini",
+      model_name: "gpt-4o-mini",
+      owned_by: "openai",
+      list_pricing: { input_cost_per_token: 0.000123, output_cost_per_token: 0.000456 },
+    });
+    const unresolved = edenAIModel({ id: "zai/glm-5", model_name: "glm-5", owned_by: "zai" });
+    const provider = {
+      ...edenai,
+      modelsDir,
+      async fetchModels() {
+        return { object: "list", data: [
+          supported,
+          { ...supported, id: "openai/gpt-4o-mini@us" },
+          unresolved,
+          { ...unresolved, id: "zai/glm-5@us" },
+        ] };
+      },
+    };
+
+    const result = await syncProvider(provider);
+    expect(result).toMatchObject({ created: 1, deleted: 1 });
+    expect(result.files.filter((file) => file.status === "deleted").map((file) => file.path)).toEqual([
+      path.join(modelsDir, "retired/model.toml"),
+    ]);
+    const content = await readFile(glmPath, "utf8");
+    expect(content).toStartWith(header);
+    expect(Bun.TOML.parse(content)).toMatchObject({
+      base_model: "zhipuai/glm-5",
+      reasoning_options: [{ type: "toggle" }],
+    });
+    expect(await Bun.file(path.join(modelsDir, "zai/glm-5@us.toml")).exists()).toBe(false);
+    expect(await Bun.file(path.join(modelsDir, "openai/gpt-4o-mini@us.toml")).exists()).toBe(true);
+  } finally {
+    await rm(root, { recursive: true, force: true });
+  }
+});
+
 test("omits Eden AI reasoning options for non-reasoning models", () => {
   const model = edenAIModel({
     id: "openai/gpt-4o-mini",
@@ -2504,8 +2599,8 @@ test("keeps only the first-party Eden AI route when the lab's own API is relayed
 
   const firstParty = collectFirstPartyBaseModels([bedrock, direct]);
   expect(firstParty).toEqual(new Set(["anthropic/claude-opus-5"]));
-  expect(buildEdenAIModel(bedrock, firstParty)).toBeUndefined();
-  expect(buildEdenAIModel(direct, firstParty)).toMatchObject({
+  expect(buildEdenAIModel(bedrock, undefined, firstParty)).toBeUndefined();
+  expect(buildEdenAIModel(direct, undefined, firstParty)).toMatchObject({
     base_model: "anthropic/claude-opus-5",
   });
 });
@@ -2522,7 +2617,7 @@ test("keeps every Eden AI route for models with no first-party relay", () => {
   const firstParty = collectFirstPartyBaseModels(models);
   expect(firstParty.size).toBe(0);
   for (const model of models) {
-    expect(buildEdenAIModel(model, firstParty)).toMatchObject({
+    expect(buildEdenAIModel(model, undefined, firstParty)).toMatchObject({
       base_model: "openai/gpt-oss-120b",
     });
   }
diff --git a/sync.md b/sync.md
index 9dae6607429..735d20a17d4 100644
--- a/sync.md
+++ b/sync.md
@@ -139,6 +139,14 @@ CI automatically picks up providers registered in `providers` in `packages/core/
 
 Actions are pinned by commit SHA. Keep new workflow actions pinned the same way.
 
+## Eden AI Notes
+
+- Source endpoint: `https://api.edenai.run/v3/models`; no authentication required.
+- Reasoning effort options are derived from the lab's provider entry or OpenRouter. A toggle-only or budget-only control is not an effort list; do not invent effort levels.
+- When the effort mapper cannot resolve controls, preserve the existing route's authored `reasoning_options` while syncing other authoritative fields. Do not replace authored toggle, effort, or budget controls with `[]`.
+- New reasoning models with neither a resolved mapping nor authored controls remain skipped for manual authoring. No empty placeholder is generated, so the normal auto-merge policy remains unchanged; legitimate always-on `[]` entries are not blanket-blocked.
+- Intentional route deduplication and removal of IDs absent from the upstream catalog are unchanged.
+
 ## CrossModel Notes
 
 CrossModel is implemented in `packages/core/src/sync/providers/crossmodel.ts`.

From c5849b5515213b3153c438eac107f34ab6da6cb0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 04:02:04 +0000
Subject: [PATCH 0046/2012] chore(sync): update Kilo model catalog (#5607)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml  | 9 ++++-----
 providers/kilo/models/qwen/qwen3.6-27b.toml              | 2 +-
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml  | 2 +-
 3 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
index 02ea9134945..5441601c6e9 100644
--- a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -11,13 +11,12 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.8
-output = 1
-cache_read = 0.4
+input = 0.25
+output = 0.75
 
 [limit]
-context = 128_000
-output = 115_200
+context = 32_000
+output = 28_800
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index fdb3246f594..7aaf82ee885 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -10,7 +10,7 @@ input = 0.45
 output = 2.7
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index a951250014b..7aa572fb995 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.075
+output = 0.1
 cache_read = 0.007
 
 [limit]

From 6393bd42d8662aab485d8642c62c7623c18380fd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 04:02:07 +0000
Subject: [PATCH 0047/2012] chore(sync): update OpenRouter model catalog
 (#5609)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml  | 7 +++----
 providers/openrouter/models/qwen/qwen3.6-27b.toml        | 9 ++++++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml       | 4 +++-
 3 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
index b32e69d51d6..c0bba4820f6 100644
--- a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -12,13 +12,12 @@ knowledge = "2024-06-30"
 open_weights = true
 
 [cost]
-input = 0.8
-output = 1
-cache_read = 0.4
+input = 0.25
+output = 0.75
 
 [limit]
 context = 128_000
-output = 115_200
+output = 28_800
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index bb7c7a435ef..d35b93066de 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -1,14 +1,17 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.6-27b"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.32
-output = 3.2
+input = 0.6
+output = 3.6
+cache_read = 0.12
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 7bd172e6fed..fe64f45e1df 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "DeepSeek V4 Flash Latest"
 description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
 family = "deepseek"
@@ -19,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.075
+output = 0.1
 cache_read = 0.007
 
 [limit]

From 4d4fa9d596063b9ff2c3dd55ca68f3c1739dd96d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 04:02:09 +0000
Subject: [PATCH 0048/2012] chore(sync): update NanoGPT model catalog (#5610)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 1464b70a207112427826f087c679cdd4bd9e056e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 04:02:10 +0000
Subject: [PATCH 0049/2012] chore(sync): update Merge Gateway model catalog
 (#5608)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 providers/merge-gateway/models/zai/glm-5.3-flash.toml      | 6 +++---
 3 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index b4bdd294444..0a7a2254ee8 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index f30ae74cdb3..35e52350151 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index af6491a5296..b17553c9180 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -7,6 +7,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.012
+output = 0.04
+cache_read = 0.003

From 8b6f297089ed7754b97470ccea7c0002d716ffa2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 06:26:39 +0000
Subject: [PATCH 0050/2012] chore(sync): update Merge Gateway model catalog
 (#5619)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/zai/glm-5.3-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index b17553c9180..af6491a5296 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -7,6 +7,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.012
-output = 0.04
-cache_read = 0.003
+input = 0.075
+output = 0.25
+cache_read = 0.015

From 93179fa985ef25fe7284fdb2e741d715ac99e4dd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 06:26:45 +0000
Subject: [PATCH 0051/2012] chore(sync): update Inceptron model catalog (#5620)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/inceptron/models/moonshotai/Kimi-K2.6.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
index e47b97eabb0..d0ba64678f6 100644
--- a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.54
+input = 0.56
 output = 3.39
-cache_read = 0.13
+cache_read = 0.17
 cache_write = 0
 
 [modalities]

From be0aa6c7b95185a3d7eb77a9723e4cf109a1dab8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 06:26:48 +0000
Subject: [PATCH 0052/2012] chore(sync): update NanoGPT model catalog (#5621)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 0eefb2845cfa9608746ef80c73f5479b6528339a Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 27 Aug 2026 16:18:21 +0800
Subject: [PATCH 0053/2012] deprecate hy4-preview

---
 models/tencent/hy4-preview.toml                      | 1 +
 providers/tencent-token-plan/models/hy4-preview.toml | 1 +
 providers/tencent-tokenhub/models/hy4-preview.toml   | 1 +
 3 files changed, 3 insertions(+)

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
index 19f5c9a36ec..18eccdf5cb7 100644
--- a/models/tencent/hy4-preview.toml
+++ b/models/tencent/hy4-preview.toml
@@ -8,6 +8,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
+status = "deprecated"
 
 [limit]
 context = 1_024_000
diff --git a/providers/tencent-token-plan/models/hy4-preview.toml b/providers/tencent-token-plan/models/hy4-preview.toml
index 2a75bc2426a..fd714b7f789 100644
--- a/providers/tencent-token-plan/models/hy4-preview.toml
+++ b/providers/tencent-token-plan/models/hy4-preview.toml
@@ -6,6 +6,7 @@
 # https://cloud.tencent.com/document/product/1823/130055
 # https://cloud.tencent.com/document/product/1823/131208
 base_model = "tencent/hy4-preview"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/tencent-tokenhub/models/hy4-preview.toml b/providers/tencent-tokenhub/models/hy4-preview.toml
index 2a75bc2426a..fd714b7f789 100644
--- a/providers/tencent-tokenhub/models/hy4-preview.toml
+++ b/providers/tencent-tokenhub/models/hy4-preview.toml
@@ -6,6 +6,7 @@
 # https://cloud.tencent.com/document/product/1823/130055
 # https://cloud.tencent.com/document/product/1823/131208
 base_model = "tencent/hy4-preview"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "toggle"

From 776123c0d1f430a1b739f1205fc862739182b2a7 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 27 Aug 2026 16:23:06 +0800
Subject: [PATCH 0054/2012] fix hy4-preview

---
 models/tencent/hy4-preview.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
index 18eccdf5cb7..19f5c9a36ec 100644
--- a/models/tencent/hy4-preview.toml
+++ b/models/tencent/hy4-preview.toml
@@ -8,7 +8,6 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-status = "deprecated"
 
 [limit]
 context = 1_024_000

From de6db2dcd0e01d14e294d1a2c30934bf4e84c7ac Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 27 Aug 2026 16:39:22 +0800
Subject: [PATCH 0055/2012] drop hy4-preview

---
 models/tencent/hy4-preview.toml               | 18 ----------------
 .../models/hy4-preview.toml                   | 21 -------------------
 .../tencent-tokenhub/models/hy4-preview.toml  | 21 -------------------
 3 files changed, 60 deletions(-)
 delete mode 100644 models/tencent/hy4-preview.toml
 delete mode 100644 providers/tencent-token-plan/models/hy4-preview.toml
 delete mode 100644 providers/tencent-tokenhub/models/hy4-preview.toml

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
deleted file mode 100644
index 19f5c9a36ec..00000000000
--- a/models/tencent/hy4-preview.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-name = "Hy4 preview"
-description = "A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities."
-family = "Hy"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = false
-
-[limit]
-context = 1_024_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/tencent-token-plan/models/hy4-preview.toml b/providers/tencent-token-plan/models/hy4-preview.toml
deleted file mode 100644
index fd714b7f789..00000000000
--- a/providers/tencent-token-plan/models/hy4-preview.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
-# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
-# Catalog is USD / million tokens (×1000).
-# Toggle: thinking.type = enabled|disabled (default disabled)
-# Effort: reasoning_effort = none|high
-# https://cloud.tencent.com/document/product/1823/130055
-# https://cloud.tencent.com/document/product/1823/131208
-base_model = "tencent/hy4-preview"
-status = "deprecated"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.834
-output = 2.501
-cache_read = 0.042
diff --git a/providers/tencent-tokenhub/models/hy4-preview.toml b/providers/tencent-tokenhub/models/hy4-preview.toml
deleted file mode 100644
index fd714b7f789..00000000000
--- a/providers/tencent-tokenhub/models/hy4-preview.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
-# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
-# Catalog is USD / million tokens (×1000).
-# Toggle: thinking.type = enabled|disabled (default disabled)
-# Effort: reasoning_effort = none|high
-# https://cloud.tencent.com/document/product/1823/130055
-# https://cloud.tencent.com/document/product/1823/131208
-base_model = "tencent/hy4-preview"
-status = "deprecated"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.834
-output = 2.501
-cache_read = 0.042

From 5640cf91a73e59e15527591b1b6a98f849a21263 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 10:33:14 +0000
Subject: [PATCH 0056/2012] chore(sync): update Kilo model catalog (#5642)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 providers/kilo/models/qwen/qwen3.5-35b-a3b.toml             | 3 +++
 providers/kilo/models/qwen/qwen3.6-27b.toml                 | 2 +-
 providers/kilo/models/qwen/qwen3.6-35b-a3b.toml             | 4 ++--
 providers/kilo/models/qwen/qwen3.8-flash.toml               | 2 +-
 5 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index e79a3173757..6f876cf5199 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index 96076003f63..43f7cfc71e4 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -9,5 +9,8 @@ values = ["none", "high"]
 input = 0.1625
 output = 1.3
 
+[limit]
+output = 235_929
+
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index 7aaf82ee885..fdb3246f594 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -10,7 +10,7 @@ input = 0.45
 output = 2.7
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index a5bbcc9f543..1385363812a 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.14
-output = 1
+input = 0.1
+output = 0.9
 cache_read = 0.05
 
 [limit]
diff --git a/providers/kilo/models/qwen/qwen3.8-flash.toml b/providers/kilo/models/qwen/qwen3.8-flash.toml
index 16ecba16400..1d0da1780ca 100644
--- a/providers/kilo/models/qwen/qwen3.8-flash.toml
+++ b/providers/kilo/models/qwen/qwen3.8-flash.toml
@@ -7,7 +7,7 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.16
+input = 0.15
 output = 0.47
 cache_read = 0.016
 cache_write = 0.2

From 9863b5cd743920421a9aa43987496153864fee2e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 10:33:17 +0000
Subject: [PATCH 0057/2012] chore(sync): update Inceptron model catalog (#5640)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/inceptron/models/moonshotai/Kimi-K2.7-Code.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/inceptron/models/moonshotai/Kimi-K2.7-Code.toml b/providers/inceptron/models/moonshotai/Kimi-K2.7-Code.toml
index d2d11d2b15c..d633e11afd0 100644
--- a/providers/inceptron/models/moonshotai/Kimi-K2.7-Code.toml
+++ b/providers/inceptron/models/moonshotai/Kimi-K2.7-Code.toml
@@ -6,9 +6,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.67
+input = 0.66
 output = 3.4
-cache_read = 0.19
+cache_read = 0.18
 cache_write = 0
 
 [modalities]

From e576a623841da8150f8e94b048a7737a63bf4a6d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 10:33:20 +0000
Subject: [PATCH 0058/2012] chore(sync): update OpenRouter model catalog
 (#5641)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k2.5.toml    | 6 +++---
 .../openrouter/models/moonshotai/kimi-k2.7-code.toml     | 4 ++--
 providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml    | 9 ++++++---
 providers/openrouter/models/qwen/qwen3.6-27b.toml        | 7 +++----
 providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml    | 6 ++++--
 providers/openrouter/models/qwen/qwen3.8-flash.toml      | 2 +-
 7 files changed, 22 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.5.toml b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
index ebb15bee06b..7943f5c3aaa 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.5.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
@@ -10,9 +10,9 @@ field = "reasoning_details"
 type = "toggle"
 
 [cost]
-input = 0.45
-output = 2.25
-cache_read = 0.07
+input = 0.6
+output = 3
+cache_read = 0.1
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
index 9af80ded38d..09933092d9a 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
@@ -3,9 +3,9 @@ temperature = true
 reasoning_options = []
 
 [cost]
-input = 0.67
+input = 0.66
 output = 3.4
-cache_read = 0.19
+cache_read = 0.18
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index 451e8493713..9c78b7280fd 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -6,9 +6,12 @@ base_model = "alibaba/qwen3.5-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.225
-output = 1.8
-cache_read = 0.225
+input = 0.25
+output = 1.25
+cache_read = 0.25
+
+[limit]
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index d35b93066de..68d7d650159 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,12 +6,11 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.6
-output = 3.6
-cache_read = 0.12
+input = 0.32
+output = 3.2
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index c110dd8430f..ac95b282785 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -1,11 +1,13 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.6-35b-a3b"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.14
-output = 1
+input = 0.1
+output = 0.9
 cache_read = 0.05
 
 [limit]
diff --git a/providers/openrouter/models/qwen/qwen3.8-flash.toml b/providers/openrouter/models/qwen/qwen3.8-flash.toml
index 8296ed361c2..620dd895e63 100644
--- a/providers/openrouter/models/qwen/qwen3.8-flash.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-flash.toml
@@ -10,7 +10,7 @@ type = "toggle"
 type = "budget_tokens"
 
 [cost]
-input = 0.16
+input = 0.15
 output = 0.47
 cache_read = 0.016
 cache_write = 0.2

From b89c2e881c64df4007f23db6335933d02da91536 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 10:34:00 +0000
Subject: [PATCH 0059/2012] chore(sync): update NanoGPT model catalog (#5643)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/TEE/gemma-4-26b-a4b-uncensored.toml     | 4 ++--
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/TEE/gemma-4-26b-a4b-uncensored.toml b/providers/nano-gpt/models/TEE/gemma-4-26b-a4b-uncensored.toml
index ec1f0e921b2..9eb99c7f733 100644
--- a/providers/nano-gpt/models/TEE/gemma-4-26b-a4b-uncensored.toml
+++ b/providers/nano-gpt/models/TEE/gemma-4-26b-a4b-uncensored.toml
@@ -4,7 +4,7 @@ description = "Open Gemma instruction model for efficient chat and self-hosted d
 family = "gemma"
 release_date = "2026-05-23"
 last_updated = "2026-05-23"
-attachment = true
+attachment = false
 reasoning = false
 tool_call = true
 structured_output = true
@@ -21,5 +21,5 @@ input = 65_536
 output = 65_536
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From d85ae6220bbb800f061bd3388025b4855b735654 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:56:34 +0000
Subject: [PATCH 0060/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#5651)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-flash-vision-exp.toml        | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml b/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..e6f23a8f258
--- /dev/null
+++ b/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,13 @@
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_050_000

From 81f5b420aa0298a99d4ffd1cdf3c2c42954e1f69 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:56:42 +0000
Subject: [PATCH 0061/2012] chore(sync): update LLM Gateway model catalog
 (#5650)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-vision-exp.toml   | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..0b1fac9b6a9
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+name = "DeepSeek V4 Flash Vision Exp (DeepSeek)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_050_000
+output = 393_216

From f5e07b0619c985d5e78cf088cd86f75afb57cc55 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:56:47 +0000
Subject: [PATCH 0062/2012] chore(sync): update Venice model catalog (#5652)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/z-ai-glm-5-3-flash.toml | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/providers/venice/models/z-ai-glm-5-3-flash.toml b/providers/venice/models/z-ai-glm-5-3-flash.toml
index dc4c775c583..9c56abf44f5 100644
--- a/providers/venice/models/z-ai-glm-5-3-flash.toml
+++ b/providers/venice/models/z-ai-glm-5-3-flash.toml
@@ -3,12 +3,15 @@ name = "GLM 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
 release_date = "2026-08-21"
 open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.09375
-output = 0.3125
-cache_read = 0.01875
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_048_576

From e3d35506e932c0a8cf6909362bfac205b7077311 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:56:56 +0000
Subject: [PATCH 0063/2012] chore(sync): update Eden AI model catalog (#5588)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../databricks/databricks-gpt-oss-120b@eu.toml      | 13 +++++++++++++
 .../databricks/databricks-gpt-oss-20b@eu.toml       | 13 +++++++++++++
 2 files changed, 26 insertions(+)
 create mode 100644 providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
 create mode 100644 providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml

diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
new file mode 100644
index 00000000000..731a49d8387
--- /dev/null
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
@@ -0,0 +1,13 @@
+base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (EU)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.15001
+output = 0.59997
+cache_read = 0.15001
+cache_write = 0.15001
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
new file mode 100644
index 00000000000..a12ca346593
--- /dev/null
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
@@ -0,0 +1,13 @@
+base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (EU)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.07
+output = 0.30002
+cache_read = 0.07
+cache_write = 0.07

From 695c779a04f18d29fa495ac010d793956eea0825 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:57:25 +0000
Subject: [PATCH 0064/2012] chore(sync): update Charm Hyper model catalog
 (#5653)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.toml                  | 6 +++---
 providers/hyper/models/kimi-k2.5.toml              | 6 +++---
 providers/hyper/models/llama-3.3-70b-instruct.toml | 6 +++---
 providers/hyper/models/minimax-m2.7.toml           | 6 +++---
 providers/hyper/models/qwen3.8-flash.toml          | 2 +-
 5 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 5ce60bdae29..9e565da68e5 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.92
-output = 2.976
-cache_write = 0.46
+input = 0.9
+output = 2.804
+cache_write = 0.45
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index 81e762294c4..debc53b807a 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.544
-output = 2.76
-cache_write = 0.272
+input = 0.5444
+output = 2.855
+cache_write = 0.2722
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/llama-3.3-70b-instruct.toml b/providers/hyper/models/llama-3.3-70b-instruct.toml
index 1620dd2fae7..7ca7f9333ed 100644
--- a/providers/hyper/models/llama-3.3-70b-instruct.toml
+++ b/providers/hyper/models/llama-3.3-70b-instruct.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.638
-output = 0.768
-cache_write = 0.319
+input = 0.6066
+output = 1.0386
+cache_write = 0.3033
 
 [limit]
 output = 12_800
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index de482c56069..8dc18f04455 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.424
-output = 1.612
-cache_write = 0.212
+input = 0.408
+output = 1.512
+cache_write = 0.204
 
 [limit]
 context = 262_100
diff --git a/providers/hyper/models/qwen3.8-flash.toml b/providers/hyper/models/qwen3.8-flash.toml
index abd863768ed..23e537cd5d4 100644
--- a/providers/hyper/models/qwen3.8-flash.toml
+++ b/providers/hyper/models/qwen3.8-flash.toml
@@ -3,7 +3,7 @@ description = "Efficient model for low-latency assistance, extraction, and routi
 reasoning_options = []
 
 [cost]
-input = 0.16
+input = 0.15
 output = 0.47
 cache_read = 0.016
 

From b402aba3172f40c806793abd3f46c350e0c4b9f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:57:36 +0000
Subject: [PATCH 0065/2012] chore(sync): update OpenRouter model catalog
 (#5655)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/qwen/qwen3.6-27b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index 68d7d650159..d35b93066de 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,11 +6,12 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.32
-output = 3.2
+input = 0.6
+output = 3.6
+cache_read = 0.12
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From f4c21ace13b03be89dc8f587491f9326afc2d14f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 14:57:42 +0000
Subject: [PATCH 0066/2012] chore(sync): update Kilo model catalog (#5654)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/minimax/minimax-m2.7:free.toml | 4 +++-
 providers/kilo/models/minimax/minimax-m3:free.toml   | 7 ++++++-
 providers/kilo/models/qwen/qwen3.6-27b.toml          | 2 +-
 3 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/minimax/minimax-m2.7:free.toml b/providers/kilo/models/minimax/minimax-m2.7:free.toml
index a0258eb4533..54ac074faba 100644
--- a/providers/kilo/models/minimax/minimax-m2.7:free.toml
+++ b/providers/kilo/models/minimax/minimax-m2.7:free.toml
@@ -10,7 +10,9 @@ values = ["high"]
 [cost]
 input = 0
 output = 0
+reasoning = 0
+cache_read = 0
 
 [limit]
 context = 196_608
-output = 176_947
+output = 196_608
diff --git a/providers/kilo/models/minimax/minimax-m3:free.toml b/providers/kilo/models/minimax/minimax-m3:free.toml
index d922c671141..1e4add53e23 100644
--- a/providers/kilo/models/minimax/minimax-m3:free.toml
+++ b/providers/kilo/models/minimax/minimax-m3:free.toml
@@ -10,6 +10,11 @@ values = ["none", "high"]
 [cost]
 input = 0
 output = 0
+reasoning = 0
+cache_read = 0
 
 [limit]
-output = 943_718
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index fdb3246f594..7aaf82ee885 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -10,7 +10,7 @@ input = 0.45
 output = 2.7
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From e221c9671b85bb2ebf6b7361e128e2165a5baee8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:23 +0000
Subject: [PATCH 0067/2012] chore(sync): update Inceptron model catalog (#5666)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/inceptron/models/moonshotai/Kimi-K2.6.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
index d0ba64678f6..1d4fb10b7bd 100644
--- a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.56
+input = 0.54
 output = 3.39
-cache_read = 0.17
+cache_read = 0.15
 cache_write = 0
 
 [modalities]

From 85226bf967039da1aa41606b4718431e07f9891e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:27 +0000
Subject: [PATCH 0068/2012] chore(sync): update NanoGPT model catalog (#5664)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/Baichuan-M2.toml    | 24 ----------------
 .../nano-gpt/models/poolside/laguna-m.1.toml  | 12 --------
 .../models/z-ai/glm-5.3-flash-uncensored.toml | 28 +++++++++++++++++++
 3 files changed, 28 insertions(+), 36 deletions(-)
 delete mode 100644 providers/nano-gpt/models/Baichuan-M2.toml
 delete mode 100644 providers/nano-gpt/models/poolside/laguna-m.1.toml
 create mode 100644 providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml

diff --git a/providers/nano-gpt/models/Baichuan-M2.toml b/providers/nano-gpt/models/Baichuan-M2.toml
deleted file mode 100644
index 4e7ccc5c1c8..00000000000
--- a/providers/nano-gpt/models/Baichuan-M2.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Baichuan M2 32B Medical"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "baichuan"
-release_date = "2025-08-19"
-last_updated = "2025-08-19"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 15.73
-output = 15.73
-cache_read = 7.865
-
-[limit]
-context = 32_768
-input = 32_768
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/poolside/laguna-m.1.toml b/providers/nano-gpt/models/poolside/laguna-m.1.toml
deleted file mode 100644
index 4b6640a8a82..00000000000
--- a/providers/nano-gpt/models/poolside/laguna-m.1.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-# Included in subscription
-base_model = "poolside/laguna-m.1"
-reasoning = false
-tool_call = false
-
-[cost]
-input = 0.2
-output = 0.4
-cache_read = 0.1
-
-[limit]
-input = 262_144
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
new file mode 100644
index 00000000000..d6252728002
--- /dev/null
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -0,0 +1,28 @@
+name = "GLM 5.3 Flash Uncensored"
+description = "GLM 5.3 Flash Uncensored is an uncensored fine-tune of the efficient 320B mixture-of-experts reasoning model, built for unrestricted chat, creative writing, coding, agentic work, tool use, and long-context tasks."
+family = "glm"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high"]
+
+[cost]
+input = 0.125
+output = 0.5
+cache_read = 0.0625
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 52772eefe1af0353ff29ea14926c8a126c648728 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:37 +0000
Subject: [PATCH 0069/2012] chore(sync): update OpenRouter model catalog
 (#5663)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731.toml      |  8 ++++--
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../inclusionai/ling-3.0-flash-fin:free.toml  | 28 +++++++++++++++++++
 .../models/mistralai/ministral-8b.toml        | 24 ----------------
 providers/openrouter/models/tencent/hy3.toml  |  6 ++--
 5 files changed, 39 insertions(+), 33 deletions(-)
 create mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-fin:free.toml
 delete mode 100644 providers/openrouter/models/mistralai/ministral-8b.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 5211487ab5f..1273df92943 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-flash-0731"
 
 [[reasoning_options]]
@@ -8,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.12
-cache_read = 0.012
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f56ac3460d0..fcb4e40e521 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07952
-output = 0.15904
-cache_read = 0.015904
+input = 0.07798
+output = 0.15596
+cache_read = 0.015596
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-fin:free.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-fin:free.toml
new file mode 100644
index 00000000000..409dc68f92e
--- /dev/null
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-fin:free.toml
@@ -0,0 +1,28 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ling 3.0 Flash Fin (free)"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "ling"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/mistralai/ministral-8b.toml b/providers/openrouter/models/mistralai/ministral-8b.toml
deleted file mode 100644
index 451ea70512d..00000000000
--- a/providers/openrouter/models/mistralai/ministral-8b.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Ministral 8B"
-description = "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads"
-family = "ministral"
-release_date = "2024-10-17"
-last_updated = "2024-10-17"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = true
-knowledge = "2024-09-30"
-open_weights = false
-
-[cost]
-input = 0.11
-output = 0.11
-
-[limit]
-context = 128_000
-output = 102_400
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 05508ec104a33f69bb65480d850d6554f0960989 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:42 +0000
Subject: [PATCH 0070/2012] chore(sync): update Merge Gateway model catalog
 (#5668)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 providers/merge-gateway/models/zai/glm-5.3-flash.toml      | 6 +++---
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index c680bd0f847..16d15cbb411 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index af6491a5296..3e406e2e86a 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -7,6 +7,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.015
+output = 0.05
+cache_read = 0.003

From 43bd326c8191925e6ee5d3a3c258518184e3268d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:44 +0000
Subject: [PATCH 0071/2012] chore(sync): update Charm Hyper model catalog
 (#5671)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-flash-0731.toml | 8 +-------
 providers/hyper/models/glm-5.2.toml                | 7 +------
 2 files changed, 2 insertions(+), 13 deletions(-)

diff --git a/providers/hyper/models/deepseek-v4-flash-0731.toml b/providers/hyper/models/deepseek-v4-flash-0731.toml
index 5bb961722a3..e539f4b8557 100644
--- a/providers/hyper/models/deepseek-v4-flash-0731.toml
+++ b/providers/hyper/models/deepseek-v4-flash-0731.toml
@@ -1,15 +1,9 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
 release_date = "2026-08-02"
 last_updated = "2026-08-02"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
+reasoning_options = []
 
 [cost]
 input = 0.44
 output = 1.32
 cache_read = 0.044
-
-[limit]
-context = 1_048_576
diff --git a/providers/hyper/models/glm-5.2.toml b/providers/hyper/models/glm-5.2.toml
index bde5c7f646d..19c9bb5d6d5 100644
--- a/providers/hyper/models/glm-5.2.toml
+++ b/providers/hyper/models/glm-5.2.toml
@@ -1,11 +1,10 @@
 base_model = "zhipuai/glm-5.2"
 release_date = "2026-06-30"
 last_updated = "2026-07-22"
-attachment = true
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "high", "max"]
+values = ["high", "xhigh"]
 
 [cost]
 input = 1.52432
@@ -13,8 +12,4 @@ output = 4.79072
 cache_read = 0.152432
 
 [limit]
-context = 1_048_576
 output = 32_768
-
-[modalities]
-input = ["text", "image"]

From 48ab90fc1150da3a9ea93f7a904889a2af31b639 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:55:50 +0000
Subject: [PATCH 0072/2012] chore(sync): update Kilo model catalog (#5669)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inclusionai/ling-3.0-flash-fin:free.toml  | 27 +++++++++++++++++++
 .../kilo/models/mistralai/ministral-8b.toml   | 23 ----------------
 providers/kilo/models/tencent/hy3.toml        |  6 ++---
 3 files changed, 30 insertions(+), 26 deletions(-)
 create mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
 delete mode 100644 providers/kilo/models/mistralai/ministral-8b.toml

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
new file mode 100644
index 00000000000..ecd3d2975fc
--- /dev/null
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
@@ -0,0 +1,27 @@
+name = "Ling 3.0 Flash Fin (free)"
+description = "Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for real-world investment..."
+family = "ling"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/mistralai/ministral-8b.toml b/providers/kilo/models/mistralai/ministral-8b.toml
deleted file mode 100644
index a6f3b6862cf..00000000000
--- a/providers/kilo/models/mistralai/ministral-8b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Mistral: Ministral 8B"
-description = "Ministral 8B is an 8B parameter model featuring a unique interleaved sliding-window attention pattern for faster, memory-efficient inference. Designed for edge use cases, it supports up to 128k context length..."
-family = "ministral"
-release_date = "2024-10-17"
-last_updated = "2024-10-17"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.11
-output = 0.11
-
-[limit]
-context = 128_000
-output = 102_400
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..731fb78b785 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 826d9fe918d9d208791314dd6892f7327c1de5b2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:57:01 +0000
Subject: [PATCH 0073/2012] chore(sync): update Eden AI model catalog (#5667)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 5 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 3fb9086c97e..76bc9a6d3e2 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.758485
-output = 0.758485
+input = 0.756925
+output = 0.756925
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index c1278fe8028..690cc4bdc6c 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.175035
-output = 0.758485
+input = 0.174675
+output = 0.756925
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 027c0f64156..6ff97d9c80d 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46676
-output = 0.933521
+input = 0.4658
+output = 0.9316
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 21734031ada..9896df51c69 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.175035
-output = 0.70014
+input = 0.174675
+output = 0.6987
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 008423b04d1..16fd188db09 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 1.050211
-output = 1.050211
+input = 1.04805
+output = 1.04805

From 14993bf176f238922b694783f68db007c75c19be Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:34:59 +0000
Subject: [PATCH 0074/2012] chore(sync): update Merge Gateway model catalog
 (#5681)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..faa70b7c1e0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -11,8 +12,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 23916d26a397eefe4faa0871367771000e946873 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:35:03 +0000
Subject: [PATCH 0075/2012] chore(sync): update Kilo model catalog (#5680)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    |  4 ++--
 .../kilo/models/thedrummer/rocinante-12b.toml | 22 -------------------
 2 files changed, 2 insertions(+), 24 deletions(-)
 delete mode 100644 providers/kilo/models/thedrummer/rocinante-12b.toml

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 4d6615c6b1e..c977cf138f9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 512_288
-output = 461_059
+context = 262_144
+output = 16_384
diff --git a/providers/kilo/models/thedrummer/rocinante-12b.toml b/providers/kilo/models/thedrummer/rocinante-12b.toml
deleted file mode 100644
index 917b917cd18..00000000000
--- a/providers/kilo/models/thedrummer/rocinante-12b.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "TheDrummer: Rocinante 12B"
-description = "Open-weight instruction model for adaptable chat and self-hosted production workloads"
-release_date = "2024-09-30"
-last_updated = "2024-09-30"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.25
-output = 0.5
-
-[limit]
-context = 65_536
-output = 58_982
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 2de1b25508349b8e10318cf55d57d03ca0a0ca79 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:35:08 +0000
Subject: [PATCH 0076/2012] chore(sync): update OpenRouter model catalog
 (#5679)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    | 12 ++++++----
 .../models/thedrummer/rocinante-12b.toml      | 23 -------------------
 3 files changed, 10 insertions(+), 31 deletions(-)
 delete mode 100644 providers/openrouter/models/thedrummer/rocinante-12b.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index fcb4e40e521..8076cd981ad 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07798
-output = 0.15596
-cache_read = 0.015596
+input = 0.0749
+output = 0.1498
+cache_read = 0.01498
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 41555319289..de4e34c6528 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3-ultra-550b-a55b"
 structured_output = true
 
@@ -12,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 3.6
-cache_read = 0.2
+input = 0.5
+output = 2.2
+cache_read = 0.1
 
 [limit]
-context = 512_288
-output = 461_059
+context = 262_144
+output = 16_384
diff --git a/providers/openrouter/models/thedrummer/rocinante-12b.toml b/providers/openrouter/models/thedrummer/rocinante-12b.toml
deleted file mode 100644
index 911a35b0533..00000000000
--- a/providers/openrouter/models/thedrummer/rocinante-12b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Rocinante 12B"
-description = "Open-weight instruction model for adaptable chat and self-hosted production workloads"
-release_date = "2024-09-30"
-last_updated = "2024-09-30"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = true
-knowledge = "2024-04-30"
-open_weights = true
-
-[cost]
-input = 0.25
-output = 0.5
-
-[limit]
-context = 65_536
-output = 58_982
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 2666ed927ac62a848befd0d8eba31970d17306d2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:35:12 +0000
Subject: [PATCH 0077/2012] chore(sync): update LLM Gateway model catalog
 (#5682)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/iceberg/gemini-3.7-flash.toml             | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml

diff --git a/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml b/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml
new file mode 100644
index 00000000000..a6672029187
--- /dev/null
+++ b/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml
@@ -0,0 +1,12 @@
+base_model = "google/gemini-3.7-flash"
+name = "Gemini 3.7 Flash (Iceberg)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.08333

From bd5b109c65d909cb85586eb6091492e918d8a120 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:35:18 +0000
Subject: [PATCH 0078/2012] chore(sync): update Eden AI model catalog (#5678)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/vertex/gemini-3.1-flash-lite.toml       | 13 +++++++++++++
 .../models/vertex/gemini-3.1-flash-lite@eu.toml    | 14 ++++++++++++++
 .../models/vertex/gemini-3.1-flash-lite@us.toml    | 14 ++++++++++++++
 3 files changed, 41 insertions(+)
 create mode 100644 providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
 create mode 100644 providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
 create mode 100644 providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml

diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
new file mode 100644
index 00000000000..c9c3c4295e6
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
@@ -0,0 +1,13 @@
+base_model = "google/gemini-3.1-flash-lite"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.25
+output = 1.5
+reasoning = 1.5
+cache_read = 0.025
+cache_write = 0.083333
+input_audio = 0.5
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
new file mode 100644
index 00000000000..f5fa68cae40
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.1-flash-lite"
+name = "Gemini 3.1 Flash Lite (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.25
+output = 1.5
+reasoning = 1.5
+cache_read = 0.025
+cache_write = 0.083333
+input_audio = 0.5
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml
new file mode 100644
index 00000000000..2c7d840b683
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.1-flash-lite"
+name = "Gemini 3.1 Flash Lite (US)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.25
+output = 1.5
+reasoning = 1.5
+cache_read = 0.025
+cache_write = 0.083333
+input_audio = 0.5

From 2a47f8241fea8238e400467a15d123e1cf0d3a6d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:35:24 +0000
Subject: [PATCH 0079/2012] chore(sync): update Charm Hyper model catalog
 (#5677)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-flash-0731.toml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/providers/hyper/models/deepseek-v4-flash-0731.toml b/providers/hyper/models/deepseek-v4-flash-0731.toml
index e539f4b8557..f089eb593d4 100644
--- a/providers/hyper/models/deepseek-v4-flash-0731.toml
+++ b/providers/hyper/models/deepseek-v4-flash-0731.toml
@@ -1,7 +1,10 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
 release_date = "2026-08-02"
 last_updated = "2026-08-02"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.44

From d701eedc4c1a473d3736ade29c6cf7528164e3b9 Mon Sep 17 00:00:00 2001
From: Rubin Simons <me@rubin55.org>
Date: Fri, 28 Aug 2026 02:37:10 +0300
Subject: [PATCH 0080/2012] Update neuralwatt catalog (#5636)

* fix(neuralwatt): remove models no longer served

Absent from https://api.neuralwatt.com/v1/models and from the model
list embedded in https://portal.neuralwatt.com/pricing as of
2026-08-27. Drops the Kimi K2.5, Kimi K2.6 and Qwen3.5 lines.

Also drops the HuggingFace-style Kimi-K2.7-Code and Qwen3.6-35B-A3B
ids, which the provider has renamed to kimi-k2.7-code and
qwen3.6-35b; both are re-added under the new ids.

Keeps deepseek-v4-pro and qwen-3.8-27b: both are access-gated
preview models, so their absence from the public catalog is
expected.

* feat(neuralwatt): add models missing from the catalog

Adds kimi-k2.7-code, kimi-k2.7-code-fast, kimi-k3-flex and
qwen3.6-35b. The first and last replace the HuggingFace-style ids
removed in the previous commit.

Pricing, limits, capabilities and reasoning controls come from
metadata on GET https://api.neuralwatt.com/v1/models. Effort lists
are reasoning.supported_efforts; both kimi-k2.7-code variants report
reasoning with no effort control, so they take an empty list.

kimi-k3-flex applies the documented 0.65 flex multiplier.

* fix(neuralwatt): update flex pricing to the 0.65 multiplier

The flex tier moved from 50% off to 35% off (a 0.65 multiplier).
Four files were still deriving flex rates at 0.5.

Rates are now 0.65x the standard rates published by
GET https://api.neuralwatt.com/v1/models, cached input included.
deepseek-v4-flash-flex and kimi-k3-flex were already correct.

https://portal.neuralwatt.com/docs/guides/flex-tier

* fix(neuralwatt): refresh pricing, limits and capabilities

Cached input is published per model as
metadata.pricing.cached_input_per_million; these files derived it as
25% of input instead, overstating it 2.5x on six models. The real
ratio is 10% for GLM, Gemma and Qwen3.6.

The GLM short variants publish max_output_tokens = 32000; they were
using their 199984 context window as the output cap.

Adds explicit structured_output from capabilities.json_mode, and
pins kimi-k2.7-code-flex to NeuralWatt's 262128 context rather than
inheriting the lab entry's 262144.

Source: GET https://api.neuralwatt.com/v1/models

* fix(neuralwatt): correct reasoning capability and controls

Effort lists come from metadata.reasoning.supported_efforts. The GLM
files claimed seven levels where the API supports three; the other
four are aliases it folds into those three, not distinct levels.

glm-5.2-fast, -short-fast and -short-fast-flex were marked
reasoning = false. They report capabilities.reasoning true with
default_effort none, so reasoning is available, just off by default.

Drops toggles on gemma-4-31b and kimi-k3: every effort list here
includes none, which means "off".

thinking_token_budget was probed against the live API. Every model
accepts it except deepseek-v4-flash and gemma-4-31b, which 400 on
the V2 model runner. This adds budget_tokens where it was missing,
including the kimi-k2.7-code variants that previously declared no
controls at all, and corrects kimi-k3, whose comment claimed the
budget was rejected.

* docs(neuralwatt): rewrite README against the live catalog

The model list named the removed K2.5, K2.6 and Qwen3.5 lines and
used the old HuggingFace-style ids. Flex was documented at a 50%
discount rather than the current 0.65 multiplier. Cache reads were
described as a flat 25% of input, but the rate is published per model
and ranges from 10% to 55.6%. Fast variants were called non-reasoning,
which holds for only two of five.

Adds sections on reasoning controls, including the aliasing in
accepted_efforts, and on access-gated preview models, so their
absence from the public catalog is not read as a delisting.

* fix(neuralwatt): correct stale kimi-k3 reasoning header

The leading comment survived the reasoning_options rewrite and
contradicted it. It documented an enable_thinking toggle, an effort
range of low|high|max, and claimed thinking_token_budget was rejected
by the V2 model runner, which are not correct (anymore).

The current effort list includes none as the off switch,
and the budget returns 200 when probed against the live API.

* fix(neuralwatt): declare the reasoning side channel correctly

Neuralwatt returns the reasoning trace on a `reasoning` field, in
streaming deltas and non-streaming messages alike. Probing the live
API shows `reasoning_content` is never populated, so the seven files
naming that field pointed clients at something that does not exist,
and eleven reasoning models declared no channel at all.

The schema's field enum offers only reasoning_content and
reasoning_details, so this uses the bare `interleaved = true`, as
providers/scx-ai does for a host with the same behaviour.

All 18 reasoning models now declare it; kimi-k3-fast and
qwen3.6-35b-fast report no reasoning and declare none.

* fix(neuralwatt): repair split effort comment on deepseek-v4-pro

Inserting interleaved = true above the reasoning_options tables
allowed for only one comment line before the header. The Effort note
on deepseek-v4-pro is two lines, so the key landed between them.

Restores the note as a contiguous block and groups interleaved with
the other top-level keys, matching the three sibling files.
---
 providers/neuralwatt/README.md                | 78 +++++++++++--------
 .../models/Qwen/Qwen3.5-397B-A17B-FP8.toml    | 27 -------
 .../models/Qwen/Qwen3.6-35B-A3B.toml          | 27 -------
 .../models/deepseek-v4-flash-flex.toml        |  1 +
 .../neuralwatt/models/deepseek-v4-flash.toml  |  1 +
 .../neuralwatt/models/deepseek-v4-pro.toml    |  4 +-
 providers/neuralwatt/models/gemma-4-31b.toml  |  5 +-
 providers/neuralwatt/models/glm-5.2-fast.toml | 11 ++-
 providers/neuralwatt/models/glm-5.2-flex.toml | 12 +--
 .../models/glm-5.2-short-fast-flex.toml       | 17 ++--
 .../neuralwatt/models/glm-5.2-short-fast.toml | 13 +++-
 .../neuralwatt/models/glm-5.2-short-flex.toml | 14 ++--
 .../neuralwatt/models/glm-5.2-short.toml      | 10 ++-
 providers/neuralwatt/models/glm-5.2.toml      |  8 +-
 .../neuralwatt/models/kimi-k2.5-fast.toml     | 23 ------
 .../neuralwatt/models/kimi-k2.6-fast.toml     | 23 ------
 .../neuralwatt/models/kimi-k2.6-flex.toml     | 27 -------
 .../models/kimi-k2.7-code-fast.toml           | 19 +++++
 .../models/kimi-k2.7-code-flex.toml           | 14 +++-
 .../neuralwatt/models/kimi-k2.7-code.toml     | 17 ++++
 providers/neuralwatt/models/kimi-k3-flex.toml | 22 ++++++
 providers/neuralwatt/models/kimi-k3.toml      | 27 ++-----
 .../models/moonshotai/Kimi-K2.5.toml          | 27 -------
 .../models/moonshotai/Kimi-K2.6.toml          | 27 -------
 .../models/moonshotai/Kimi-K2.7-Code.toml     | 11 ---
 providers/neuralwatt/models/qwen-3.8-27b.toml |  4 +-
 .../neuralwatt/models/qwen3.5-397b-fast.toml  | 23 ------
 .../neuralwatt/models/qwen3.6-35b-fast.toml   |  3 +-
 providers/neuralwatt/models/qwen3.6-35b.toml  | 21 +++++
 providers/neuralwatt/provider.toml            |  6 +-
 30 files changed, 212 insertions(+), 310 deletions(-)
 delete mode 100644 providers/neuralwatt/models/Qwen/Qwen3.5-397B-A17B-FP8.toml
 delete mode 100644 providers/neuralwatt/models/Qwen/Qwen3.6-35B-A3B.toml
 delete mode 100644 providers/neuralwatt/models/kimi-k2.5-fast.toml
 delete mode 100644 providers/neuralwatt/models/kimi-k2.6-fast.toml
 delete mode 100644 providers/neuralwatt/models/kimi-k2.6-flex.toml
 create mode 100644 providers/neuralwatt/models/kimi-k2.7-code-fast.toml
 create mode 100644 providers/neuralwatt/models/kimi-k2.7-code.toml
 create mode 100644 providers/neuralwatt/models/kimi-k3-flex.toml
 delete mode 100644 providers/neuralwatt/models/moonshotai/Kimi-K2.5.toml
 delete mode 100644 providers/neuralwatt/models/moonshotai/Kimi-K2.6.toml
 delete mode 100644 providers/neuralwatt/models/moonshotai/Kimi-K2.7-Code.toml
 delete mode 100644 providers/neuralwatt/models/qwen3.5-397b-fast.toml
 create mode 100644 providers/neuralwatt/models/qwen3.6-35b.toml

diff --git a/providers/neuralwatt/README.md b/providers/neuralwatt/README.md
index e3e2397c581..f27507bb677 100644
--- a/providers/neuralwatt/README.md
+++ b/providers/neuralwatt/README.md
@@ -8,37 +8,51 @@ Provider Details
 - Environment variable: NEURALWATT_API_KEY
 - Documentation: https://portal.neuralwatt.com/docs
 
-Model Categories
-
-Reasoning Models (with interleaved thinking):
-- glm-5.2 — GLM 5.2, reasoning enabled
-- glm-5.2-short — GLM 5.2 Short, reasoning enabled
-- moonshotai/Kimi-K2.5 — Kimi K2.5, reasoning + image input
-- moonshotai/Kimi-K2.6 — Kimi K2.6, reasoning + image input
-- moonshotai/Kimi-K2.7-Code — Kimi K2.7 Code, reasoning + image input
-- Qwen/Qwen3.5-397B-A17B-FP8 — Qwen3.5 397B, reasoning enabled
-- Qwen/Qwen3.6-35B-A3B — Qwen3.6 35B A3B, reasoning enabled
-
-Fast Variants (optimized for speed, non-reasoning):
-- glm-5.2-fast — GLM 5.2 Fast
-- glm-5.2-short-fast — GLM 5.2 Short Fast
-- kimi-k2.5-fast — Kimi K2.5 Fast, image input
-- kimi-k2.6-fast — Kimi K2.6 Fast, image input
-- qwen3.5-397b-fast — Qwen3.5 397B Fast
-- qwen3.6-35b-fast — Qwen3.6 35B Fast
-
-Flex Variants (streaming required, discounted):
-- glm-5.2-flex — GLM 5.2 Flex, reasoning enabled
-- glm-5.2-short-flex — GLM 5.2 Short Flex, reasoning enabled
-- glm-5.2-short-fast-flex — GLM 5.2 Short Fast Flex
-- kimi-k2.6-flex — Kimi K2.6 Flex, reasoning + image input
-- kimi-k2.7-code-flex — Kimi K2.7 Code Flex, reasoning + image input
+Standard Models
+- deepseek-v4-flash: DeepSeek V4 Flash, 1M context
+- deepseek-v4-pro: DeepSeek V4 Pro, 1M context, private preview
+- gemma-4-31b: Gemma 4 31B, image input, 262K context
+- glm-5.2: GLM 5.2, 1M context
+- glm-5.2-short: GLM 5.2 Short, 200K context, 32K output cap
+- kimi-k2.7-code: Kimi K2.7 Code, image input, 262K context
+- kimi-k3: Kimi K3, image input, 1M context
+- qwen-3.8-27b: Qwen 3.8 27B, image input, 262K context, private preview
+- qwen3.6-35b: Qwen3.6 35B, image input, 131K context
+
+Fast Variants
+Tuned for lower latency. What that means differs per model:
+- glm-5.2-fast, glm-5.2-short-fast: reasoning stays available, but defaults off
+- kimi-k2.7-code-fast: reasoning capped to a short budget and cannot be disabled
+- kimi-k3-fast, qwen3.6-35b-fast: no reasoning
+
+Flex Variants
+Discounted, best-effort latency. Requests may be held under load.
+- deepseek-v4-flash-flex, glm-5.2-flex, glm-5.2-short-flex,
+  glm-5.2-short-fast-flex, kimi-k2.7-code-flex, kimi-k3-flex
+
+Pricing
+- Rates come from metadata.pricing on GET /v1/models. Preview models require an
+  authenticated request to appear.
+- Cached input is published per model as cached_input_per_million. It is 10% of
+  the input price on most models, but 20% on deepseek-v4-flash and 55.6% on
+  qwen-3.8-27b, so read the field rather than applying a ratio.
+- There is no separate cache-write charge.
+- Flex is billed at 0.65x the standard rate (35% off), cached input included.
+- Flex requires stream = true. Non-streaming requests fall through to the
+  standard tier and are billed at standard rates.
+
+Reasoning Controls
+- reasoning_effort accepts the levels in metadata.reasoning.supported_efforts.
+  Every list that offers reasoning also offers none, which is the off switch.
+- accepted_efforts is wider than supported_efforts. The extra values are
+  aliases the API folds into a supported level, not distinct levels.
+- thinking_token_budget is accepted on every model except deepseek-v4-flash and
+  gemma-4-31b, which reject it with a 400 on the V2 model runner.
+- kimi-k2.7-code and its variants expose no effort control, only a budget.
 
 Notes
-- Standard model IDs, pricing, and limits are sourced directly from the Neuralwatt API; flex pricing applies the official 0.5 docs multiplier to the corresponding standard rates
-- Cache reads are billed at 25% of the input token price; there is no separate cache-write charge
-- Neuralwatt provides real-time energy consumption data (Joules/kWh) per request
-- "Fast" variants are optimized for lower latency without reasoning
-- Flex requires streaming; non-streaming requests fall back to the standard tier
-- Official Neuralwatt docs currently describe Flex as a 50% token-pricing discount, including cached-input billing
-- Vision models support image input via OpenAI-compatible API
+- Preview models are access-gated and stay out of the public catalog until
+  access is granted. See https://portal.neuralwatt.com/docs/guides/preview-models
+- Neuralwatt reports real-time energy consumption (Joules/kWh) per request.
+- Models that report no max_output_tokens are capped only by their context
+  window, and are cataloged with output equal to context.
diff --git a/providers/neuralwatt/models/Qwen/Qwen3.5-397B-A17B-FP8.toml b/providers/neuralwatt/models/Qwen/Qwen3.5-397B-A17B-FP8.toml
deleted file mode 100644
index a4790f34439..00000000000
--- a/providers/neuralwatt/models/Qwen/Qwen3.5-397B-A17B-FP8.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Qwen3.5 397B A17B FP8"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2026-02-01"
-last_updated = "2026-02-01"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.69
-output = 4.14
-cache_read = 0.1725
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
\ No newline at end of file
diff --git a/providers/neuralwatt/models/Qwen/Qwen3.6-35B-A3B.toml b/providers/neuralwatt/models/Qwen/Qwen3.6-35B-A3B.toml
deleted file mode 100644
index 17f34727fba..00000000000
--- a/providers/neuralwatt/models/Qwen/Qwen3.6-35B-A3B.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Qwen3.6 35B A3B"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen3.6"
-release_date = "2026-04-01"
-last_updated = "2026-04-01"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.29
-output = 1.15
-cache_read = 0.0725
-
-[limit]
-context = 131_056
-output = 131_056
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
\ No newline at end of file
diff --git a/providers/neuralwatt/models/deepseek-v4-flash-flex.toml b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
index a69bb2ad43d..090d4d2f3f7 100644
--- a/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
+++ b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
@@ -13,6 +13,7 @@
 # https://portal.neuralwatt.com/docs/api/chat-completions
 base_model = "deepseek/deepseek-v4-flash"
 name = "DeepSeek V4 Flash Flex"
+interleaved = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/neuralwatt/models/deepseek-v4-flash.toml b/providers/neuralwatt/models/deepseek-v4-flash.toml
index dfd88aa481b..5d30b1cbdca 100644
--- a/providers/neuralwatt/models/deepseek-v4-flash.toml
+++ b/providers/neuralwatt/models/deepseek-v4-flash.toml
@@ -10,6 +10,7 @@
 # https://portal.neuralwatt.com/pricing
 # https://portal.neuralwatt.com/docs/api/chat-completions
 base_model = "deepseek/deepseek-v4-flash"
+interleaved = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/neuralwatt/models/deepseek-v4-pro.toml b/providers/neuralwatt/models/deepseek-v4-pro.toml
index a4ec5cb806f..7869e9bb888 100644
--- a/providers/neuralwatt/models/deepseek-v4-pro.toml
+++ b/providers/neuralwatt/models/deepseek-v4-pro.toml
@@ -13,6 +13,7 @@
 # https://portal.neuralwatt.com/docs/guides/preview-models
 base_model = "deepseek/deepseek-v4-pro"
 status = "beta"
+interleaved = true
 
 # Effort: reasoning_effort = none|low|high|max (off is effort=none; aliases
 # medium/minimal -> low, xhigh -> high)
@@ -24,9 +25,6 @@ values = ["none", "low", "high", "max"]
 [[reasoning_options]]
 type = "budget_tokens"
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 1.0
 output = 3.0
diff --git a/providers/neuralwatt/models/gemma-4-31b.toml b/providers/neuralwatt/models/gemma-4-31b.toml
index 179f7f33a8d..0da74c962b9 100644
--- a/providers/neuralwatt/models/gemma-4-31b.toml
+++ b/providers/neuralwatt/models/gemma-4-31b.toml
@@ -1,11 +1,12 @@
 base_model = "google/gemma-4-31b-it"
 name = "Gemma 4 31B"
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "effort", values = ["none", "max"] }]
+interleaved = true
 
 [cost]
 input = 0.144
 output = 0.42
-cache_read = 0.036
+cache_read = 0.0144
 
 [limit]
 context = 262_128
diff --git a/providers/neuralwatt/models/glm-5.2-fast.toml b/providers/neuralwatt/models/glm-5.2-fast.toml
index e31157c077d..f43574adc89 100644
--- a/providers/neuralwatt/models/glm-5.2-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-fast.toml
@@ -4,15 +4,22 @@ family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
-reasoning = false
+reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [
+  { type = "effort", values = ["none", "high", "max"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
 input = 1.45
 output = 4.5
-cache_read = 0.3625
+cache_read = 0.145
 
 [limit]
 context = 1_048_560
diff --git a/providers/neuralwatt/models/glm-5.2-flex.toml b/providers/neuralwatt/models/glm-5.2-flex.toml
index 52a7eb9a8d3..47465b4023f 100644
--- a/providers/neuralwatt/models/glm-5.2-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-flex.toml
@@ -5,19 +5,21 @@ release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
 reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
 reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+  { type = "effort", values = ["none", "high", "max"] },
   { type = "budget_tokens" },
 ]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
-input = 0.725
-output = 2.25
-cache_read = 0.18125
+input = 0.9425
+output = 2.925
+cache_read = 0.09425
 
 [limit]
 context = 1_048_560
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
index a3c330a85b6..36151854e7c 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
@@ -4,19 +4,26 @@ family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
-reasoning = false
+reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [
+  { type = "effort", values = ["none", "high", "max"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
-input = 0.725
-output = 2.25
-cache_read = 0.18125
+input = 0.9425
+output = 2.925
+cache_read = 0.09425
 
 [limit]
 context = 199_984
-output = 199_984
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast.toml b/providers/neuralwatt/models/glm-5.2-short-fast.toml
index 52e5c586126..12d05d328dc 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast.toml
@@ -4,19 +4,26 @@ family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
-reasoning = false
+reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [
+  { type = "effort", values = ["none", "high", "max"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
 input = 1.45
 output = 4.5
-cache_read = 0.3625
+cache_read = 0.145
 
 [limit]
 context = 199_984
-output = 199_984
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short-flex.toml b/providers/neuralwatt/models/glm-5.2-short-flex.toml
index c673e3a4141..1e96b70f2ef 100644
--- a/providers/neuralwatt/models/glm-5.2-short-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-flex.toml
@@ -5,23 +5,25 @@ release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
 reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
 reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+  { type = "effort", values = ["none", "high", "max"] },
   { type = "budget_tokens" },
 ]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
-input = 0.725
-output = 2.25
-cache_read = 0.18125
+input = 0.9425
+output = 2.925
+cache_read = 0.09425
 
 [limit]
 context = 199_984
-output = 199_984
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short.toml b/providers/neuralwatt/models/glm-5.2-short.toml
index 729cfd09b69..857cc0d70cd 100644
--- a/providers/neuralwatt/models/glm-5.2-short.toml
+++ b/providers/neuralwatt/models/glm-5.2-short.toml
@@ -5,23 +5,25 @@ release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
 reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
 reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+  { type = "effort", values = ["none", "high", "max"] },
   { type = "budget_tokens" },
 ]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
 input = 1.45
 output = 4.5
-cache_read = 0.3625
+cache_read = 0.145
 
 [limit]
 context = 199_984
-output = 199_984
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2.toml b/providers/neuralwatt/models/glm-5.2.toml
index b500c2bc329..0f3f23cfd75 100644
--- a/providers/neuralwatt/models/glm-5.2.toml
+++ b/providers/neuralwatt/models/glm-5.2.toml
@@ -5,19 +5,21 @@ release_date = "2026-06-17"
 last_updated = "2026-06-17"
 attachment = false
 reasoning = true
+# Budget: thinking_token_budget (integer reasoning tokens)
 reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+  { type = "effort", values = ["none", "high", "max"] },
   { type = "budget_tokens" },
 ]
+interleaved = true
 temperature = true
 tool_call = true
+structured_output = false
 open_weights = true
 
 [cost]
 input = 1.45
 output = 4.5
-cache_read = 0.3625
+cache_read = 0.145
 
 [limit]
 context = 1_048_560
diff --git a/providers/neuralwatt/models/kimi-k2.5-fast.toml b/providers/neuralwatt/models/kimi-k2.5-fast.toml
deleted file mode 100644
index 75da1e5d69f..00000000000
--- a/providers/neuralwatt/models/kimi-k2.5-fast.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Kimi K2.5 Fast"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-01-27"
-last_updated = "2026-01-27"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.52
-output = 2.59
-cache_read = 0.13
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/kimi-k2.6-fast.toml b/providers/neuralwatt/models/kimi-k2.6-fast.toml
deleted file mode 100644
index b5f8319ad65..00000000000
--- a/providers/neuralwatt/models/kimi-k2.6-fast.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Kimi K2.6 Fast"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-04-21"
-last_updated = "2026-04-21"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.69
-output = 3.22
-cache_read = 0.1725
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/kimi-k2.6-flex.toml b/providers/neuralwatt/models/kimi-k2.6-flex.toml
deleted file mode 100644
index f0e4c6f8088..00000000000
--- a/providers/neuralwatt/models/kimi-k2.6-flex.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Kimi K2.6 Flex"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-04-21"
-last_updated = "2026-04-21"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.345
-output = 1.61
-cache_read = 0.08625
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/neuralwatt/models/kimi-k2.7-code-fast.toml b/providers/neuralwatt/models/kimi-k2.7-code-fast.toml
new file mode 100644
index 00000000000..5bb7381dad9
--- /dev/null
+++ b/providers/neuralwatt/models/kimi-k2.7-code-fast.toml
@@ -0,0 +1,19 @@
+base_model = "moonshotai/kimi-k2.7-code"
+name = "Kimi K2.7 Code Fast"
+description = "Kimi K2.7 Code with reasoning capped to a short budget for lower latency; reasoning cannot be disabled on this model"
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [{ type = "budget_tokens" }]
+interleaved = true
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.095
+
+[limit]
+context = 262_128
+output = 262_128
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k2.7-code-flex.toml b/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
index df1d02c4764..cf9cefa44aa 100644
--- a/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
+++ b/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
@@ -1,11 +1,17 @@
 base_model = "moonshotai/kimi-k2.7-code"
 name = "Kimi K2.7 Code Flex"
-reasoning_options = []
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [{ type = "budget_tokens" }]
+interleaved = true
 
 [cost]
-input = 0.475
-output = 2
-cache_read = 0.11875
+input = 0.6175
+output = 2.6
+cache_read = 0.06175
+
+[limit]
+context = 262_128
+output = 262_128
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/neuralwatt/models/kimi-k2.7-code.toml b/providers/neuralwatt/models/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..099d845c043
--- /dev/null
+++ b/providers/neuralwatt/models/kimi-k2.7-code.toml
@@ -0,0 +1,17 @@
+base_model = "moonshotai/kimi-k2.7-code"
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [{ type = "budget_tokens" }]
+interleaved = true
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.095
+
+[limit]
+context = 262_128
+output = 262_128
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k3-flex.toml b/providers/neuralwatt/models/kimi-k3-flex.toml
new file mode 100644
index 00000000000..23752742a08
--- /dev/null
+++ b/providers/neuralwatt/models/kimi-k3-flex.toml
@@ -0,0 +1,22 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 Flex"
+description = "Kimi K3 on the flex tier: discounted, best-effort latency, requests may be held under load"
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [
+  { type = "effort", values = ["none", "low", "high", "max"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
+
+[cost]
+input = 1.95
+output = 9.75
+cache_read = 0.195
+
+[limit]
+context = 1_048_560
+output = 1_048_560
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k3.toml b/providers/neuralwatt/models/kimi-k3.toml
index 0f1ef9c488b..ce4d8d81577 100644
--- a/providers/neuralwatt/models/kimi-k3.toml
+++ b/providers/neuralwatt/models/kimi-k3.toml
@@ -1,24 +1,11 @@
-# Reasoning controls verified against the Neuralwatt live API on 2026-08-05.
-# Toggle: chat_template_kwargs.enable_thinking = true|false — setting false on
-# kimi-k3 drops reasoning to zero chars on a proof task (vs ~1,900 at max).
-# Effort: reasoning_effort = low|high|max — max produces deep reasoning while
-# low/high produce minimal reasoning; matches first-party moonshotai baseline.
-# thinking_token_budget is documented but rejected by the current vLLM V2
-# runner ("not yet supported by the V2 model runner"), so it is not declared.
-# Sources:
-# https://portal.neuralwatt.com/docs/api/chat-completions
-# https://platform.kimi.ai/docs/guide/use-reasoning-effort
+# Effort: reasoning_effort = none|low|high|max (none disables thinking)
+# Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "moonshotai/kimi-k3"
-
-[[reasoning_options]]
-type = "toggle" # API: {"chat_template_kwargs": {"enable_thinking": false}}
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
+reasoning_options = [
+  { type = "effort", values = ["none", "low", "high", "max"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
 
 [cost]
 input = 3.0
diff --git a/providers/neuralwatt/models/moonshotai/Kimi-K2.5.toml b/providers/neuralwatt/models/moonshotai/Kimi-K2.5.toml
deleted file mode 100644
index ca750d34648..00000000000
--- a/providers/neuralwatt/models/moonshotai/Kimi-K2.5.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Kimi K2.5"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-01-27"
-last_updated = "2026-01-27"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.52
-output = 2.59
-cache_read = 0.13
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
\ No newline at end of file
diff --git a/providers/neuralwatt/models/moonshotai/Kimi-K2.6.toml b/providers/neuralwatt/models/moonshotai/Kimi-K2.6.toml
deleted file mode 100644
index 1abe7364f0e..00000000000
--- a/providers/neuralwatt/models/moonshotai/Kimi-K2.6.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Kimi K2.6"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-04-21"
-last_updated = "2026-04-21"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.69
-output = 3.22
-cache_read = 0.1725
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
\ No newline at end of file
diff --git a/providers/neuralwatt/models/moonshotai/Kimi-K2.7-Code.toml b/providers/neuralwatt/models/moonshotai/Kimi-K2.7-Code.toml
deleted file mode 100644
index 021f06e3028..00000000000
--- a/providers/neuralwatt/models/moonshotai/Kimi-K2.7-Code.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "moonshotai/kimi-k2.7-code"
-reasoning_options = []
-
-[cost]
-input = 0.95
-output = 4
-cache_read = 0.2375
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/qwen-3.8-27b.toml b/providers/neuralwatt/models/qwen-3.8-27b.toml
index 16289280aec..baa1ccb1e23 100644
--- a/providers/neuralwatt/models/qwen-3.8-27b.toml
+++ b/providers/neuralwatt/models/qwen-3.8-27b.toml
@@ -15,6 +15,7 @@
 # https://portal.neuralwatt.com/docs/guides/preview-models
 base_model = "alibaba/qwen3.8-27b"
 status = "beta"
+interleaved = true
 
 [[reasoning_options]]
 type = "effort"
@@ -23,9 +24,6 @@ values = ["none", "low", "medium", "xhigh"]
 [[reasoning_options]]
 type = "budget_tokens"
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.45
 output = 3.2
diff --git a/providers/neuralwatt/models/qwen3.5-397b-fast.toml b/providers/neuralwatt/models/qwen3.5-397b-fast.toml
deleted file mode 100644
index 522dca0f9aa..00000000000
--- a/providers/neuralwatt/models/qwen3.5-397b-fast.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen3.5 397B Fast"
-description = "Efficient Qwen model for fast chat, extraction, and high-volume workloads"
-family = "qwen"
-release_date = "2026-02-01"
-last_updated = "2026-02-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = true
-
-[cost]
-input = 0.69
-output = 4.14
-cache_read = 0.1725
-
-[limit]
-context = 262_128
-output = 262_128
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/qwen3.6-35b-fast.toml b/providers/neuralwatt/models/qwen3.6-35b-fast.toml
index 51e31b5fcf6..c0536ac3b67 100644
--- a/providers/neuralwatt/models/qwen3.6-35b-fast.toml
+++ b/providers/neuralwatt/models/qwen3.6-35b-fast.toml
@@ -7,12 +7,13 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = true
 
 [cost]
 input = 0.29
 output = 1.15
-cache_read = 0.0725
+cache_read = 0.029
 
 [limit]
 context = 131_056
diff --git a/providers/neuralwatt/models/qwen3.6-35b.toml b/providers/neuralwatt/models/qwen3.6-35b.toml
new file mode 100644
index 00000000000..3daca1e8ce2
--- /dev/null
+++ b/providers/neuralwatt/models/qwen3.6-35b.toml
@@ -0,0 +1,21 @@
+base_model = "alibaba/qwen3.6-35b-a3b"
+name = "Qwen3.6 35B"
+# Budget: thinking_token_budget (integer reasoning tokens)
+reasoning_options = [
+  { type = "effort", values = ["none", "high"] },
+  { type = "budget_tokens" },
+]
+interleaved = true
+
+[cost]
+input = 0.29
+output = 1.15
+cache_read = 0.029
+
+[limit]
+context = 131_056
+output = 131_056
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/provider.toml b/providers/neuralwatt/provider.toml
index c5916a38564..aae26883ba3 100644
--- a/providers/neuralwatt/provider.toml
+++ b/providers/neuralwatt/provider.toml
@@ -8,8 +8,10 @@ npm = "@ai-sdk/openai-compatible"
 # Flex tier uses either a `-flex` model ID or `service_tier = "flex"` and
 # requires `stream = true`; non-streaming requests fall through to the standard
 # tier. The flex-tier guide states a 0.65 pricing multiplier (35% off), applied
-# to token billing and energy billing alike (accessed 2026-08-27); earlier
-# flex entries predate the move from 0.5.
+# to token billing and energy billing alike (accessed 2026-08-27).
+# The reasoning trace comes back on a `reasoning` field, streaming and
+# non-streaming alike, never `reasoning_content`/`reasoning_details`, so
+# `interleaved` is the bare boolean rather than a field name.
 # https://portal.neuralwatt.com/docs/api/chat-completions (accessed 2026-06-25)
 # https://portal.neuralwatt.com/docs/guides/flex-tier (accessed 2026-08-27)
 # https://portal.neuralwatt.com/docs/energy-methodology (accessed 2026-06-28)

From 1df64ddff1807ee24251a7cbd25b9e5594fdd40e Mon Sep 17 00:00:00 2001
From: hndr <53109993+hndrr@users.noreply.github.com>
Date: Fri, 28 Aug 2026 08:38:05 +0900
Subject: [PATCH 0081/2012] add Qwen3.8 and GLM 5.3 Flash to Modal (#5660)

---
 .../modal/models/Qwen/Qwen3.8-2.4T-A95B.toml  | 25 +++++++++++++++++++
 .../modal/models/zai-org/GLM-5.3-Flash.toml   | 24 ++++++++++++++++++
 2 files changed, 49 insertions(+)
 create mode 100644 providers/modal/models/Qwen/Qwen3.8-2.4T-A95B.toml
 create mode 100644 providers/modal/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/modal/models/Qwen/Qwen3.8-2.4T-A95B.toml b/providers/modal/models/Qwen/Qwen3.8-2.4T-A95B.toml
new file mode 100644
index 00000000000..af307963ce2
--- /dev/null
+++ b/providers/modal/models/Qwen/Qwen3.8-2.4T-A95B.toml
@@ -0,0 +1,25 @@
+# Modal Shared Endpoint model ID: Qwen/Qwen3.8-2.4T-A95B.
+# Pricing is USD per 1M tokens: prompt $2.00, cached prompt $0.25,
+# and completion $6.00.
+# Modal documents text-only input, context up to 1,010,000 tokens, and
+# adjustable reasoning effort with low, medium, and xhigh levels.
+# https://modal.com/library/qwen/qwen3-8-max
+# https://modal.com/blog/qwen3-8-2-4t-a95b-now-available-on-modal
+# https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8-Max"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.0
+output = 6.0
+cache_read = 0.25
+
+[limit]
+context = 1_010_000
diff --git a/providers/modal/models/zai-org/GLM-5.3-Flash.toml b/providers/modal/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..a4d4379375a
--- /dev/null
+++ b/providers/modal/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,24 @@
+# Modal Shared Endpoint model ID: zai-org/GLM-5.3-Flash.
+# Pricing is USD per 1M tokens: prompt $0.45, cached prompt $0.09,
+# and completion $1.50.
+# Modal documents text, image, and video input with a 1M-token context window.
+# Effort: reasoning_effort = low|high|max; defaults to max.
+# https://modal.com/library/zai/glm-5-3-flash
+# https://huggingface.co/zai-org/GLM-5.3-Flash
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.45
+output = 1.5
+cache_read = 0.09
+
+[modalities]
+input = ["text", "image", "video"]

From 7af48014c4b96c27ed4d479d795620c00c69e2c8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E8=BD=BB=E5=B0=98?=
 <53001305+SlightDust@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:38:21 +0800
Subject: [PATCH 0082/2012] feat: add DeepSeek-V4-Pro-0813 to SCNet Token Plan
 (#5662)

---
 .../models/DeepSeek-V4-Pro-0813.toml                | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/scnet-token-plan/models/DeepSeek-V4-Pro-0813.toml

diff --git a/providers/scnet-token-plan/models/DeepSeek-V4-Pro-0813.toml b/providers/scnet-token-plan/models/DeepSeek-V4-Pro-0813.toml
new file mode 100644
index 00000000000..ebcbe85a7e1
--- /dev/null
+++ b/providers/scnet-token-plan/models/DeepSeek-V4-Pro-0813.toml
@@ -0,0 +1,13 @@
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort = high|max
+# https://www.scnet.cn/ac/openapi/doc/2.0/moduleapi/api/chat.html (accessed 2026-08-28)
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0

From d050e33943b9e65d955c36560a1eb64a54680e08 Mon Sep 17 00:00:00 2001
From: bingus-bongo <100236235+bingus-bongo@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:39:20 +0800
Subject: [PATCH 0083/2012] Add TokenGo OpenAI-compatible provider (#5676)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Add TokenGo OpenAI-compatible provider

Catalog TokenGo as a gateway with currentColor logo, provider metadata, and 13 hosted models using lab base_model plus TokenGo USD pricing.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Cite TokenGo pricing sources and drop unverified Qwen budget.

Map each model’s USD rates to the public Model Hub plus console cache ratios, and stop claiming DashScope thinking_budget on this OpenAI-compatible gateway.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Document TokenGo reasoning extra_body wire paths.

Restore the Qwen toggle with enable_thinking and state each model's TokenGo request field so clients can form valid openai-compat reasoning requests.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Align GLM-5.3 Flash hub comment with corrected output.

---------

Co-authored-by: Eddie Zheng <eddiezheng@Eddies-MacBook-Pro.local>
Co-authored-by: Cursor <cursoragent@cursor.com>
---
 providers/tokengo/logo.svg                    |  3 +++
 .../models/deepseek/deepseek-v3.1.toml        | 14 ++++++++++++++
 .../models/deepseek/deepseek-v3.2.toml        | 14 ++++++++++++++
 .../models/deepseek/deepseek-v4-flash.toml    | 19 +++++++++++++++++++
 .../models/deepseek/deepseek-v4-pro.toml      | 18 ++++++++++++++++++
 .../tokengo/models/minimax/minimax-m2.5.toml  | 11 +++++++++++
 .../tokengo/models/moonshotai/kimi-k2.6.toml  | 14 ++++++++++++++
 .../tokengo/models/moonshotai/kimi-k3.toml    | 19 +++++++++++++++++++
 .../models/qwen/qwen3.5-397b-a17b.toml        | 15 +++++++++++++++
 providers/tokengo/models/z-ai/glm-5.1.toml    | 14 ++++++++++++++
 providers/tokengo/models/z-ai/glm-5.2.toml    | 15 +++++++++++++++
 .../tokengo/models/z-ai/glm-5.3-flash.toml    | 15 +++++++++++++++
 providers/tokengo/models/z-ai/glm-5.3.toml    | 15 +++++++++++++++
 providers/tokengo/models/z-ai/glm-5.toml      | 14 ++++++++++++++
 providers/tokengo/provider.toml               | 17 +++++++++++++++++
 15 files changed, 217 insertions(+)
 create mode 100644 providers/tokengo/logo.svg
 create mode 100644 providers/tokengo/models/deepseek/deepseek-v3.1.toml
 create mode 100644 providers/tokengo/models/deepseek/deepseek-v3.2.toml
 create mode 100644 providers/tokengo/models/deepseek/deepseek-v4-flash.toml
 create mode 100644 providers/tokengo/models/deepseek/deepseek-v4-pro.toml
 create mode 100644 providers/tokengo/models/minimax/minimax-m2.5.toml
 create mode 100644 providers/tokengo/models/moonshotai/kimi-k2.6.toml
 create mode 100644 providers/tokengo/models/moonshotai/kimi-k3.toml
 create mode 100644 providers/tokengo/models/qwen/qwen3.5-397b-a17b.toml
 create mode 100644 providers/tokengo/models/z-ai/glm-5.1.toml
 create mode 100644 providers/tokengo/models/z-ai/glm-5.2.toml
 create mode 100644 providers/tokengo/models/z-ai/glm-5.3-flash.toml
 create mode 100644 providers/tokengo/models/z-ai/glm-5.3.toml
 create mode 100644 providers/tokengo/models/z-ai/glm-5.toml
 create mode 100644 providers/tokengo/provider.toml

diff --git a/providers/tokengo/logo.svg b/providers/tokengo/logo.svg
new file mode 100644
index 00000000000..c751e6b5d49
--- /dev/null
+++ b/providers/tokengo/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path fill-rule="evenodd" d="M12 1 23 12 12 23 1 12Zm-4.47 7.44h9.05v2.34h-3.37v5.84h-2.32v-5.84H7.53Z"/>
+</svg>
diff --git a/providers/tokengo/models/deepseek/deepseek-v3.1.toml b/providers/tokengo/models/deepseek/deepseek-v3.1.toml
new file mode 100644
index 00000000000..2f6f5fb3054
--- /dev/null
+++ b/providers/tokengo/models/deepseek/deepseek-v3.1.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking.type = enabled|disabled
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $0.19/$0.71: https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.19/$0.71/$0.06 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same hybrid on/off as typical DeepSeek V3.1 openai-compat peers.
+base_model = "deepseek/deepseek-v3.1"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.19
+output = 0.71
+cache_read = 0.06
diff --git a/providers/tokengo/models/deepseek/deepseek-v3.2.toml b/providers/tokengo/models/deepseek/deepseek-v3.2.toml
new file mode 100644
index 00000000000..a936fe664d5
--- /dev/null
+++ b/providers/tokengo/models/deepseek/deepseek-v3.2.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking.type = enabled|disabled
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub (rounded $0.22/$0.33): https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.2174/$0.326/$0.06 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same hybrid on/off as typical DeepSeek V3.2 openai-compat peers.
+base_model = "deepseek/deepseek-v3.2"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.2174
+output = 0.326
+cache_read = 0.06
diff --git a/providers/tokengo/models/deepseek/deepseek-v4-flash.toml b/providers/tokengo/models/deepseek/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..367e9b40ecd
--- /dev/null
+++ b/providers/tokengo/models/deepseek/deepseek-v4-flash.toml
@@ -0,0 +1,19 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub (rounded $0.10/$0.20): https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.098/$0.196/$0.028 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same controls as typical DeepSeek V4 Flash openai-compat peers.
+base_model = "deepseek/deepseek-v4-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.098
+output = 0.196
+cache_read = 0.028
diff --git a/providers/tokengo/models/deepseek/deepseek-v4-pro.toml b/providers/tokengo/models/deepseek/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..a1361522479
--- /dev/null
+++ b/providers/tokengo/models/deepseek/deepseek-v4-pro.toml
@@ -0,0 +1,18 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub (rounded $0.43/$0.87): https://www.tokengo.com/models
+# - Exact console rates $0.435/$0.87 (no cache ratio listed)
+# TokenGo OpenAI-compatible extra_body; same controls as typical DeepSeek V4 Pro openai-compat peers.
+base_model = "deepseek/deepseek-v4-pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 0.435
+output = 0.87
diff --git a/providers/tokengo/models/minimax/minimax-m2.5.toml b/providers/tokengo/models/minimax/minimax-m2.5.toml
new file mode 100644
index 00000000000..92c384bee18
--- /dev/null
+++ b/providers/tokengo/models/minimax/minimax-m2.5.toml
@@ -0,0 +1,11 @@
+# Always-on thinking; no TokenGo on/off or effort field (same as typical MiniMax-M2.5 peers).
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $0.30/$1.20: https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.30/$1.20/$0.03 (TokenGo model table ratios)
+base_model = "minimax/MiniMax-M2.5"
+reasoning_options = []
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.03
diff --git a/providers/tokengo/models/moonshotai/kimi-k2.6.toml b/providers/tokengo/models/moonshotai/kimi-k2.6.toml
new file mode 100644
index 00000000000..e41d4f4bbdf
--- /dev/null
+++ b/providers/tokengo/models/moonshotai/kimi-k2.6.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking.type = enabled|disabled
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $0.95/$4.00: https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.95/$4.00/$0.16 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same on/off as typical Kimi K2.6 openai-compat peers.
+base_model = "moonshotai/kimi-k2.6"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.95
+output = 4.0
+cache_read = 0.16
diff --git a/providers/tokengo/models/moonshotai/kimi-k3.toml b/providers/tokengo/models/moonshotai/kimi-k3.toml
new file mode 100644
index 00000000000..c7bb5114efd
--- /dev/null
+++ b/providers/tokengo/models/moonshotai/kimi-k3.toml
@@ -0,0 +1,19 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $3.00/$15.00: https://www.tokengo.com/models
+# - Exact console rates incl. cache $3.00/$15.00/$0.30 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same controls as typical Kimi K3 openai-compat peers.
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 3.0
+output = 15.0
+cache_read = 0.3
diff --git a/providers/tokengo/models/qwen/qwen3.5-397b-a17b.toml b/providers/tokengo/models/qwen/qwen3.5-397b-a17b.toml
new file mode 100644
index 00000000000..75c9d1d86f2
--- /dev/null
+++ b/providers/tokengo/models/qwen/qwen3.5-397b-a17b.toml
@@ -0,0 +1,15 @@
+# Toggle: enable_thinking true|false
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $0.40/$2.65: https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.40/$2.65/$0.20 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same on/off control as typical Qwen relays.
+# No thinking_budget on this host (generic relays omit DashScope budget).
+base_model = "alibaba/qwen3.5-397b-a17b"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.4
+output = 2.65
+cache_read = 0.2
diff --git a/providers/tokengo/models/z-ai/glm-5.1.toml b/providers/tokengo/models/z-ai/glm-5.1.toml
new file mode 100644
index 00000000000..17bf4336a01
--- /dev/null
+++ b/providers/tokengo/models/z-ai/glm-5.1.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking.type = enabled|disabled
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $1.40/$4.40: https://www.tokengo.com/models
+# - Exact console rates incl. cache $1.40/$4.40/$0.26 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same on/off as typical GLM-5.1 openai-compat peers.
+base_model = "zhipuai/glm-5.1"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/tokengo/models/z-ai/glm-5.2.toml b/providers/tokengo/models/z-ai/glm-5.2.toml
new file mode 100644
index 00000000000..2ab99ffa116
--- /dev/null
+++ b/providers/tokengo/models/z-ai/glm-5.2.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $1.40/$4.40: https://www.tokengo.com/models
+# - Exact console rates incl. cache $1.40/$4.40/$0.26 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same effort set as typical GLM-5.2 openai-compat peers.
+base_model = "zhipuai/glm-5.2"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/tokengo/models/z-ai/glm-5.3-flash.toml b/providers/tokengo/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..ed041093586
--- /dev/null
+++ b/providers/tokengo/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = low|high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub (rounded $0.07/$0.02): https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.075/$0.025/$0.015 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same effort set as typical GLM-5.3-Flash openai-compat peers.
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.025
+cache_read = 0.015
diff --git a/providers/tokengo/models/z-ai/glm-5.3.toml b/providers/tokengo/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..9f49bfaa96d
--- /dev/null
+++ b/providers/tokengo/models/z-ai/glm-5.3.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = low|high|max
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $1.40/$4.40: https://www.tokengo.com/models
+# - Exact console rates incl. cache $1.40/$4.40/$0.26 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same effort set as typical GLM-5.3 openai-compat peers.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/tokengo/models/z-ai/glm-5.toml b/providers/tokengo/models/z-ai/glm-5.toml
new file mode 100644
index 00000000000..fd48f0c5225
--- /dev/null
+++ b/providers/tokengo/models/z-ai/glm-5.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking.type = enabled|disabled
+# Pricing (USD/MTok, accessed 2026-08-27):
+# - Public hub $0.89/$3.26: https://www.tokengo.com/models
+# - Exact console rates incl. cache $0.89/$3.2647/$0.2226 (TokenGo model table ratios)
+# TokenGo OpenAI-compatible extra_body; same on/off as typical GLM-5 openai-compat peers.
+base_model = "zhipuai/glm-5"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.89
+output = 3.2647
+cache_read = 0.2226
diff --git a/providers/tokengo/provider.toml b/providers/tokengo/provider.toml
new file mode 100644
index 00000000000..a73c9b417b1
--- /dev/null
+++ b/providers/tokengo/provider.toml
@@ -0,0 +1,17 @@
+# TokenGo is an OpenAI-compatible multi-model gateway (POST /v1/chat/completions).
+# Sources (accessed 2026-08-27), and what each supports:
+# - https://www.tokengo.com/docs/documentation/getting-started/introduction
+#   Unified OpenAI-compatible API; pay-as-you-go USD per million tokens; Model Hub for rates.
+# - https://www.tokengo.com/docs/documentation/getting-started/quickstart
+#   Chat completions base shape (model + messages).
+# - https://www.tokengo.com/models
+#   Public Model Hub input/output list prices (some cards round). Cache is not shown.
+# Exact USD/MTok including cache_read come from TokenGo's first-party console model
+# table (model/completion/cache ratios) accessed 2026-08-27.
+# Reasoning extra_body matches typical openai-compat peers per model (thinking.type /
+# reasoning_effort / enable_thinking). See each model file for the exact wire path.
+name = "TokenGo"
+env = ["TOKENGO_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+doc = "https://www.tokengo.com/docs"
+api = "https://api.tokengo.com/v1"

From 4fe69cc5856a861b9b942d4f2efe1a6b25c12da8 Mon Sep 17 00:00:00 2001
From: Jaber Jaber <103749727+jaberjaber23@users.noreply.github.com>
Date: Fri, 28 Aug 2026 02:39:35 +0300
Subject: [PATCH 0084/2012] feat(runinfra): add GLM 5.3 Flash and Ornith 1.5
 35B A3B, correct the Nemotron cached price (#5674)

* feat(runinfra): add GLM 5.3 Flash and Ornith 1.5 35B A3B

GLM 5.3 Flash overrides the existing zhipuai/glm-5.3-flash lab entry with the
context window, output budget, prices and input modalities this deployment was
launched with. Ornith 1.5 35B A3B is self-contained because no Ornith 1.5 lab
entry exists upstream, matching the shape nano-gpt already uses.

Every reasoning, caching and capability claim was probed live on 2026-08-27 at
temperature 0 with three repeats per level; the evidence is in each file comment.
GLM distinguishes only low and high and refuses to disable reasoning, which its
own 400 body states. Ornith is an on/off toggle: none completes in 4 tokens
against 78 for every other value including the default.

* fix(runinfra): publish the Nemotron cached input price

Prefix caching is live on this serve and priced at $0.01 per million cached
input tokens, but the entry carried no cache_read, so the catalog understated
what the model costs to run with a repeated prefix. Confirmed by probe rather
than declaration: an identical 4,200-token prompt repeated 6 times reported
cached_tokens 0 then 4,352 on every attempt after the first.

* fix(ornith): add the Ornith 1.5 lab entry and make the provider file override-only

The review bot is right. RunInfra did not create this checkpoint, it is nameable
on Hugging Face and already hosted elsewhere, so a full inline provider
definition skipped the required lab entry. Matching the shape another provider
already uses is not an exemption.

Ornith 1.5 35B A3B now has a lab file under deepreinforce, the lab that makes
the Ornith family and already carries the 1.0 variants. Model facts come from
the Hugging Face repo ornith-ai/Ornith-1.5-35B-A3B read 2026-08-27: created
2026-08-18, last modified 2026-08-23, license mit, tagged image-text-to-text
and conversational.

The provider file is now base_model plus real deltas only: prices, the reasoning
toggle, the interleaved field, and the output budget this deployment was launched
with. Context and modalities match the lab entry and are no longer restated.
---
 models/deepreinforce/ornith-1.5-35b-a3b.toml  | 28 ++++++++++++++
 ...A-Nemotron-3.5-Lightning-30B-A3B-BF16.toml |  4 ++
 .../models/ornith-ai/Ornith-1.5-35B-A3B.toml  | 26 +++++++++++++
 .../models/zai-org/GLM-5.3-Flash.toml         | 38 +++++++++++++++++++
 4 files changed, 96 insertions(+)
 create mode 100644 models/deepreinforce/ornith-1.5-35b-a3b.toml
 create mode 100644 providers/runinfra/models/ornith-ai/Ornith-1.5-35B-A3B.toml
 create mode 100644 providers/runinfra/models/zai-org/GLM-5.3-Flash.toml

diff --git a/models/deepreinforce/ornith-1.5-35b-a3b.toml b/models/deepreinforce/ornith-1.5-35b-a3b.toml
new file mode 100644
index 00000000000..ff436d1bf5d
--- /dev/null
+++ b/models/deepreinforce/ornith-1.5-35b-a3b.toml
@@ -0,0 +1,28 @@
+name = "Ornith 1.5 35B A3B"
+description = "Mixture-of-experts coding-reasoning model for agentic software tasks, tool use, and image understanding"
+family = "ornith"
+release_date = "2026-08-18"
+last_updated = "2026-08-23"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+license = "MIT"
+
+[limit]
+context = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B"
+
+[[links]]
+label = "Model card"
+url = "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B"
+type = "model_card"
diff --git a/providers/runinfra/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.toml b/providers/runinfra/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.toml
index 2cc8095c64c..b84500c0d2b 100644
--- a/providers/runinfra/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.toml
+++ b/providers/runinfra/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.toml
@@ -6,6 +6,9 @@
 # own run-to-run variance (baseline 570/642 tokens across repeats, medium 588/642), so no graded
 # control is claimed. enable_thinking / chat_template_kwargs / thinking_budget are accepted but
 # have no effect. The control is a toggle.
+# cache_read is the cached input price, added 2026-08-27 after prefix caching was confirmed on
+# this serve rather than assumed: an identical 4,200-token prompt repeated 6 times reported
+# cached_tokens 0 on the first attempt and 4,352 on all five that followed.
 base_model = "nvidia/nemotron-3.5-lightning"
 
 [[reasoning_options]]
@@ -14,6 +17,7 @@ type = "toggle"
 [cost]
 input = 0.05
 output = 0.15
+cache_read = 0.01
 
 [limit]
 output = 32_768
diff --git a/providers/runinfra/models/ornith-ai/Ornith-1.5-35B-A3B.toml b/providers/runinfra/models/ornith-ai/Ornith-1.5-35B-A3B.toml
new file mode 100644
index 00000000000..5323a7bc9ae
--- /dev/null
+++ b/providers/runinfra/models/ornith-ai/Ornith-1.5-35B-A3B.toml
@@ -0,0 +1,26 @@
+# Source: https://runinfra.ai/inference-api/ornith-1-5-35b
+# Reasoning control, probed live on this serve 2026-08-27 (temperature 0, fixed prompt,
+# 3 repeats per level, reproducible to the token): reasoning_effort = "none" completes in 4
+# tokens while the omitted default completes in 78, and "low", "medium", "high", "xhigh" and
+# "max" all complete in 78, identical to the default. So the control is an on/off toggle and
+# no graded ladder is claimed. Values outside the accepted set are refused (probed "bogus": 400).
+# Reasoning streams on a "reasoning_content" delta field (probed live).
+# cache_read is the cached input price and prefix caching is real, not declared: an identical
+# 4,200-token prompt repeated 6 times reported cached_tokens 0 then 6,976 on every later attempt.
+# The output budget is what THIS deployment was launched with; the context window matches the
+# lab entry and is therefore not restated here.
+base_model = "deepreinforce/ornith-1.5-35b-a3b"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.10
+output = 0.40
+cache_read = 0.01
+
+[limit]
+output = 32_768
diff --git a/providers/runinfra/models/zai-org/GLM-5.3-Flash.toml b/providers/runinfra/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..36f839e5808
--- /dev/null
+++ b/providers/runinfra/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,38 @@
+# Source: https://runinfra.ai/inference-api/glm-5-3-flash
+# Reasoning control, probed live on this serve 2026-08-27 (temperature 0, fixed prompt,
+# 3 repeats per level, completion tokens reproducible to the token at every level):
+# "low" 4, "high" 12, and "medium" / "xhigh" / "max" / the omitted default all 30. So only
+# "low" and "high" are distinct rungs; every other accepted value means maximum effort.
+# The serve states this itself when reasoning is switched off: reasoning_effort = "none"
+# returns HTTP 400 'reasoning_effort "none" is not supported: GLM 5.3 Flash cannot disable
+# thinking, and this serve treats every value outside "low" and "high" as maximum effort.'
+# Values outside the accepted set are refused rather than folded (probed "ultra", "bogus": 400).
+# Reasoning is interleaved into the stream, on a "reasoning" delta field rather than
+# "reasoning_content" (probed live). interleaved is declared as the bare boolean because the
+# schema's field enum covers only "reasoning_content" and "reasoning_details", and naming
+# either of those here would misdescribe what this serve actually emits.
+# Tool calling and structured output verified live the same day: a get_weather call
+# round-tripped with correct arguments, json_object parsed, and a strict json_schema validated.
+# cache_read is the cached input price and prefix caching is real, not declared: an identical
+# 4,200-token prompt repeated 6 times reported cached_tokens 0 then 5,760 on every later attempt.
+# context and output are what THIS deployment was launched with, both below the lab entry's
+# figures. Input modalities list only text and image because those are the two we serve and
+# proved; the lab entry's video and pdf are not claimed here.
+base_model = "zhipuai/glm-5.3-flash"
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.10
+output = 0.40
+cache_read = 0.01
+
+[limit]
+context = 1_048_576
+output = 32_768
+
+[modalities]
+input = ["text", "image"]

From 154531a871801fb1c3a0476834525c62f25fe8e3 Mon Sep 17 00:00:00 2001
From: Raphael Palma <117413668+raphaelgpalma@users.noreply.github.com>
Date: Thu, 27 Aug 2026 20:39:42 -0300
Subject: [PATCH 0085/2012] feat(ollama-cloud): add glm-5.3-flash (#5675)

GLM-5.3-Flash (320B-A18B, MIT, released 2026-08-26) is served on Ollama Cloud as glm-5.3-flash:cloud. Inherits shared metadata from zhipuai/glm-5.3-flash; overrides open_weights (weights are public on HuggingFace) and adds the effort-based reasoning options and reasoning_content interleave that Ollama's OpenAI-compatible endpoint returns. Verified against https://ollama.com/v1.
---
 providers/ollama-cloud/models/glm-5.3-flash.toml | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/ollama-cloud/models/glm-5.3-flash.toml

diff --git a/providers/ollama-cloud/models/glm-5.3-flash.toml b/providers/ollama-cloud/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..fa66dc84ecc
--- /dev/null
+++ b/providers/ollama-cloud/models/glm-5.3-flash.toml
@@ -0,0 +1,9 @@
+base_model = "zhipuai/glm-5.3-flash"
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"

From 417db2332192c8843122c873d5a63f202e4b1019 Mon Sep 17 00:00:00 2001
From: Zain Hasan <zain_has@hotmail.com>
Date: Thu, 27 Aug 2026 19:40:11 -0400
Subject: [PATCH 0086/2012] [Together AI] add glm 5.3 flash (#5634)

---
 .../models/zai-org/GLM-5.3-Flash.toml           | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/togetherai/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/togetherai/models/zai-org/GLM-5.3-Flash.toml b/providers/togetherai/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..5b8b80ebcb7
--- /dev/null
+++ b/providers/togetherai/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.50
+cache_read = 0.03
+
+[limit]
+context = 1_048_575
+output = 400_000
+
+[modalities]
+input = ["text", "image", "video"]

From 03d0e09da3fd42a5828690112fb10c5884038b1c Mon Sep 17 00:00:00 2001
From: Andrew Szeto <github@jabagawee.com>
Date: Thu, 27 Aug 2026 16:40:39 -0700
Subject: [PATCH 0087/2012] feat(github-copilot): Add sync provider for GitHub
 Copilot (#5623)

* feat(github-copilot): sync pricing from the github/docs pricing table

GitHub Copilot has no pricing API, but the docs pricing page is
rendered from a YAML data file in the public github/docs repo, so this
module syncs [cost] from that file and leaves everything else in the
TOMLs hand-authored. Copilot pricing has needed manual updates a few times
recently (#3282, #3965, #5186), so this should help automate that.

* chore(sync): update GitHub Copilot model catalog
---
 packages/core/src/sync/index.ts               |   5 +-
 .../core/src/sync/providers/github-copilot.ts | 163 +++++++++++++
 packages/core/test/github-copilot.test.ts     | 228 ++++++++++++++++++
 .../models/gemini-3.6-flash.toml              |  17 +-
 .../github-copilot/models/gpt-5.6-luna.toml   |  16 +-
 .../github-copilot/models/gpt-5.6-sol.toml    |  26 +-
 .../github-copilot/models/gpt-5.6-terra.toml  |  20 +-
 sync.md                                       |  10 +
 8 files changed, 453 insertions(+), 32 deletions(-)
 create mode 100644 packages/core/src/sync/providers/github-copilot.ts
 create mode 100644 packages/core/test/github-copilot.test.ts

diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 6a72c914858..c5f116f377a 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -16,6 +16,7 @@ import { deepinfra } from "./providers/deepinfra.js";
 import { digitalocean } from "./providers/digitalocean.js";
 import { edenai } from "./providers/edenai.js";
 import { empiriolabs } from "./providers/empiriolabs.js";
+import { githubCopilot } from "./providers/github-copilot.js";
 import { google } from "./providers/google.js";
 import { hyper } from "./providers/hyper.js";
 import { huggingface } from "./providers/huggingface.js";
@@ -135,6 +136,7 @@ export const providers: {
   digitalocean: SyncProvider<any>;
   edenai: SyncProvider<any>;
   empiriolabs: SyncProvider<any>;
+  "github-copilot": SyncProvider<any>;
   google: SyncProvider<any>;
   hyper: SyncProvider<any>;
   huggingface: SyncProvider<any>;
@@ -167,6 +169,7 @@ export const providers: {
   digitalocean,
   edenai,
   empiriolabs,
+  "github-copilot": githubCopilot,
   google,
   hyper,
   huggingface,
@@ -207,7 +210,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "google", "hyper", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
diff --git a/packages/core/src/sync/providers/github-copilot.ts b/packages/core/src/sync/providers/github-copilot.ts
new file mode 100644
index 00000000000..5cd810e0f1e
--- /dev/null
+++ b/packages/core/src/sync/providers/github-copilot.ts
@@ -0,0 +1,163 @@
+import { z } from "zod";
+
+import type { ExistingModel, SyncProvider, SyncedModel } from "../index.js";
+
+const PRICING_ENDPOINT = "https://raw.githubusercontent.com/github/docs/main/data/tables/copilot/models-and-pricing.yml";
+
+const NOT_APPLICABLE = "Not applicable";
+
+const Price = z.string().regex(/^\$\d+(?:\.\d+)?$/u);
+
+export const GitHubCopilotPricingRow = z.object({
+  model: z.string().min(1),
+  provider: z.string().min(1),
+  release_status: z.string().min(1),
+  category: z.string().min(1),
+  threshold: z.string().optional(),
+  tier: z.string().optional(),
+  input: Price,
+  cached_input: Price,
+  output: Price,
+  cache_write: z.union([Price, z.literal(NOT_APPLICABLE)]).optional(),
+  notes: z.string().optional(),
+}).passthrough();
+
+export type GitHubCopilotPricingRow = z.infer<typeof GitHubCopilotPricingRow>;
+
+export interface GitHubCopilotPricingModel {
+  slug: string;
+  releaseStatus: string;
+  rows: GitHubCopilotPricingRow[];
+}
+
+// Map names in pricing YAML to actual filenames in repo
+const FILE_ALIASES: Record<string, string> = {
+  "mai-code-1-flash": "mai-code-1-flash-picker",
+};
+
+const IGNORED_ROWS = new Set([
+  // Goes in [experimental.modes.fast] under claude-opus-4.8
+  "claude-opus-4.8-fast-mode-preview",
+  // No matching model in repo
+  "raptor-mini",
+]);
+
+export function githubCopilotModelSlug(name: string) {
+  return name
+    .replace(/\[\^[^\]]*\]/gu, "")
+    .toLowerCase()
+    .replace(/[^a-z0-9.]+/gu, "-")
+    .replace(/^-+|-+$/gu, "");
+}
+
+function price(value: string) {
+  return value === NOT_APPLICABLE ? undefined : Number(value.slice(1));
+}
+
+function rowCost(row: GitHubCopilotPricingRow) {
+  return {
+    input: price(row.input),
+    output: price(row.output),
+    cache_read: price(row.cached_input),
+    cache_write: row.cache_write === undefined ? undefined : price(row.cache_write),
+  };
+}
+
+function longContextThresholdSize(threshold: string | undefined, slug: string) {
+  const match = threshold?.match(/^>\s*(\d+(?:\.\d+)?)\s*([KM])$/u);
+  if (!match) throw new Error(`Unparseable long-context threshold for ${slug}: ${threshold}`);
+  return Number(match[1]) * (match[2] === "K" ? 1_000 : 1_000_000);
+}
+
+export function buildGitHubCopilotCost(model: GitHubCopilotPricingModel) {
+  const defaults: GitHubCopilotPricingRow[] = [];
+  const longContext: GitHubCopilotPricingRow[] = [];
+  for (const row of model.rows) {
+    const tier = row.tier ?? "Default";
+    if (tier === "Default") defaults.push(row);
+    else if (tier === "Long context") longContext.push(row);
+    else throw new Error(`Unknown pricing tier for ${model.slug}: ${row.tier}`);
+  }
+  const base = defaults[0];
+  if (base === undefined || defaults.length > 1) {
+    throw new Error(`Expected exactly one default pricing row for ${model.slug}, found ${defaults.length}`);
+  }
+
+  const tiers = longContext
+    .map((row) => ({
+      tier: { type: "context" as const, size: longContextThresholdSize(row.threshold, model.slug) },
+      ...rowCost(row),
+    }))
+    .sort((a, b) => a.tier.size - b.tier.size);
+
+  return { ...rowCost(base), tiers: tiers.length > 0 ? tiers : undefined };
+}
+
+export function parseGitHubCopilotPricing(raw: unknown) {
+  const rows = z.array(GitHubCopilotPricingRow).parse(raw);
+  const models = new Map<string, GitHubCopilotPricingModel>();
+  for (const row of rows) {
+    const slug = githubCopilotModelSlug(row.model);
+    const model = models.get(slug) ?? { slug, releaseStatus: row.release_status, rows: [] };
+    model.rows.push(row);
+    models.set(slug, model);
+  }
+  return [...models.values()];
+}
+
+export function buildGitHubCopilotModel(
+  model: GitHubCopilotPricingModel,
+  authored: ExistingModel,
+): SyncedModel {
+  // Only update token rates, leaving audio/reasoning rates and all other
+  // fields untouched.
+  const cost = { ...authored.cost, ...buildGitHubCopilotCost(model) };
+  return { ...authored, cost } as SyncedModel;
+}
+
+export const githubCopilot = {
+  id: "github-copilot",
+  name: "GitHub Copilot",
+  modelsDir: "providers/github-copilot/models",
+  skipCreates: true,
+  deleteMissing: false,
+  sourceID(model) {
+    return IGNORED_ROWS.has(model.slug) ? undefined : model.slug;
+  },
+  skippedNotice(ids) {
+    if (ids.length === 0) return [];
+    return [
+      `${ids.length} Copilot pricing table models have no local catalog file: ${ids.map((id) => `\`${id}\``).join(", ")}`,
+    ];
+  },
+  missingNotice(paths) {
+    if (paths.length === 0) return [];
+    return [
+      `${paths.length} local Copilot models are missing from the docs pricing table and were retained: ${paths.map((path) => `\`${path}\``).join(", ")}`,
+    ];
+  },
+  async fetchModels() {
+    const response = await fetch(PRICING_ENDPOINT);
+    if (!response.ok) {
+      throw new Error(`Copilot pricing request failed: ${response.status} ${response.statusText}`);
+    }
+    return Bun.YAML.parse(await response.text());
+  },
+  parseModels: parseGitHubCopilotPricing,
+  translateModel(model, context) {
+    if (IGNORED_ROWS.has(model.slug)) return undefined;
+    const candidates = [
+      model.slug,
+      FILE_ALIASES[model.slug],
+      model.releaseStatus === "Public preview" ? `${model.slug}-preview` : undefined,
+    ].filter((candidate) => candidate !== undefined);
+    const id = candidates.find((candidate) => context.authored(candidate) !== undefined);
+    const authored = id === undefined ? undefined : context.authored(id);
+    if (id === undefined || authored === undefined) return undefined;
+    return {
+      id,
+      model: buildGitHubCopilotModel(model, authored),
+      header: "# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing\n",
+    };
+  },
+} satisfies SyncProvider<GitHubCopilotPricingModel>;
diff --git a/packages/core/test/github-copilot.test.ts b/packages/core/test/github-copilot.test.ts
new file mode 100644
index 00000000000..d524665af3b
--- /dev/null
+++ b/packages/core/test/github-copilot.test.ts
@@ -0,0 +1,228 @@
+import { expect, test } from "bun:test";
+
+import {
+  buildGitHubCopilotCost,
+  githubCopilotModelSlug,
+  githubCopilot,
+  parseGitHubCopilotPricing,
+  type GitHubCopilotPricingRow,
+} from "../src/sync/providers/github-copilot.js";
+import type { ExistingModel } from "../src/sync/index.js";
+
+function row(overrides: Partial<GitHubCopilotPricingRow>): GitHubCopilotPricingRow {
+  return {
+    model: "GPT-5.6 Terra",
+    provider: "openai",
+    release_status: "GA",
+    category: "Versatile",
+    input: "$2.00",
+    cached_input: "$0.20",
+    output: "$12.00",
+    ...overrides,
+  };
+}
+
+test("slugifies display names into catalog filenames", () => {
+  expect(githubCopilotModelSlug("GPT-5.6 Sol[^gpt-56-sol-promo]")).toBe("gpt-5.6-sol");
+  expect(githubCopilotModelSlug("GPT-5 mini")).toBe("gpt-5-mini");
+  expect(githubCopilotModelSlug("GPT-5.3-Codex")).toBe("gpt-5.3-codex");
+  expect(githubCopilotModelSlug("Claude Opus 4.8 (fast mode) (preview)")).toBe("claude-opus-4.8-fast-mode-preview");
+  expect(githubCopilotModelSlug("MAI-Code-1.1-Flash")).toBe("mai-code-1.1-flash");
+  expect(githubCopilotModelSlug("Kimi K2.7 Code")).toBe("kimi-k2.7-code");
+});
+
+test("groups tier rows under one model", () => {
+  const models = parseGitHubCopilotPricing([
+    row({ threshold: "≤ 272K", tier: "Default" }),
+    row({ threshold: "> 272K", tier: "Long context", input: "$4.00", cached_input: "$0.40", output: "$18.00" }),
+    row({ model: "Claude Sonnet 5", provider: "anthropic", input: "$2.00", output: "$10.00", cache_write: "$2.50" }),
+  ]);
+  expect(models.map((model) => model.slug)).toEqual(["gpt-5.6-terra", "claude-sonnet-5"]);
+  expect(models[0]?.rows).toHaveLength(2);
+});
+
+test("builds flat cost with cache_write and Not applicable handling", () => {
+  const [model] = parseGitHubCopilotPricing([
+    row({ model: "GPT-5.4 mini", input: "$0.75", cached_input: "$0.075", output: "$4.50", cache_write: "Not applicable" }),
+  ]);
+  expect(buildGitHubCopilotCost(model!)).toEqual({
+    input: 0.75,
+    output: 4.5,
+    cache_read: 0.075,
+    cache_write: undefined,
+    tiers: undefined,
+  });
+});
+
+test("builds long-context tiers from threshold rows", () => {
+  const [model] = parseGitHubCopilotPricing([
+    row({ threshold: "≤ 272K", tier: "Default", cache_write: "$2.50" }),
+    row({ threshold: "> 272K", tier: "Long context", input: "$4.00", cached_input: "$0.40", output: "$18.00", cache_write: "$5.00" }),
+  ]);
+  expect(buildGitHubCopilotCost(model!)).toEqual({
+    input: 2,
+    output: 12,
+    cache_read: 0.2,
+    cache_write: 2.5,
+    tiers: [{
+      tier: { type: "context", size: 272_000 },
+      input: 4,
+      output: 18,
+      cache_read: 0.4,
+      cache_write: 5,
+    }],
+  });
+});
+
+test("rejects malformed tables instead of writing garbage", () => {
+  const build = (rows: GitHubCopilotPricingRow[]) => {
+    const models = parseGitHubCopilotPricing(rows);
+    return models.map((model) => buildGitHubCopilotCost(model));
+  };
+  // Unknown tier label.
+  expect(() => build([row({ tier: "Standard" })])).toThrow(/Unknown pricing tier/u);
+  // Two default rows for one model.
+  expect(() => build([row({}), row({})])).toThrow(/exactly one default pricing row/u);
+  // Long-context row without a parseable threshold.
+  expect(() => build([
+    row({ threshold: "≤ 272K", tier: "Default" }),
+    row({ threshold: "272K+", tier: "Long context" }),
+  ])).toThrow(/Unparseable long-context threshold/u);
+  // Price strings are schema-validated before translation.
+  expect(() => build([row({ input: "$1,000.00" })])).toThrow();
+  expect(() => build([row({ input: "Included" })])).toThrow();
+});
+
+function translationContext(files: Record<string, ExistingModel>) {
+  return {
+    existing: (id: string) => files[id],
+    authored: (id: string) => files[id],
+  };
+}
+
+const authoredTerra: ExistingModel = {
+  base_model: "openai/gpt-5.6-terra",
+  cost: { input: 1, output: 1, cache_read: 1 },
+};
+
+test("updates cost on the authored file and preserves audio rates", () => {
+  const [model] = parseGitHubCopilotPricing([row({ cache_write: "$2.50" })]);
+  const translated = githubCopilot.translateModel(model!, translationContext({
+    "gpt-5.6-terra": {
+      ...authoredTerra,
+      cost: { input: 1, output: 1, reasoning: 3, cache_read: 1, input_audio: 1.5, output_audio: 6 },
+    },
+  }));
+  expect(translated?.id).toBe("gpt-5.6-terra");
+  expect(translated?.model.cost).toMatchObject({
+    input: 2,
+    output: 12,
+    reasoning: 3,
+    cache_read: 0.2,
+    cache_write: 2.5,
+    input_audio: 1.5,
+    output_audio: 6,
+  });
+  expect((translated?.model as ExistingModel).base_model).toBe("openai/gpt-5.6-terra");
+});
+
+test("clears authored tiers and cache_write the table no longer lists", () => {
+  const [model] = parseGitHubCopilotPricing([row({ cache_write: "Not applicable" })]);
+  const translated = githubCopilot.translateModel(model!, translationContext({
+    "gpt-5.6-terra": {
+      ...authoredTerra,
+      cost: {
+        input: 1,
+        output: 1,
+        cache_read: 1,
+        cache_write: 9,
+        tiers: [{ tier: { type: "context", size: 272_000 }, input: 9, output: 9 }],
+      },
+    },
+  }));
+  const cost = translated?.model.cost;
+  expect(cost?.input).toBe(2);
+  // Stale authored values must not survive the spread; the runner strips the
+  // explicit undefineds before writing.
+  expect(cost?.cache_write).toBeUndefined();
+  expect(cost?.tiers).toBeUndefined();
+});
+
+test("resolves preview and alias filenames", () => {
+  const preview = parseGitHubCopilotPricing([
+    row({ model: "Gemini 3.1 Pro", provider: "google", release_status: "Public preview" }),
+  ]);
+  expect(githubCopilot.translateModel(preview[0]!, translationContext({
+    "gemini-3.1-pro-preview": authoredTerra,
+  }))?.id).toBe("gemini-3.1-pro-preview");
+
+  const alias = parseGitHubCopilotPricing([row({ model: "MAI-Code-1-Flash", provider: "microsoft" })]);
+  expect(githubCopilot.translateModel(alias[0]!, translationContext({
+    "mai-code-1-flash-picker": authoredTerra,
+  }))?.id).toBe("mai-code-1-flash-picker");
+
+  // An exact slug match wins over both fallbacks.
+  expect(githubCopilot.translateModel(preview[0]!, translationContext({
+    "gemini-3.1-pro": authoredTerra,
+    "gemini-3.1-pro-preview": authoredTerra,
+  }))?.id).toBe("gemini-3.1-pro");
+  expect(githubCopilot.translateModel(alias[0]!, translationContext({
+    "mai-code-1-flash": authoredTerra,
+    "mai-code-1-flash-picker": authoredTerra,
+  }))?.id).toBe("mai-code-1-flash");
+});
+
+test("skips ignored rows silently and unmatched rows with an ID", () => {
+  const [fastMode] = parseGitHubCopilotPricing([
+    row({ model: "Claude Opus 4.8 (fast mode) (preview)", provider: "anthropic" }),
+  ]);
+  expect(githubCopilot.translateModel(fastMode!, translationContext({}))).toBeUndefined();
+  expect(githubCopilot.sourceID(fastMode!)).toBeUndefined();
+
+  const [unmatched] = parseGitHubCopilotPricing([row({ model: "Brand New Model" })]);
+  expect(githubCopilot.translateModel(unmatched!, translationContext({}))).toBeUndefined();
+  expect(githubCopilot.sourceID(unmatched!)).toBe("brand-new-model");
+});
+
+const pricingYaml = `
+- model: 'GPT-5.6 Sol[^gpt-56-sol-promo]'
+  provider: openai
+  release_status: GA
+  category: Powerful
+  threshold: '≤ 272K'
+  tier: Default
+  input: $2.00
+  cached_input: $0.20
+  output: $10.00
+  cache_write: $2.50
+
+- model: 'GPT-5.6 Sol[^gpt-56-sol-promo]'
+  provider: openai
+  release_status: GA
+  category: Powerful
+  threshold: '> 272K'
+  tier: 'Long context'
+  input: $4.00
+  cached_input: $0.40
+  output: $15.00
+  cache_write: $5.00
+`;
+
+test("parses rows straight from the docs YAML format", () => {
+  const models = parseGitHubCopilotPricing(Bun.YAML.parse(pricingYaml));
+  expect(models).toHaveLength(1);
+  expect(models[0]?.slug).toBe("gpt-5.6-sol");
+  expect(buildGitHubCopilotCost(models[0]!)).toEqual({
+    input: 2,
+    output: 10,
+    cache_read: 0.2,
+    cache_write: 2.5,
+    tiers: [{
+      tier: { type: "context", size: 272_000 },
+      input: 4,
+      output: 15,
+      cache_read: 0.4,
+      cache_write: 5,
+    }],
+  });
+});
diff --git a/providers/github-copilot/models/gemini-3.6-flash.toml b/providers/github-copilot/models/gemini-3.6-flash.toml
index e6fab531054..bbb6d9fcae4 100644
--- a/providers/github-copilot/models/gemini-3.6-flash.toml
+++ b/providers/github-copilot/models/gemini-3.6-flash.toml
@@ -1,10 +1,19 @@
+# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
 base_model = "google/gemini-3.6-flash"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }, { type = "budget_tokens", min = 256, max = 32_000 }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 256
+max = 32_000
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
+input = 0.75
+output = 3.75
+cache_read = 0.075
 
 [limit]
 context = 1_000_000
diff --git a/providers/github-copilot/models/gpt-5.6-luna.toml b/providers/github-copilot/models/gpt-5.6-luna.toml
index 2521a9108ab..9adeea21314 100644
--- a/providers/github-copilot/models/gpt-5.6-luna.toml
+++ b/providers/github-copilot/models/gpt-5.6-luna.toml
@@ -1,15 +1,19 @@
 # Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
-# GPT-5.6 Luna reduced rates (matches OpenAI 2026-07-30 cut); long-context threshold >200K
 base_model = "openai/gpt-5.6-luna"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.20
-output = 1.20
+input = 0.2
+output = 1.2
 cache_read = 0.02
+cache_write = 0.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 0.40
-output = 1.80
+input = 0.4
+output = 1.8
 cache_read = 0.04
+cache_write = 0.5
diff --git a/providers/github-copilot/models/gpt-5.6-sol.toml b/providers/github-copilot/models/gpt-5.6-sol.toml
index 0e083daa0c9..f4370368104 100644
--- a/providers/github-copilot/models/gpt-5.6-sol.toml
+++ b/providers/github-copilot/models/gpt-5.6-sol.toml
@@ -1,19 +1,19 @@
-# Source: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
-# GPT-5.6 Sol promotional pricing (50% off standard rates) through September 3, 2026.
-# Default ≤272K: $2.50 / $0.25 / $3.125 / $15.00 (input / cache_read / cache_write / output)
-# Long context >272K: $5.00 / $0.50 / $6.25 / $22.50
+# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
 base_model = "openai/gpt-5.6-sol"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.5
-output = 15
-cache_read = 0.25
-cache_write = 3.125
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 5
-output = 22.5
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
diff --git a/providers/github-copilot/models/gpt-5.6-terra.toml b/providers/github-copilot/models/gpt-5.6-terra.toml
index ad20e81d801..a61acebebe7 100644
--- a/providers/github-copilot/models/gpt-5.6-terra.toml
+++ b/providers/github-copilot/models/gpt-5.6-terra.toml
@@ -1,15 +1,19 @@
 # Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
-# GPT-5.6 Terra reduced rates (matches OpenAI 2026-07-30 cut); long-context threshold >272K
 base_model = "openai/gpt-5.6-terra"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.00
-output = 12.00
-cache_read = 0.20
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 4.00
-output = 18.00
-cache_read = 0.40
+input = 4
+output = 18
+cache_read = 0.4
+cache_write = 5
diff --git a/sync.md b/sync.md
index 735d20a17d4..a9c59774309 100644
--- a/sync.md
+++ b/sync.md
@@ -19,6 +19,7 @@ The grouped sync targets are available for local convenience, but CI syncs each
 - `bun models:sync kilo` syncs only Kilo.
 - `bun models:sync merge-gateway` syncs only Merge Gateway.
 - `bun models:sync openai` syncs only OpenAI catalog availability.
+- `bun models:sync github-copilot` syncs only GitHub Copilot pricing.
 - `bun models:sync tinfoil` syncs only Tinfoil.
 - `bun models:sync aggregators --dry-run` prints changes without writing model files.
 - `bun models:sync aggregators --new-only` creates new model files but skips updates and removals.
@@ -215,6 +216,15 @@ Google is implemented in `packages/core/src/sync/providers/google.ts`.
 - New Google API models are not created automatically (`skipCreates`) and do not open missing-model issues because the endpoint is not lifecycle-authoritative.
 - Missing-model tracking is limited to recognizable public model families; opaque API codenames such as `ajax`, `perseus`, and `thorin` are ignored.
 
+## GitHub Copilot Notes
+
+GitHub Copilot is implemented in `packages/core/src/sync/providers/github-copilot.ts`.
+
+- Source: `https://raw.githubusercontent.com/github/docs/main/data/tables/copilot/models-and-pricing.yml`
+- The YML contains only token rates, so the sync only updates `[cost]`: `input`, `cached_input` (as `cache_read`), `cache_write`, `output`, and long-context rows as `cost.tiers`.
+- Display names are converted to file IDs, with minimal special case logic to match existing model entries.
+- Unmatched rows open missing-model issues, and local entries missing from the source are kept.
+
 ## xAI Notes
 
 xAI is implemented in `packages/core/src/sync/providers/xai.ts`.

From 0aecf6f42dc7bda4c019c5c54911f6ef25a2761f Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Fri, 28 Aug 2026 07:40:53 +0800
Subject: [PATCH 0088/2012] feat(vivgrid): add glm-5.3-flash (#5616)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Vivgrid now serves Zhipu AI's GLM-5.3-Flash through its unified
OpenAI-compatible endpoint. Override-only entry on the existing
zhipuai/glm-5.3-flash lab metadata.

Pricing (USD per 1M tokens, https://docs.vivgrid.com/models):
input $0.15 / cached input $0.04 / output $0.50.

Reasoning controls and the reasoning_content side channel follow the
sibling providers/vivgrid/models/glm-5.3.toml entry — same host, same
lab, same generation, same OpenAI-compatible surface.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/vivgrid/models/glm-5.3-flash.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/vivgrid/models/glm-5.3-flash.toml

diff --git a/providers/vivgrid/models/glm-5.3-flash.toml b/providers/vivgrid/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..c96e44f1b38
--- /dev/null
+++ b/providers/vivgrid/models/glm-5.3-flash.toml
@@ -0,0 +1,12 @@
+# Pricing per https://docs.vivgrid.com/models (accessed 2026-08-27):
+# input $0.15 / cached input $0.04 / output $0.50 per 1M tokens.
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.04

From cee0830fa542918861a91f2f05327dd15830fafa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:42:03 -0500
Subject: [PATCH 0089/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5672)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): expose Ling Flash Fin reasoning toggle

Replace empty reasoning options with the binary reasoning.enabled control on both Ling 3.0 Flash Fin routes. Do not advertise unsupported graded effort or token budgets.

Extract shared InclusionAI model metadata as required by the catalog guidelines, retaining Vercel-specific limits and prices.

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/inclusionai/ling-3.0-flash-fin.toml    | 20 +++++++++++++++++++
 .../inclusionai/ling-3.0-flash-fin-free.toml  | 14 +++++++++++++
 .../inclusionai/ling-3.0-flash-fin.toml       | 13 ++++++++++++
 3 files changed, 47 insertions(+)
 create mode 100644 models/inclusionai/ling-3.0-flash-fin.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-fin-free.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-fin.toml

diff --git a/models/inclusionai/ling-3.0-flash-fin.toml b/models/inclusionai/ling-3.0-flash-fin.toml
new file mode 100644
index 00000000000..2e98f3b45d5
--- /dev/null
+++ b/models/inclusionai/ling-3.0-flash-fin.toml
@@ -0,0 +1,20 @@
+# https://novita.ai/models/model-detail/inclusionai-ling-3.0-flash-fin
+# https://openrouter.ai/api/v1/models/inclusionai/ling-3.0-flash-fin:free/endpoints
+# https://vercel.com/changelog/ling-3-0-flash-fin-now-available-on-ai-gateway-for-free
+name = "Ling 3.0 Flash Fin"
+description = "Finance-enhanced model for financial research, multi-step investment workflows, and long-horizon planning and execution"
+family = "ling"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-fin-free.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-fin-free.toml
new file mode 100644
index 00000000000..ac37b48734e
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-fin-free.toml
@@ -0,0 +1,14 @@
+# Toggle: reasoning.enabled = true|false (Chat Completions API)
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# https://ai-gateway.vercel.sh/v1/models/inclusionai/ling-3.0-flash-fin-free/endpoints
+base_model = "inclusionai/ling-3.0-flash-fin"
+name = "Ling 3.0 Flash Fin (Free)"
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-fin.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-fin.toml
new file mode 100644
index 00000000000..757facd2dbc
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-fin.toml
@@ -0,0 +1,13 @@
+# Toggle: reasoning.enabled = true|false (Chat Completions API)
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# https://ai-gateway.vercel.sh/v1/models/inclusionai/ling-3.0-flash-fin/endpoints
+base_model = "inclusionai/ling-3.0-flash-fin"
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000

From 3c500e1f99f3f9b608e4d58599bb11c67a8518f3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 18:42:28 -0500
Subject: [PATCH 0090/2012] chore(sync): update Requesty model catalog (#5670)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/claude-fable-5.toml   |  8 ++++----
 .../requesty/models/claude-fable-5@eu.toml      |  8 ++++----
 providers/requesty/models/claude-haiku-4-5.toml |  8 ++++----
 .../requesty/models/claude-haiku-4-5@eu.toml    |  8 ++++----
 providers/requesty/models/claude-opus-4-1.toml  |  8 ++++----
 providers/requesty/models/claude-opus-4-5.toml  |  8 ++++----
 .../requesty/models/claude-opus-4-5@eu.toml     |  8 ++++----
 providers/requesty/models/claude-opus-4-6.toml  |  8 ++++----
 .../requesty/models/claude-opus-4-6@eu.toml     |  8 ++++----
 providers/requesty/models/claude-opus-4-7.toml  |  8 ++++----
 .../requesty/models/claude-opus-4-7@eu.toml     |  8 ++++----
 providers/requesty/models/claude-opus-4-8.toml  |  8 ++++----
 .../requesty/models/claude-opus-4-8@eu.toml     |  8 ++++----
 providers/requesty/models/claude-opus-5.toml    |  8 ++++----
 providers/requesty/models/claude-opus-5@eu.toml |  8 ++++----
 .../requesty/models/claude-sonnet-4-5.toml      | 16 ++++++++--------
 .../requesty/models/claude-sonnet-4-5@eu.toml   | 16 ++++++++--------
 .../requesty/models/claude-sonnet-4-6.toml      |  8 ++++----
 .../requesty/models/claude-sonnet-4-6@eu.toml   |  8 ++++----
 .../requesty/models/claude-sonnet-4@eu.toml     | 16 ++++++++--------
 providers/requesty/models/claude-sonnet-5.toml  |  8 ++++----
 .../requesty/models/claude-sonnet-5@eu.toml     |  8 ++++----
 .../requesty/models/deepseek-v4-flash-0731.toml |  6 +++---
 .../models/deepseek-v4-flash-0731@eu.toml       |  6 +++---
 .../requesty/models/deepseek-v4-flash.toml      |  6 +++---
 .../requesty/models/deepseek-v4-pro-0813.toml   |  6 +++---
 providers/requesty/models/deepseek-v4-pro.toml  |  6 +++---
 .../requesty/models/deepseek-v4-pro@eu.toml     |  6 +++---
 providers/requesty/models/devstral-latest.toml  |  6 +++---
 .../requesty/models/devstral-latest@eu.toml     |  6 +++---
 providers/requesty/models/fugu-ultra.toml       | 12 ++++++------
 .../requesty/models/gemini-2.5-flash@eu.toml    |  8 ++++----
 .../requesty/models/gemini-3-pro-image.toml     | 16 ++++++++--------
 .../requesty/models/gemini-3.1-flash-image.toml |  8 ++++----
 .../requesty/models/gemini-3.1-flash-lite.toml  | 16 ++++++++--------
 .../models/gemini-3.1-flash-lite@eu.toml        | 16 ++++++++--------
 .../requesty/models/gemini-3.1-pro-preview.toml | 16 ++++++++--------
 .../requesty/models/gemini-3.5-flash-lite.toml  |  6 +++---
 .../models/gemini-3.5-flash-lite@eu.toml        |  6 +++---
 providers/requesty/models/gemini-3.5-flash.toml |  8 ++++----
 .../requesty/models/gemini-3.5-flash@eu.toml    |  8 ++++----
 providers/requesty/models/gemini-3.6-flash.toml |  6 +++---
 providers/requesty/models/gemini-3.7-flash.toml |  6 +++---
 .../requesty/models/gemini-3.7-flash@eu.toml    |  6 +++---
 .../requesty/models/gemma-4-26b-a4b-it.toml     |  6 +++---
 providers/requesty/models/glm-5.1.toml          |  6 +++---
 providers/requesty/models/glm-5.1@eu.toml       |  6 +++---
 providers/requesty/models/glm-5.2-fast.toml     |  6 +++---
 providers/requesty/models/glm-5.2.toml          |  6 +++---
 providers/requesty/models/glm-5.2@eu.toml       |  6 +++---
 providers/requesty/models/glm-5.3-flash.toml    |  6 +++---
 providers/requesty/models/glm-5.3-flash@eu.toml | 17 +++++++++++++++++
 providers/requesty/models/glm-5.3.toml          |  6 +++---
 providers/requesty/models/gpt-4.1-mini@eu.toml  |  6 +++---
 providers/requesty/models/gpt-4.1-nano@eu.toml  |  6 +++---
 providers/requesty/models/gpt-4.1@eu.toml       |  6 +++---
 providers/requesty/models/gpt-4o-mini@eu.toml   |  6 +++---
 providers/requesty/models/gpt-5-mini@eu.toml    |  6 +++---
 providers/requesty/models/gpt-5-nano@eu.toml    |  6 +++---
 providers/requesty/models/gpt-5.1@eu.toml       |  6 +++---
 providers/requesty/models/gpt-5.3-chat.toml     |  6 +++---
 providers/requesty/models/gpt-5.3-codex.toml    |  6 +++---
 providers/requesty/models/gpt-5.4-mini.toml     |  6 +++---
 providers/requesty/models/gpt-5.4-nano.toml     |  6 +++---
 providers/requesty/models/gpt-5.4-pro.toml      |  6 +++---
 providers/requesty/models/gpt-5.4.toml          |  6 +++---
 providers/requesty/models/gpt-5.4@eu.toml       | 12 ++++++------
 providers/requesty/models/gpt-5.5-pro.toml      |  4 ++--
 providers/requesty/models/gpt-5.5.toml          |  6 +++---
 providers/requesty/models/gpt-5.5@eu.toml       | 12 ++++++------
 providers/requesty/models/gpt-5.6-luna.toml     | 12 ++++++------
 providers/requesty/models/gpt-5.6-luna@eu.toml  |  6 +++---
 providers/requesty/models/gpt-5.6-sol.toml      | 16 ++++++++--------
 providers/requesty/models/gpt-5.6-sol@eu.toml   |  6 +++---
 providers/requesty/models/gpt-5.6-terra.toml    | 12 ++++++------
 providers/requesty/models/gpt-5.6-terra@eu.toml |  6 +++---
 providers/requesty/models/gpt-5@eu.toml         |  6 +++---
 providers/requesty/models/grok-4.2-beta.toml    | 16 ++++++++--------
 providers/requesty/models/grok-4.3.toml         | 16 ++++++++--------
 providers/requesty/models/grok-4.5.toml         | 16 ++++++++--------
 providers/requesty/models/grok-4.6.toml         | 16 ++++++++--------
 providers/requesty/models/grok-build-0.1.toml   |  6 +++---
 providers/requesty/models/hy3.toml              |  6 +++---
 providers/requesty/models/inkling-256k.toml     |  6 +++---
 providers/requesty/models/inkling.toml          |  6 +++---
 providers/requesty/models/kat-coder-pro.toml    |  4 ++--
 providers/requesty/models/kimi-k2.6.toml        |  6 +++---
 providers/requesty/models/kimi-k2.6@eu.toml     |  6 +++---
 providers/requesty/models/kimi-k2.7-code.toml   |  6 +++---
 .../requesty/models/kimi-k2.7-code@eu.toml      |  6 +++---
 providers/requesty/models/kimi-k3.toml          |  6 +++---
 providers/requesty/models/kimi-k3@eu.toml       |  6 +++---
 providers/requesty/models/ling-2.6-1t.toml      |  4 ++--
 providers/requesty/models/ling-2.6-flash.toml   |  4 ++--
 providers/requesty/models/mimo-v2.5-pro.toml    |  6 +++---
 providers/requesty/models/mimo-v2.5.toml        |  6 +++---
 .../requesty/models/minimax-m2.7-highspeed.toml |  8 ++++----
 providers/requesty/models/minimax-m2.7.toml     |  8 ++++----
 providers/requesty/models/minimax-m3.toml       |  6 +++---
 providers/requesty/models/minimax-m3@eu.toml    |  6 +++---
 .../requesty/models/mistral-medium-3-5.toml     |  6 +++---
 .../requesty/models/mistral-medium-3-5@eu.toml  |  6 +++---
 .../requesty/models/mistral-medium-latest.toml  |  6 +++---
 .../models/mistral-medium-latest@eu.toml        |  6 +++---
 .../requesty/models/mistral-small-2603.toml     |  6 +++---
 .../requesty/models/mistral-small-2603@eu.toml  |  6 +++---
 .../requesty/models/nemotron-3-nano-omni.toml   |  6 +++---
 .../models/nemotron-3-nano-omni@eu.toml         |  6 +++---
 .../requesty/models/nemotron-3-ultra-nvfp4.toml |  6 +++---
 .../models/nemotron-lightning-3.5-30b-a3b.toml  |  6 +++---
 .../nvidia-nemotron-3-super-120b-a12b.toml      |  4 ++--
 .../models/nvidia-nemotron-3-ultra.toml         |  4 ++--
 providers/requesty/models/o4-mini@eu.toml       |  6 +++---
 providers/requesty/models/qwen3.5-27b.toml      |  4 ++--
 providers/requesty/models/qwen3.5-2b.toml       |  4 ++--
 providers/requesty/models/qwen3.5-35b-a3b.toml  |  6 +++---
 providers/requesty/models/qwen3.6-plus.toml     |  8 ++++----
 providers/requesty/models/qwen3.7-max.toml      |  8 ++++----
 providers/requesty/models/qwen3.7-plus.toml     |  8 ++++----
 .../requesty/models/qwen3.8-2.4T-A95B.toml      |  6 +++---
 providers/requesty/models/qwen3.8-max.toml      |  8 ++++----
 providers/requesty/models/ring-2.6-1t.toml      |  4 ++--
 providers/requesty/models/seed-1.8.toml         |  6 +++---
 providers/requesty/models/seed-2.0-code.toml    |  6 +++---
 providers/requesty/models/seed-2.0-mini.toml    |  6 +++---
 providers/requesty/models/seed-2.0-pro.toml     |  6 +++---
 providers/requesty/models/step-3.7-flash.toml   |  6 +++---
 .../models/thinkingcap-qwen3.6-27b.toml         |  6 +++---
 .../models/thinkingcap-qwen3.6-27b@eu.toml      |  6 +++---
 129 files changed, 496 insertions(+), 479 deletions(-)
 create mode 100644 providers/requesty/models/glm-5.3-flash@eu.toml

diff --git a/providers/requesty/models/claude-fable-5.toml b/providers/requesty/models/claude-fable-5.toml
index e2a107d8842..685f20fb08b 100644
--- a/providers/requesty/models/claude-fable-5.toml
+++ b/providers/requesty/models/claude-fable-5.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 9
-output = 45
-cache_read = 0.9
-cache_write = 11.25
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
diff --git a/providers/requesty/models/claude-fable-5@eu.toml b/providers/requesty/models/claude-fable-5@eu.toml
index 35f9764cd66..7e5d5f41f71 100644
--- a/providers/requesty/models/claude-fable-5@eu.toml
+++ b/providers/requesty/models/claude-fable-5@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 9.9
-output = 49.5
-cache_read = 0.99
-cache_write = 12.375
+input = 11
+output = 55
+cache_read = 1.1
+cache_write = 13.75
diff --git a/providers/requesty/models/claude-haiku-4-5.toml b/providers/requesty/models/claude-haiku-4-5.toml
index 67ec34cf709..ef99a1dbd00 100644
--- a/providers/requesty/models/claude-haiku-4-5.toml
+++ b/providers/requesty/models/claude-haiku-4-5.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.9
-output = 4.5
-cache_read = 0.09
-cache_write = 1.125
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/requesty/models/claude-haiku-4-5@eu.toml b/providers/requesty/models/claude-haiku-4-5@eu.toml
index 5cc8db46cdd..db382199492 100644
--- a/providers/requesty/models/claude-haiku-4-5@eu.toml
+++ b/providers/requesty/models/claude-haiku-4-5@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.99
-output = 4.95
-cache_read = 0.099
-cache_write = 1.2375
+input = 1.1
+output = 5.5
+cache_read = 0.11
+cache_write = 1.375
diff --git a/providers/requesty/models/claude-opus-4-1.toml b/providers/requesty/models/claude-opus-4-1.toml
index 18401d20661..bde914f89d9 100644
--- a/providers/requesty/models/claude-opus-4-1.toml
+++ b/providers/requesty/models/claude-opus-4-1.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 13.5
-output = 67.5
-cache_read = 1.35
-cache_write = 16.875
+input = 15
+output = 75
+cache_read = 1.5
+cache_write = 18.75
diff --git a/providers/requesty/models/claude-opus-4-5.toml b/providers/requesty/models/claude-opus-4-5.toml
index 222b87f29e1..f00d78dd6bb 100644
--- a/providers/requesty/models/claude-opus-4-5.toml
+++ b/providers/requesty/models/claude-opus-4-5.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-cache_write = 5.625
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-4-5@eu.toml b/providers/requesty/models/claude-opus-4-5@eu.toml
index 93abb718801..58430132a60 100644
--- a/providers/requesty/models/claude-opus-4-5@eu.toml
+++ b/providers/requesty/models/claude-opus-4-5@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 24.75
-cache_read = 0.495
-cache_write = 6.1875
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/requesty/models/claude-opus-4-6.toml b/providers/requesty/models/claude-opus-4-6.toml
index b82af39fee8..e43a8709a23 100644
--- a/providers/requesty/models/claude-opus-4-6.toml
+++ b/providers/requesty/models/claude-opus-4-6.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-cache_write = 5.625
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-4-6@eu.toml b/providers/requesty/models/claude-opus-4-6@eu.toml
index 56d5804900f..f5e44965220 100644
--- a/providers/requesty/models/claude-opus-4-6@eu.toml
+++ b/providers/requesty/models/claude-opus-4-6@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 24.75
-cache_read = 0.495
-cache_write = 6.1875
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/requesty/models/claude-opus-4-7.toml b/providers/requesty/models/claude-opus-4-7.toml
index 9b33e0c52b5..4c464ebf796 100644
--- a/providers/requesty/models/claude-opus-4-7.toml
+++ b/providers/requesty/models/claude-opus-4-7.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-cache_write = 5.625
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-4-7@eu.toml b/providers/requesty/models/claude-opus-4-7@eu.toml
index 514bce6a512..b2067197c30 100644
--- a/providers/requesty/models/claude-opus-4-7@eu.toml
+++ b/providers/requesty/models/claude-opus-4-7@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 24.75
-cache_read = 0.495
-cache_write = 6.1875
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/requesty/models/claude-opus-4-8.toml b/providers/requesty/models/claude-opus-4-8.toml
index a77ef72f101..32911565b01 100644
--- a/providers/requesty/models/claude-opus-4-8.toml
+++ b/providers/requesty/models/claude-opus-4-8.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-cache_write = 5.625
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-4-8@eu.toml b/providers/requesty/models/claude-opus-4-8@eu.toml
index ec2bef267b6..af4708b3bfd 100644
--- a/providers/requesty/models/claude-opus-4-8@eu.toml
+++ b/providers/requesty/models/claude-opus-4-8@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 24.75
-cache_read = 0.495
-cache_write = 6.1875
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/requesty/models/claude-opus-5.toml b/providers/requesty/models/claude-opus-5.toml
index b765dada372..5807ccae254 100644
--- a/providers/requesty/models/claude-opus-5.toml
+++ b/providers/requesty/models/claude-opus-5.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-cache_write = 5.625
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-5@eu.toml b/providers/requesty/models/claude-opus-5@eu.toml
index f78e6d571a0..efe07665120 100644
--- a/providers/requesty/models/claude-opus-5@eu.toml
+++ b/providers/requesty/models/claude-opus-5@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 24.75
-cache_read = 0.495
-cache_write = 6.1875
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/requesty/models/claude-sonnet-4-5.toml b/providers/requesty/models/claude-sonnet-4-5.toml
index c509d9f175e..fb2d6fa1c69 100644
--- a/providers/requesty/models/claude-sonnet-4-5.toml
+++ b/providers/requesty/models/claude-sonnet-4-5.toml
@@ -9,17 +9,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.7
-output = 13.5
-cache_read = 0.27
-cache_write = 3.375
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 5.4
-output = 20.25
-cache_read = 0.54
-cache_write = 6.75
+input = 6
+output = 22.5
+cache_read = 0.6
+cache_write = 7.5
 
 [limit]
 context = 1_000_000
diff --git a/providers/requesty/models/claude-sonnet-4-5@eu.toml b/providers/requesty/models/claude-sonnet-4-5@eu.toml
index 418061c557d..cadc9b45bb9 100644
--- a/providers/requesty/models/claude-sonnet-4-5@eu.toml
+++ b/providers/requesty/models/claude-sonnet-4-5@eu.toml
@@ -10,17 +10,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.97
-output = 14.85
-cache_read = 0.27
-cache_write = 3.7125
+input = 3.3
+output = 16.5
+cache_read = 0.3
+cache_write = 4.125
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 5.94
-output = 22.275
-cache_read = 0.54
-cache_write = 7.425
+input = 6.6
+output = 24.75
+cache_read = 0.6
+cache_write = 8.25
 
 [limit]
 context = 1_000_000
diff --git a/providers/requesty/models/claude-sonnet-4-6.toml b/providers/requesty/models/claude-sonnet-4-6.toml
index 33f7728a6f0..011600033b7 100644
--- a/providers/requesty/models/claude-sonnet-4-6.toml
+++ b/providers/requesty/models/claude-sonnet-4-6.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.7
-output = 13.5
-cache_read = 0.27
-cache_write = 3.375
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [limit]
 output = 128_000
diff --git a/providers/requesty/models/claude-sonnet-4-6@eu.toml b/providers/requesty/models/claude-sonnet-4-6@eu.toml
index 94e97bb862a..fb541539029 100644
--- a/providers/requesty/models/claude-sonnet-4-6@eu.toml
+++ b/providers/requesty/models/claude-sonnet-4-6@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.97
-output = 14.85
-cache_read = 0.27
-cache_write = 3.7125
+input = 3.3
+output = 16.5
+cache_read = 0.3
+cache_write = 4.125
diff --git a/providers/requesty/models/claude-sonnet-4@eu.toml b/providers/requesty/models/claude-sonnet-4@eu.toml
index 082d254b653..c66eab29d92 100644
--- a/providers/requesty/models/claude-sonnet-4@eu.toml
+++ b/providers/requesty/models/claude-sonnet-4@eu.toml
@@ -10,17 +10,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.7
-output = 13.5
-cache_read = 0.27
-cache_write = 3.375
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 5.4
-output = 20.25
-cache_read = 0.54
-cache_write = 6.75
+input = 6
+output = 22.5
+cache_read = 0.6
+cache_write = 7.5
 
 [limit]
 context = 1_000_000
diff --git a/providers/requesty/models/claude-sonnet-5.toml b/providers/requesty/models/claude-sonnet-5.toml
index b5edcbfe133..ef081972290 100644
--- a/providers/requesty/models/claude-sonnet-5.toml
+++ b/providers/requesty/models/claude-sonnet-5.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 9
-cache_read = 0.18
-cache_write = 2.25
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/requesty/models/claude-sonnet-5@eu.toml b/providers/requesty/models/claude-sonnet-5@eu.toml
index e7592a83115..d4db42a626b 100644
--- a/providers/requesty/models/claude-sonnet-5@eu.toml
+++ b/providers/requesty/models/claude-sonnet-5@eu.toml
@@ -10,7 +10,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.98
-output = 9.9
-cache_read = 0.198
-cache_write = 2.475
+input = 2.2
+output = 11
+cache_read = 0.22
+cache_write = 2.75
diff --git a/providers/requesty/models/deepseek-v4-flash-0731.toml b/providers/requesty/models/deepseek-v4-flash-0731.toml
index 8509a5326fd..d73d297b44e 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.126
-output = 0.252
-cache_read = 0.063
+input = 0.14
+output = 0.28
+cache_read = 0.07
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/deepseek-v4-flash-0731@eu.toml b/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
index 50bdcdee824..e33044e4663 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.126
-output = 0.252
-cache_read = 0.063
+input = 0.14
+output = 0.28
+cache_read = 0.07
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/deepseek-v4-flash.toml b/providers/requesty/models/deepseek-v4-flash.toml
index cba6b291fb2..991114cd014 100644
--- a/providers/requesty/models/deepseek-v4-flash.toml
+++ b/providers/requesty/models/deepseek-v4-flash.toml
@@ -3,6 +3,6 @@ reasoning = false
 structured_output = false
 
 [cost]
-input = 0.396
-output = 1.188
-cache_read = 0.0126
+input = 0.44
+output = 1.32
+cache_read = 0.014
diff --git a/providers/requesty/models/deepseek-v4-pro-0813.toml b/providers/requesty/models/deepseek-v4-pro-0813.toml
index 80f2c6e61e7..117c7233a1e 100644
--- a/providers/requesty/models/deepseek-v4-pro-0813.toml
+++ b/providers/requesty/models/deepseek-v4-pro-0813.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.188
-output = 3.564
-cache_read = 0.0396
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 output = 131_072
diff --git a/providers/requesty/models/deepseek-v4-pro.toml b/providers/requesty/models/deepseek-v4-pro.toml
index 82bea369ce9..6b68c88eba5 100644
--- a/providers/requesty/models/deepseek-v4-pro.toml
+++ b/providers/requesty/models/deepseek-v4-pro.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.188
-output = 3.564
-cache_read = 0.0396
+input = 1.32
+output = 3.96
+cache_read = 0.044
diff --git a/providers/requesty/models/deepseek-v4-pro@eu.toml b/providers/requesty/models/deepseek-v4-pro@eu.toml
index 8d322b19c98..5ac0f1fe8f6 100644
--- a/providers/requesty/models/deepseek-v4-pro@eu.toml
+++ b/providers/requesty/models/deepseek-v4-pro@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.575
-output = 3.15
-cache_read = 0.396
+input = 1.75
+output = 3.5
+cache_read = 0.44
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/devstral-latest.toml b/providers/requesty/models/devstral-latest.toml
index 3dd0b187efb..6b8094c8fb4 100644
--- a/providers/requesty/models/devstral-latest.toml
+++ b/providers/requesty/models/devstral-latest.toml
@@ -10,9 +10,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.396
-output = 1.98
-cache_read = 0.396
+input = 0.44
+output = 2.2
+cache_read = 0.44
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/devstral-latest@eu.toml b/providers/requesty/models/devstral-latest@eu.toml
index 040860df3e3..046e77deebd 100644
--- a/providers/requesty/models/devstral-latest@eu.toml
+++ b/providers/requesty/models/devstral-latest@eu.toml
@@ -10,9 +10,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.396
-output = 1.98
-cache_read = 0.396
+input = 0.44
+output = 2.2
+cache_read = 0.44
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/fugu-ultra.toml b/providers/requesty/models/fugu-ultra.toml
index a36b7df3b9c..bc82853c108 100644
--- a/providers/requesty/models/fugu-ultra.toml
+++ b/providers/requesty/models/fugu-ultra.toml
@@ -3,15 +3,15 @@ reasoning = false
 structured_output = false
 
 [cost]
-input = 4.5
-output = 27
-cache_read = 0.45
+input = 5
+output = 30
+cache_read = 0.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 9
-output = 40.5
-cache_read = 0.9
+input = 10
+output = 45
+cache_read = 1
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/gemini-2.5-flash@eu.toml b/providers/requesty/models/gemini-2.5-flash@eu.toml
index 2416c428431..fdf5313133a 100644
--- a/providers/requesty/models/gemini-2.5-flash@eu.toml
+++ b/providers/requesty/models/gemini-2.5-flash@eu.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.27
-output = 2.25
-cache_read = 0.0675
-cache_write = 0.495
+input = 0.3
+output = 2.5
+cache_read = 0.075
+cache_write = 0.55
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3-pro-image.toml b/providers/requesty/models/gemini-3-pro-image.toml
index 7adf6cd3c9d..21eb5f8c6ce 100644
--- a/providers/requesty/models/gemini-3-pro-image.toml
+++ b/providers/requesty/models/gemini-3-pro-image.toml
@@ -10,17 +10,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 10.8
-cache_read = 0.18
-cache_write = 4.05
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 4.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 3.6
-output = 16.2
-cache_read = 0.18
-cache_write = 4.05
+input = 4
+output = 18
+cache_read = 0.2
+cache_write = 4.5
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/gemini-3.1-flash-image.toml b/providers/requesty/models/gemini-3.1-flash-image.toml
index 9d3631718de..c3b869b378a 100644
--- a/providers/requesty/models/gemini-3.1-flash-image.toml
+++ b/providers/requesty/models/gemini-3.1-flash-image.toml
@@ -10,10 +10,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 1.8
+input = 0.5
+output = 2
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 0.45
-output = 1.8
+input = 0.5
+output = 2
diff --git a/providers/requesty/models/gemini-3.1-flash-lite.toml b/providers/requesty/models/gemini-3.1-flash-lite.toml
index ffbedc9c3fc..f45173eeab0 100644
--- a/providers/requesty/models/gemini-3.1-flash-lite.toml
+++ b/providers/requesty/models/gemini-3.1-flash-lite.toml
@@ -8,17 +8,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.225
-output = 1.35
-cache_read = 0.0225
-cache_write = 0.074997
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.08333
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 0.45
-output = 2.025
-cache_read = 0.0225
-cache_write = 0.074997
+input = 0.5
+output = 2.25
+cache_read = 0.025
+cache_write = 0.08333
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.1-flash-lite@eu.toml b/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
index 7e6b034a598..a5170790571 100644
--- a/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
+++ b/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
@@ -9,17 +9,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.2475
-output = 1.485
-cache_read = 0.02475
-cache_write = 0.082497
+input = 0.275
+output = 1.65
+cache_read = 0.0275
+cache_write = 0.091663
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 0.495
-output = 2.2275
-cache_read = 0.02475
-cache_write = 0.082497
+input = 0.55
+output = 2.475
+cache_read = 0.0275
+cache_write = 0.091663
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.1-pro-preview.toml b/providers/requesty/models/gemini-3.1-pro-preview.toml
index 56dc44aa600..45392c2634d 100644
--- a/providers/requesty/models/gemini-3.1-pro-preview.toml
+++ b/providers/requesty/models/gemini-3.1-pro-preview.toml
@@ -8,17 +8,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 10.8
-cache_read = 0.18
-cache_write = 4.05
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 4.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 3.6
-output = 16.2
-cache_read = 0.18
-cache_write = 4.05
+input = 4
+output = 18
+cache_read = 0.2
+cache_write = 4.5
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.5-flash-lite.toml b/providers/requesty/models/gemini-3.5-flash-lite.toml
index c6aada13d23..3373ffd068d 100644
--- a/providers/requesty/models/gemini-3.5-flash-lite.toml
+++ b/providers/requesty/models/gemini-3.5-flash-lite.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.27
-output = 2.25
-cache_read = 0.027
+input = 0.3
+output = 2.5
+cache_read = 0.03
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.5-flash-lite@eu.toml b/providers/requesty/models/gemini-3.5-flash-lite@eu.toml
index 41fcc13f5ce..e8db9e82467 100644
--- a/providers/requesty/models/gemini-3.5-flash-lite@eu.toml
+++ b/providers/requesty/models/gemini-3.5-flash-lite@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.297
-output = 2.475
-cache_read = 0.0297
+input = 0.33
+output = 2.75
+cache_read = 0.033
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.5-flash.toml b/providers/requesty/models/gemini-3.5-flash.toml
index 2ecd9693695..905b0d8f732 100644
--- a/providers/requesty/models/gemini-3.5-flash.toml
+++ b/providers/requesty/models/gemini-3.5-flash.toml
@@ -8,10 +8,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.35
-output = 8.1
-cache_read = 0.135
-cache_write = 1.4247
+input = 1.5
+output = 9
+cache_read = 0.15
+cache_write = 1.583
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.5-flash@eu.toml b/providers/requesty/models/gemini-3.5-flash@eu.toml
index 28011a05ce6..60cb1f7244a 100644
--- a/providers/requesty/models/gemini-3.5-flash@eu.toml
+++ b/providers/requesty/models/gemini-3.5-flash@eu.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.485
-output = 8.91
-cache_read = 0.1485
-cache_write = 1.56717
+input = 1.65
+output = 9.9
+cache_read = 0.165
+cache_write = 1.7413
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.6-flash.toml b/providers/requesty/models/gemini-3.6-flash.toml
index 656f28894a4..7d7082c4824 100644
--- a/providers/requesty/models/gemini-3.6-flash.toml
+++ b/providers/requesty/models/gemini-3.6-flash.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.35
-output = 6.3
-cache_read = 0.135
+input = 1.5
+output = 7
+cache_read = 0.15
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.7-flash.toml b/providers/requesty/models/gemini-3.7-flash.toml
index 2f1ca9bfdf1..7a3857db113 100644
--- a/providers/requesty/models/gemini-3.7-flash.toml
+++ b/providers/requesty/models/gemini-3.7-flash.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 3
-cache_read = 0.06
+input = 0.75
+output = 3.75
+cache_read = 0.075
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.7-flash@eu.toml b/providers/requesty/models/gemini-3.7-flash@eu.toml
index 097d1384ab0..c7f3a15dbf3 100644
--- a/providers/requesty/models/gemini-3.7-flash@eu.toml
+++ b/providers/requesty/models/gemini-3.7-flash@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.66
-output = 3.3
-cache_read = 0.066
+input = 0.825
+output = 4.125
+cache_read = 0.0825
 
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemma-4-26b-a4b-it.toml b/providers/requesty/models/gemma-4-26b-a4b-it.toml
index 6c8e260546d..923ea60bfd8 100644
--- a/providers/requesty/models/gemma-4-26b-a4b-it.toml
+++ b/providers/requesty/models/gemma-4-26b-a4b-it.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.063
-output = 0.306
-cache_read = 0.063
+input = 0.07
+output = 0.34
+cache_read = 0.07
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/glm-5.1.toml b/providers/requesty/models/glm-5.1.toml
index 76d3d1d3014..1dd7588e579 100644
--- a/providers/requesty/models/glm-5.1.toml
+++ b/providers/requesty/models/glm-5.1.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.26
-output = 3.96
-cache_read = 0.234
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
 output = 128_000
diff --git a/providers/requesty/models/glm-5.1@eu.toml b/providers/requesty/models/glm-5.1@eu.toml
index 9e261e54bc5..d4fb0081f69 100644
--- a/providers/requesty/models/glm-5.1@eu.toml
+++ b/providers/requesty/models/glm-5.1@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.26
-output = 3.96
-cache_read = 1.26
+input = 1.4
+output = 4.4
+cache_read = 1.4
 
 [limit]
 output = 200_000
diff --git a/providers/requesty/models/glm-5.2-fast.toml b/providers/requesty/models/glm-5.2-fast.toml
index 85fd830a947..48637ee68dd 100644
--- a/providers/requesty/models/glm-5.2-fast.toml
+++ b/providers/requesty/models/glm-5.2-fast.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.89
-output = 5.94
-cache_read = 0.189
+input = 2.1
+output = 6.6
+cache_read = 0.21
 
 [limit]
 context = 1_000_000
diff --git a/providers/requesty/models/glm-5.2.toml b/providers/requesty/models/glm-5.2.toml
index fa677c6eb8b..bab0f699fff 100644
--- a/providers/requesty/models/glm-5.2.toml
+++ b/providers/requesty/models/glm-5.2.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.08
-output = 3.78
-cache_read = 0.234
+input = 1.2
+output = 4.2
+cache_read = 0.26
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/glm-5.2@eu.toml b/providers/requesty/models/glm-5.2@eu.toml
index fe9f804d8bf..7d648ff7b6c 100644
--- a/providers/requesty/models/glm-5.2@eu.toml
+++ b/providers/requesty/models/glm-5.2@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.08
-output = 3.78
-cache_read = 0.234
+input = 1.2
+output = 4.2
+cache_read = 0.26
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
index 93e48f11df3..394d0fef42b 100644
--- a/providers/requesty/models/glm-5.3-flash.toml
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.135
-output = 0.45
-cache_read = 0.027
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 output = 128_000
diff --git a/providers/requesty/models/glm-5.3-flash@eu.toml b/providers/requesty/models/glm-5.3-flash@eu.toml
new file mode 100644
index 00000000000..d58a76089fe
--- /dev/null
+++ b/providers/requesty/models/glm-5.3-flash@eu.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3-Flash (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.07
+
+[limit]
+output = 262_144
diff --git a/providers/requesty/models/glm-5.3.toml b/providers/requesty/models/glm-5.3.toml
index e6d529fafbd..baa2d5f5503 100644
--- a/providers/requesty/models/glm-5.3.toml
+++ b/providers/requesty/models/glm-5.3.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.26
-output = 3.96
-cache_read = 0.234
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
 output = 128_000
diff --git a/providers/requesty/models/gpt-4.1-mini@eu.toml b/providers/requesty/models/gpt-4.1-mini@eu.toml
index f0cc4b9b339..8f8e3096b5d 100644
--- a/providers/requesty/models/gpt-4.1-mini@eu.toml
+++ b/providers/requesty/models/gpt-4.1-mini@eu.toml
@@ -2,6 +2,6 @@ base_model = "openai/gpt-4.1-mini"
 name = "GPT-4.1 mini (EU)"
 
 [cost]
-input = 0.396
-output = 1.584
-cache_read = 0.099
+input = 0.44
+output = 1.76
+cache_read = 0.11
diff --git a/providers/requesty/models/gpt-4.1-nano@eu.toml b/providers/requesty/models/gpt-4.1-nano@eu.toml
index 26ab9bbe24e..69357d3eb93 100644
--- a/providers/requesty/models/gpt-4.1-nano@eu.toml
+++ b/providers/requesty/models/gpt-4.1-nano@eu.toml
@@ -2,6 +2,6 @@ base_model = "openai/gpt-4.1-nano"
 name = "GPT-4.1 nano (EU)"
 
 [cost]
-input = 0.099
-output = 0.396
-cache_read = 0.02475
+input = 0.11
+output = 0.44
+cache_read = 0.0275
diff --git a/providers/requesty/models/gpt-4.1@eu.toml b/providers/requesty/models/gpt-4.1@eu.toml
index 7929c847fcf..2bff8e5fc5d 100644
--- a/providers/requesty/models/gpt-4.1@eu.toml
+++ b/providers/requesty/models/gpt-4.1@eu.toml
@@ -2,6 +2,6 @@ base_model = "openai/gpt-4.1"
 name = "GPT-4.1 (EU)"
 
 [cost]
-input = 1.98
-output = 7.92
-cache_read = 0.495
+input = 2.2
+output = 8.8
+cache_read = 0.55
diff --git a/providers/requesty/models/gpt-4o-mini@eu.toml b/providers/requesty/models/gpt-4o-mini@eu.toml
index c20471482eb..9af1072d8b5 100644
--- a/providers/requesty/models/gpt-4o-mini@eu.toml
+++ b/providers/requesty/models/gpt-4o-mini@eu.toml
@@ -2,9 +2,9 @@ base_model = "openai/gpt-4o-mini"
 name = "GPT-4o mini (EU)"
 
 [cost]
-input = 0.1485
-output = 0.594
-cache_read = 0.07425
+input = 0.165
+output = 0.66
+cache_read = 0.0825
 
 [limit]
 output = 16_000
diff --git a/providers/requesty/models/gpt-5-mini@eu.toml b/providers/requesty/models/gpt-5-mini@eu.toml
index 949de050d68..5ebb06e4d3b 100644
--- a/providers/requesty/models/gpt-5-mini@eu.toml
+++ b/providers/requesty/models/gpt-5-mini@eu.toml
@@ -10,9 +10,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.2475
-output = 1.98
-cache_read = 0.02475
+input = 0.275
+output = 2.2
+cache_read = 0.0275
 
 [limit]
 context = 200_000
diff --git a/providers/requesty/models/gpt-5-nano@eu.toml b/providers/requesty/models/gpt-5-nano@eu.toml
index 6af0f44d3bd..15ad36d1521 100644
--- a/providers/requesty/models/gpt-5-nano@eu.toml
+++ b/providers/requesty/models/gpt-5-nano@eu.toml
@@ -10,9 +10,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.0495
-output = 0.396
-cache_read = 0.00495
+input = 0.055
+output = 0.44
+cache_read = 0.0055
 
 [limit]
 context = 200_000
diff --git a/providers/requesty/models/gpt-5.1@eu.toml b/providers/requesty/models/gpt-5.1@eu.toml
index 7cd1ddb2422..e190b9c134e 100644
--- a/providers/requesty/models/gpt-5.1@eu.toml
+++ b/providers/requesty/models/gpt-5.1@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.2375
-output = 9.9
-cache_read = 0.12375
+input = 1.375
+output = 11
+cache_read = 0.1375
diff --git a/providers/requesty/models/gpt-5.3-chat.toml b/providers/requesty/models/gpt-5.3-chat.toml
index c045c9e2f0f..c2069c739f1 100644
--- a/providers/requesty/models/gpt-5.3-chat.toml
+++ b/providers/requesty/models/gpt-5.3-chat.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.575
-output = 12.6
-cache_read = 0.1575
+input = 1.75
+output = 14
+cache_read = 0.175
 
 [limit]
 context = 128_000
diff --git a/providers/requesty/models/gpt-5.3-codex.toml b/providers/requesty/models/gpt-5.3-codex.toml
index 69bbfb229f8..af89482abab 100644
--- a/providers/requesty/models/gpt-5.3-codex.toml
+++ b/providers/requesty/models/gpt-5.3-codex.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.575
-output = 12.6
-cache_read = 0.1575
+input = 1.75
+output = 14
+cache_read = 0.175
diff --git a/providers/requesty/models/gpt-5.4-mini.toml b/providers/requesty/models/gpt-5.4-mini.toml
index 47ba8aceac2..44fdb6905d9 100644
--- a/providers/requesty/models/gpt-5.4-mini.toml
+++ b/providers/requesty/models/gpt-5.4-mini.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.675
-output = 4.05
-cache_read = 0.0675
+input = 0.75
+output = 4.5
+cache_read = 0.075
diff --git a/providers/requesty/models/gpt-5.4-nano.toml b/providers/requesty/models/gpt-5.4-nano.toml
index de9f1e15ca4..3f0bd3d107a 100644
--- a/providers/requesty/models/gpt-5.4-nano.toml
+++ b/providers/requesty/models/gpt-5.4-nano.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.18
-output = 1.125
-cache_read = 0.018
+input = 0.2
+output = 1.25
+cache_read = 0.02
diff --git a/providers/requesty/models/gpt-5.4-pro.toml b/providers/requesty/models/gpt-5.4-pro.toml
index a0772bd26be..9e8c264fe8a 100644
--- a/providers/requesty/models/gpt-5.4-pro.toml
+++ b/providers/requesty/models/gpt-5.4-pro.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 27
-output = 162
-cache_read = 27
+input = 30
+output = 180
+cache_read = 30
diff --git a/providers/requesty/models/gpt-5.4.toml b/providers/requesty/models/gpt-5.4.toml
index acf18e0aee5..f11cbe1c51e 100644
--- a/providers/requesty/models/gpt-5.4.toml
+++ b/providers/requesty/models/gpt-5.4.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.475
-output = 14.85
-cache_read = 0.2475
+input = 2.75
+output = 16.5
+cache_read = 0.275
diff --git a/providers/requesty/models/gpt-5.4@eu.toml b/providers/requesty/models/gpt-5.4@eu.toml
index 086b0c91c3a..60c008d3a16 100644
--- a/providers/requesty/models/gpt-5.4@eu.toml
+++ b/providers/requesty/models/gpt-5.4@eu.toml
@@ -9,12 +9,12 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.25
-output = 13.5
-cache_read = 0.225
+input = 2.5
+output = 15
+cache_read = 0.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 4.5
-output = 20.25
-cache_read = 0.45
+input = 5
+output = 22.5
+cache_read = 0.5
diff --git a/providers/requesty/models/gpt-5.5-pro.toml b/providers/requesty/models/gpt-5.5-pro.toml
index 0e803920cb9..b1e1efca4e5 100644
--- a/providers/requesty/models/gpt-5.5-pro.toml
+++ b/providers/requesty/models/gpt-5.5-pro.toml
@@ -8,5 +8,5 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 27
-output = 162
+input = 30
+output = 180
diff --git a/providers/requesty/models/gpt-5.5.toml b/providers/requesty/models/gpt-5.5.toml
index 256baf26db9..28f1a5586f2 100644
--- a/providers/requesty/models/gpt-5.5.toml
+++ b/providers/requesty/models/gpt-5.5.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 29.7
-cache_read = 0.495
+input = 5.5
+output = 33
+cache_read = 0.55
diff --git a/providers/requesty/models/gpt-5.5@eu.toml b/providers/requesty/models/gpt-5.5@eu.toml
index 2f37849b152..b0a78ccfad5 100644
--- a/providers/requesty/models/gpt-5.5@eu.toml
+++ b/providers/requesty/models/gpt-5.5@eu.toml
@@ -9,12 +9,12 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.5
-output = 27
-cache_read = 0.45
+input = 5
+output = 30
+cache_read = 0.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 9
-output = 40.5
-cache_read = 0.9
+input = 10
+output = 45
+cache_read = 1
diff --git a/providers/requesty/models/gpt-5.6-luna.toml b/providers/requesty/models/gpt-5.6-luna.toml
index 9623d14e700..1d24843159f 100644
--- a/providers/requesty/models/gpt-5.6-luna.toml
+++ b/providers/requesty/models/gpt-5.6-luna.toml
@@ -8,12 +8,12 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.18
-output = 1.08
-cache_read = 0.018
+input = 0.2
+output = 1.2
+cache_read = 0.02
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 0.36
-output = 1.62
-cache_read = 0.036
+input = 0.4
+output = 1.8
+cache_read = 0.04
diff --git a/providers/requesty/models/gpt-5.6-luna@eu.toml b/providers/requesty/models/gpt-5.6-luna@eu.toml
index 165b2e273d1..9bf48d8107d 100644
--- a/providers/requesty/models/gpt-5.6-luna@eu.toml
+++ b/providers/requesty/models/gpt-5.6-luna@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.198
-output = 1.188
-cache_read = 0.0198
+input = 0.22
+output = 1.32
+cache_read = 0.022
diff --git a/providers/requesty/models/gpt-5.6-sol.toml b/providers/requesty/models/gpt-5.6-sol.toml
index 42276e2c175..47018de319c 100644
--- a/providers/requesty/models/gpt-5.6-sol.toml
+++ b/providers/requesty/models/gpt-5.6-sol.toml
@@ -8,14 +8,14 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 3.6
-output = 18
-cache_read = 0.36
-cache_write = 4.5
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 7.2
-output = 27
-cache_read = 0.72
-cache_write = 9
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10
diff --git a/providers/requesty/models/gpt-5.6-sol@eu.toml b/providers/requesty/models/gpt-5.6-sol@eu.toml
index 5a42d72505e..ae290093a78 100644
--- a/providers/requesty/models/gpt-5.6-sol@eu.toml
+++ b/providers/requesty/models/gpt-5.6-sol@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 4.95
-output = 29.7
-cache_read = 0.495
+input = 5.5
+output = 33
+cache_read = 0.55
diff --git a/providers/requesty/models/gpt-5.6-terra.toml b/providers/requesty/models/gpt-5.6-terra.toml
index c58e6be96d6..72544acd7e7 100644
--- a/providers/requesty/models/gpt-5.6-terra.toml
+++ b/providers/requesty/models/gpt-5.6-terra.toml
@@ -8,12 +8,12 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 10.8
-cache_read = 0.18
+input = 2
+output = 12
+cache_read = 0.2
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 3.6
-output = 16.2
-cache_read = 0.36
+input = 4
+output = 18
+cache_read = 0.4
diff --git a/providers/requesty/models/gpt-5.6-terra@eu.toml b/providers/requesty/models/gpt-5.6-terra@eu.toml
index 88b926e3c21..d75520d2460 100644
--- a/providers/requesty/models/gpt-5.6-terra@eu.toml
+++ b/providers/requesty/models/gpt-5.6-terra@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.98
-output = 11.88
-cache_read = 0.198
+input = 2.2
+output = 13.2
+cache_read = 0.22
diff --git a/providers/requesty/models/gpt-5@eu.toml b/providers/requesty/models/gpt-5@eu.toml
index e694ec44140..c03d73bc598 100644
--- a/providers/requesty/models/gpt-5@eu.toml
+++ b/providers/requesty/models/gpt-5@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.2375
-output = 9.9
-cache_read = 0.12375
+input = 1.375
+output = 11
+cache_read = 0.1375
diff --git a/providers/requesty/models/grok-4.2-beta.toml b/providers/requesty/models/grok-4.2-beta.toml
index bd3edb62044..6577aa20daa 100644
--- a/providers/requesty/models/grok-4.2-beta.toml
+++ b/providers/requesty/models/grok-4.2-beta.toml
@@ -17,17 +17,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 5.4
-cache_read = 0.18
-cache_write = 1.8
+input = 2
+output = 6
+cache_read = 0.2
+cache_write = 2
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 3.6
-output = 10.8
-cache_read = 0.36
-cache_write = 3.6
+input = 4
+output = 12
+cache_read = 0.4
+cache_write = 4
 
 [limit]
 context = 2_000_000
diff --git a/providers/requesty/models/grok-4.3.toml b/providers/requesty/models/grok-4.3.toml
index 19ecb03eaa7..c9f02e4eef9 100644
--- a/providers/requesty/models/grok-4.3.toml
+++ b/providers/requesty/models/grok-4.3.toml
@@ -8,17 +8,17 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.125
-output = 2.25
-cache_read = 0.18
-cache_write = 1.125
+input = 1.25
+output = 2.5
+cache_read = 0.2
+cache_write = 1.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 2.25
-output = 4.5
-cache_read = 0.36
-cache_write = 2.25
+input = 2.5
+output = 5
+cache_read = 0.4
+cache_write = 2.5
 
 [limit]
 output = 1_000_000
diff --git a/providers/requesty/models/grok-4.5.toml b/providers/requesty/models/grok-4.5.toml
index 4e20a50c6a5..3d6bd0b85da 100644
--- a/providers/requesty/models/grok-4.5.toml
+++ b/providers/requesty/models/grok-4.5.toml
@@ -8,14 +8,14 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 5.4
-cache_read = 0.45
-cache_write = 1.8
+input = 2
+output = 6
+cache_read = 0.5
+cache_write = 2
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 3.6
-output = 10.8
-cache_read = 0.9
-cache_write = 3.6
+input = 4
+output = 12
+cache_read = 1
+cache_write = 4
diff --git a/providers/requesty/models/grok-4.6.toml b/providers/requesty/models/grok-4.6.toml
index 3d9a628cded..9f48a48a76d 100644
--- a/providers/requesty/models/grok-4.6.toml
+++ b/providers/requesty/models/grok-4.6.toml
@@ -8,14 +8,14 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 5.4
-cache_read = 0.45
-cache_write = 1.8
+input = 2
+output = 6
+cache_read = 0.5
+cache_write = 2
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 3.6
-output = 10.8
-cache_read = 0.9
-cache_write = 3.6
+input = 4
+output = 12
+cache_read = 1
+cache_write = 4
diff --git a/providers/requesty/models/grok-build-0.1.toml b/providers/requesty/models/grok-build-0.1.toml
index 580edcd07f4..1fbae26140f 100644
--- a/providers/requesty/models/grok-build-0.1.toml
+++ b/providers/requesty/models/grok-build-0.1.toml
@@ -2,6 +2,6 @@ base_model = "xai/grok-build-0.1"
 reasoning = false
 
 [cost]
-input = 0.9
-output = 1.8
-cache_read = 0.09
+input = 1
+output = 2
+cache_read = 0.1
diff --git a/providers/requesty/models/hy3.toml b/providers/requesty/models/hy3.toml
index d8e372d57f5..c6be2366f65 100644
--- a/providers/requesty/models/hy3.toml
+++ b/providers/requesty/models/hy3.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.126
-output = 0.522
-cache_read = 0.0315
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/inkling-256k.toml b/providers/requesty/models/inkling-256k.toml
index fd1df26c03e..45e77a32fba 100644
--- a/providers/requesty/models/inkling-256k.toml
+++ b/providers/requesty/models/inkling-256k.toml
@@ -16,9 +16,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.496
-output = 3.744
-cache_read = 0.2992
+input = 1.87
+output = 4.68
+cache_read = 0.374
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/inkling.toml b/providers/requesty/models/inkling.toml
index 428a090baf4..65b2371a24f 100644
--- a/providers/requesty/models/inkling.toml
+++ b/providers/requesty/models/inkling.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.683
-output = 4.212
-cache_read = 0.3366
+input = 1.87
+output = 4.68
+cache_read = 0.374
 
 [limit]
 context = 65_536
diff --git a/providers/requesty/models/kat-coder-pro.toml b/providers/requesty/models/kat-coder-pro.toml
index c390576ce4f..3ec47a7db38 100644
--- a/providers/requesty/models/kat-coder-pro.toml
+++ b/providers/requesty/models/kat-coder-pro.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0.27
-output = 1.08
+input = 0.3
+output = 1.2
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/kimi-k2.6.toml b/providers/requesty/models/kimi-k2.6.toml
index 279bf6221bb..e29bfdedc0c 100644
--- a/providers/requesty/models/kimi-k2.6.toml
+++ b/providers/requesty/models/kimi-k2.6.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.855
-output = 3.6
-cache_read = 0.144
+input = 0.95
+output = 4
+cache_read = 0.16
diff --git a/providers/requesty/models/kimi-k2.6@eu.toml b/providers/requesty/models/kimi-k2.6@eu.toml
index 9ce827e892c..9770226aa2b 100644
--- a/providers/requesty/models/kimi-k2.6@eu.toml
+++ b/providers/requesty/models/kimi-k2.6@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.855
-output = 3.6
-cache_read = 0.855
+input = 0.95
+output = 4
+cache_read = 0.95
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/kimi-k2.7-code.toml b/providers/requesty/models/kimi-k2.7-code.toml
index cd6be544d01..7dae44e6ff6 100644
--- a/providers/requesty/models/kimi-k2.7-code.toml
+++ b/providers/requesty/models/kimi-k2.7-code.toml
@@ -8,6 +8,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.855
-output = 3.6
-cache_read = 0.171
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/requesty/models/kimi-k2.7-code@eu.toml b/providers/requesty/models/kimi-k2.7-code@eu.toml
index 9037afe922f..411c0773552 100644
--- a/providers/requesty/models/kimi-k2.7-code@eu.toml
+++ b/providers/requesty/models/kimi-k2.7-code@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.125
-output = 4.05
-cache_read = 0.279
+input = 1.25
+output = 4.5
+cache_read = 0.31
diff --git a/providers/requesty/models/kimi-k3.toml b/providers/requesty/models/kimi-k3.toml
index 3bb366be140..b1999cda9f0 100644
--- a/providers/requesty/models/kimi-k3.toml
+++ b/providers/requesty/models/kimi-k3.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.025
-output = 10.125
-cache_read = 0.2025
+input = 2.25
+output = 11.25
+cache_read = 0.225
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/kimi-k3@eu.toml b/providers/requesty/models/kimi-k3@eu.toml
index 92428865b62..bb3b723880d 100644
--- a/providers/requesty/models/kimi-k3@eu.toml
+++ b/providers/requesty/models/kimi-k3@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.025
-output = 10.125
-cache_read = 0.2025
+input = 2.25
+output = 11.25
+cache_read = 0.225
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/ling-2.6-1t.toml b/providers/requesty/models/ling-2.6-1t.toml
index 6d03156b1dd..c85bef2c3f2 100644
--- a/providers/requesty/models/ling-2.6-1t.toml
+++ b/providers/requesty/models/ling-2.6-1t.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0.27
-output = 2.25
+input = 0.3
+output = 2.5
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/ling-2.6-flash.toml b/providers/requesty/models/ling-2.6-flash.toml
index fbcfbe27c1b..87bbeddcbf2 100644
--- a/providers/requesty/models/ling-2.6-flash.toml
+++ b/providers/requesty/models/ling-2.6-flash.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0.09
-output = 0.27
+input = 0.1
+output = 0.3
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/mimo-v2.5-pro.toml b/providers/requesty/models/mimo-v2.5-pro.toml
index 80847758912..96a7a3599ce 100644
--- a/providers/requesty/models/mimo-v2.5-pro.toml
+++ b/providers/requesty/models/mimo-v2.5-pro.toml
@@ -3,6 +3,6 @@ reasoning = false
 structured_output = false
 
 [cost]
-input = 0.3915
-output = 0.783
-cache_read = 0.00324
+input = 0.435
+output = 0.87
+cache_read = 0.0036
diff --git a/providers/requesty/models/mimo-v2.5.toml b/providers/requesty/models/mimo-v2.5.toml
index 7c4ac60eabf..4ba7b3143bc 100644
--- a/providers/requesty/models/mimo-v2.5.toml
+++ b/providers/requesty/models/mimo-v2.5.toml
@@ -3,6 +3,6 @@ reasoning = false
 structured_output = false
 
 [cost]
-input = 0.126
-output = 0.252
-cache_read = 0.00252
+input = 0.14
+output = 0.28
+cache_read = 0.0028
diff --git a/providers/requesty/models/minimax-m2.7-highspeed.toml b/providers/requesty/models/minimax-m2.7-highspeed.toml
index 505b7a466e9..d13335dcdc7 100644
--- a/providers/requesty/models/minimax-m2.7-highspeed.toml
+++ b/providers/requesty/models/minimax-m2.7-highspeed.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.54
-output = 2.16
-cache_read = 0.054
-cache_write = 1.08
+input = 0.6
+output = 2.4
+cache_read = 0.06
+cache_write = 1.2
 
 [limit]
 context = 200_000
diff --git a/providers/requesty/models/minimax-m2.7.toml b/providers/requesty/models/minimax-m2.7.toml
index b650b63f65a..2245aa8535d 100644
--- a/providers/requesty/models/minimax-m2.7.toml
+++ b/providers/requesty/models/minimax-m2.7.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.27
-output = 1.08
-cache_read = 0.054
-cache_write = 1.08
+input = 0.3
+output = 1.2
+cache_read = 0.06
+cache_write = 1.2
 
 [limit]
 context = 200_000
diff --git a/providers/requesty/models/minimax-m3.toml b/providers/requesty/models/minimax-m3.toml
index d726705568a..370ba1ce3df 100644
--- a/providers/requesty/models/minimax-m3.toml
+++ b/providers/requesty/models/minimax-m3.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.24
-output = 0.96
-cache_read = 0.048
+input = 0.3
+output = 1.2
+cache_read = 0.06
 
 [limit]
 context = 1_000_000
diff --git a/providers/requesty/models/minimax-m3@eu.toml b/providers/requesty/models/minimax-m3@eu.toml
index 2b240e1d13b..404aceb28f1 100644
--- a/providers/requesty/models/minimax-m3@eu.toml
+++ b/providers/requesty/models/minimax-m3@eu.toml
@@ -10,9 +10,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.36
-output = 1.8
-cache_read = 0.09
+input = 0.4
+output = 2
+cache_read = 0.1
 
 [limit]
 output = 1_048_576
diff --git a/providers/requesty/models/mistral-medium-3-5.toml b/providers/requesty/models/mistral-medium-3-5.toml
index cd3cc407a6e..846955e7dcf 100644
--- a/providers/requesty/models/mistral-medium-3-5.toml
+++ b/providers/requesty/models/mistral-medium-3-5.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.485
-output = 7.425
-cache_read = 1.485
+input = 1.65
+output = 8.25
+cache_read = 1.65
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/mistral-medium-3-5@eu.toml b/providers/requesty/models/mistral-medium-3-5@eu.toml
index ad3d129f3a1..41dfecc58cb 100644
--- a/providers/requesty/models/mistral-medium-3-5@eu.toml
+++ b/providers/requesty/models/mistral-medium-3-5@eu.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.485
-output = 7.425
-cache_read = 1.485
+input = 1.65
+output = 8.25
+cache_read = 1.65
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/mistral-medium-latest.toml b/providers/requesty/models/mistral-medium-latest.toml
index 1cc3aa69115..e841391c811 100644
--- a/providers/requesty/models/mistral-medium-latest.toml
+++ b/providers/requesty/models/mistral-medium-latest.toml
@@ -2,9 +2,9 @@ base_model = "mistral/mistral-medium-latest"
 reasoning = false
 
 [cost]
-input = 0.396
-output = 1.98
-cache_read = 0.396
+input = 0.44
+output = 2.2
+cache_read = 0.44
 
 [limit]
 context = 131_072
diff --git a/providers/requesty/models/mistral-medium-latest@eu.toml b/providers/requesty/models/mistral-medium-latest@eu.toml
index 5ca123b9452..406cd97163f 100644
--- a/providers/requesty/models/mistral-medium-latest@eu.toml
+++ b/providers/requesty/models/mistral-medium-latest@eu.toml
@@ -3,9 +3,9 @@ name = "Mistral Medium (latest) (EU)"
 reasoning = false
 
 [cost]
-input = 0.396
-output = 1.98
-cache_read = 0.396
+input = 0.44
+output = 2.2
+cache_read = 0.44
 
 [limit]
 context = 131_072
diff --git a/providers/requesty/models/mistral-small-2603.toml b/providers/requesty/models/mistral-small-2603.toml
index cae2472d897..7a4b26451ff 100644
--- a/providers/requesty/models/mistral-small-2603.toml
+++ b/providers/requesty/models/mistral-small-2603.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.1485
-output = 0.594
-cache_read = 0.1485
+input = 0.165
+output = 0.66
+cache_read = 0.165
diff --git a/providers/requesty/models/mistral-small-2603@eu.toml b/providers/requesty/models/mistral-small-2603@eu.toml
index 345822d0ae6..66339a0378a 100644
--- a/providers/requesty/models/mistral-small-2603@eu.toml
+++ b/providers/requesty/models/mistral-small-2603@eu.toml
@@ -10,6 +10,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.1485
-output = 0.594
-cache_read = 0.1485
+input = 0.165
+output = 0.66
+cache_read = 0.165
diff --git a/providers/requesty/models/nemotron-3-nano-omni.toml b/providers/requesty/models/nemotron-3-nano-omni.toml
index 575ee74c903..f8ca025e68f 100644
--- a/providers/requesty/models/nemotron-3-nano-omni.toml
+++ b/providers/requesty/models/nemotron-3-nano-omni.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.054
-output = 0.216
-cache_read = 0.054
+input = 0.06
+output = 0.24
+cache_read = 0.06
 
 [limit]
 context = 300_000
diff --git a/providers/requesty/models/nemotron-3-nano-omni@eu.toml b/providers/requesty/models/nemotron-3-nano-omni@eu.toml
index 68fc6a8b2e4..1422a67a67c 100644
--- a/providers/requesty/models/nemotron-3-nano-omni@eu.toml
+++ b/providers/requesty/models/nemotron-3-nano-omni@eu.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.054
-output = 0.216
-cache_read = 0.054
+input = 0.06
+output = 0.24
+cache_read = 0.06
 
 [limit]
 context = 300_000
diff --git a/providers/requesty/models/nemotron-3-ultra-nvfp4.toml b/providers/requesty/models/nemotron-3-ultra-nvfp4.toml
index b0cbe8dfa37..3fc92c2cc8b 100644
--- a/providers/requesty/models/nemotron-3-ultra-nvfp4.toml
+++ b/providers/requesty/models/nemotron-3-ultra-nvfp4.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.54
-output = 2.16
-cache_read = 0.108
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/nemotron-lightning-3.5-30b-a3b.toml b/providers/requesty/models/nemotron-lightning-3.5-30b-a3b.toml
index 34588b5cc51..d55a5302b32 100644
--- a/providers/requesty/models/nemotron-lightning-3.5-30b-a3b.toml
+++ b/providers/requesty/models/nemotron-lightning-3.5-30b-a3b.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.045
-output = 0.18
-cache_read = 0.009
+input = 0.05
+output = 0.2
+cache_read = 0.01
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/nvidia-nemotron-3-super-120b-a12b.toml b/providers/requesty/models/nvidia-nemotron-3-super-120b-a12b.toml
index 6a0e782c9ee..d7fbf31d042 100644
--- a/providers/requesty/models/nvidia-nemotron-3-super-120b-a12b.toml
+++ b/providers/requesty/models/nvidia-nemotron-3-super-120b-a12b.toml
@@ -17,8 +17,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.09
-output = 0.45
+input = 0.1
+output = 0.5
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/nvidia-nemotron-3-ultra.toml b/providers/requesty/models/nvidia-nemotron-3-ultra.toml
index 844902a8626..92717df9c6f 100644
--- a/providers/requesty/models/nvidia-nemotron-3-ultra.toml
+++ b/providers/requesty/models/nvidia-nemotron-3-ultra.toml
@@ -17,8 +17,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 2.25
+input = 0.5
+output = 2.5
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/o4-mini@eu.toml b/providers/requesty/models/o4-mini@eu.toml
index 39db9fef85e..7e8324087de 100644
--- a/providers/requesty/models/o4-mini@eu.toml
+++ b/providers/requesty/models/o4-mini@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.089
-output = 4.356
-cache_read = 0.27225
+input = 1.21
+output = 4.84
+cache_read = 0.3025
diff --git a/providers/requesty/models/qwen3.5-27b.toml b/providers/requesty/models/qwen3.5-27b.toml
index bb7844b8e91..e5366eb2b08 100644
--- a/providers/requesty/models/qwen3.5-27b.toml
+++ b/providers/requesty/models/qwen3.5-27b.toml
@@ -8,8 +8,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.234
-output = 2.34
+input = 0.26
+output = 2.6
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/qwen3.5-2b.toml b/providers/requesty/models/qwen3.5-2b.toml
index dcb8bf202d0..781016ffc86 100644
--- a/providers/requesty/models/qwen3.5-2b.toml
+++ b/providers/requesty/models/qwen3.5-2b.toml
@@ -17,8 +17,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.018
-output = 0.09
+input = 0.02
+output = 0.1
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/qwen3.5-35b-a3b.toml b/providers/requesty/models/qwen3.5-35b-a3b.toml
index 568c5343e3e..ca01a2e54e1 100644
--- a/providers/requesty/models/qwen3.5-35b-a3b.toml
+++ b/providers/requesty/models/qwen3.5-35b-a3b.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.126
-output = 0.9
-cache_read = 0.045
+input = 0.14
+output = 1
+cache_read = 0.05
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/qwen3.6-plus.toml b/providers/requesty/models/qwen3.6-plus.toml
index fac83531f59..c9b20433cd1 100644
--- a/providers/requesty/models/qwen3.6-plus.toml
+++ b/providers/requesty/models/qwen3.6-plus.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 2.7
-cache_read = 0.045
-cache_write = 0.5625
+input = 0.5
+output = 3
+cache_read = 0.05
+cache_write = 0.625
diff --git a/providers/requesty/models/qwen3.7-max.toml b/providers/requesty/models/qwen3.7-max.toml
index b6205364fb9..b0d779243e7 100644
--- a/providers/requesty/models/qwen3.7-max.toml
+++ b/providers/requesty/models/qwen3.7-max.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.25
-output = 6.75
-cache_read = 0.225
-cache_write = 2.8125
+input = 2.5
+output = 7.5
+cache_read = 0.25
+cache_write = 3.125
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/qwen3.7-plus.toml b/providers/requesty/models/qwen3.7-plus.toml
index 7e53c0a59df..009846440eb 100644
--- a/providers/requesty/models/qwen3.7-plus.toml
+++ b/providers/requesty/models/qwen3.7-plus.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.28
-output = 1.12
-cache_read = 0.028
-cache_write = 0.35
+input = 0.32
+output = 1.28
+cache_read = 0.032
+cache_write = 0.4
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/qwen3.8-2.4T-A95B.toml b/providers/requesty/models/qwen3.8-2.4T-A95B.toml
index 02cdb3ba648..a28dc96e242 100644
--- a/providers/requesty/models/qwen3.8-2.4T-A95B.toml
+++ b/providers/requesty/models/qwen3.8-2.4T-A95B.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 5.4
-cache_read = 0.18
+input = 2
+output = 6
+cache_read = 0.2
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/qwen3.8-max.toml b/providers/requesty/models/qwen3.8-max.toml
index bfc96dcb1e9..c9f9d76a6e8 100644
--- a/providers/requesty/models/qwen3.8-max.toml
+++ b/providers/requesty/models/qwen3.8-max.toml
@@ -9,10 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.8
-output = 5.4
-cache_read = 0.225
-cache_write = 2.25
+input = 2
+output = 6
+cache_read = 0.25
+cache_write = 2.5
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/ring-2.6-1t.toml b/providers/requesty/models/ring-2.6-1t.toml
index fed1a0ad2c7..8f33bc84633 100644
--- a/providers/requesty/models/ring-2.6-1t.toml
+++ b/providers/requesty/models/ring-2.6-1t.toml
@@ -17,8 +17,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.27
-output = 2.25
+input = 0.3
+output = 2.5
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/seed-1.8.toml b/providers/requesty/models/seed-1.8.toml
index 4b8470af3d8..2472c31bb2a 100644
--- a/providers/requesty/models/seed-1.8.toml
+++ b/providers/requesty/models/seed-1.8.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.225
-output = 1.8
-cache_read = 0.045
+input = 0.25
+output = 2
+cache_read = 0.05
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/seed-2.0-code.toml b/providers/requesty/models/seed-2.0-code.toml
index 252f4fa5c1d..c64d1a091ab 100644
--- a/providers/requesty/models/seed-2.0-code.toml
+++ b/providers/requesty/models/seed-2.0-code.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 2.7
-cache_read = 0.09
+input = 0.5
+output = 3
+cache_read = 0.1
 
 [limit]
 context = 256_000
diff --git a/providers/requesty/models/seed-2.0-mini.toml b/providers/requesty/models/seed-2.0-mini.toml
index 86f36d0d349..4e3dfb0ea14 100644
--- a/providers/requesty/models/seed-2.0-mini.toml
+++ b/providers/requesty/models/seed-2.0-mini.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.09
-output = 0.36
-cache_read = 0.018
+input = 0.1
+output = 0.4
+cache_read = 0.02
 
 [limit]
 output = 256_000
diff --git a/providers/requesty/models/seed-2.0-pro.toml b/providers/requesty/models/seed-2.0-pro.toml
index f4d76fe612e..b7091c5fb8c 100644
--- a/providers/requesty/models/seed-2.0-pro.toml
+++ b/providers/requesty/models/seed-2.0-pro.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 2.7
-cache_read = 0.09
+input = 0.5
+output = 3
+cache_read = 0.1
 
 [limit]
 output = 256_000
diff --git a/providers/requesty/models/step-3.7-flash.toml b/providers/requesty/models/step-3.7-flash.toml
index cd2b74f8b6b..10e1b1afba5 100644
--- a/providers/requesty/models/step-3.7-flash.toml
+++ b/providers/requesty/models/step-3.7-flash.toml
@@ -10,9 +10,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.18
-output = 1.035
-cache_read = 0.036
+input = 0.2
+output = 1.15
+cache_read = 0.04
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
index 9810394d7c2..b34410d395c 100644
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
+++ b/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.36
-output = 2.7
-cache_read = 0.234
+input = 0.4
+output = 3
+cache_read = 0.26
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
index 0f5cc42ea4a..ba21875fd54 100644
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
+++ b/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
@@ -17,9 +17,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.36
-output = 2.7
-cache_read = 0.234
+input = 0.4
+output = 3
+cache_read = 0.26
 
 [limit]
 context = 262_144

From 309c1bd857d9e503718dead70629e63f321a8de9 Mon Sep 17 00:00:00 2001
From: Rubin Simons <me@rubin55.org>
Date: Fri, 28 Aug 2026 03:26:50 +0300
Subject: [PATCH 0091/2012] Update orcarouter catalog (#5637)

* fix(orcarouter): remove models no longer served

Absent from both https://api.orcarouter.ai/v1/models and
https://api.orcarouter.ai/api/pricing as of 2026-08-27.

* feat(orcarouter): add Fusion and free router models

Adds fusion, fusion-mini, fusion-flash and free, all present in
/v1/models. Descriptions come from the public model pages.

The Fusion slugs carry no [cost]: they bill as the sum of the models
actually run, at upstream cost, no markup.

https://www.orcarouter.ai/models/orcarouter/fusion
https://docs.orcarouter.ai/routing/fusion
https://docs.orcarouter.ai/routing/free-models

* feat(orcarouter): add models missing from the catalog

Adds 38 models served by OrcaRouter, currently absent from models.dev,
including Claude Opus 5 / Sonnet 5 / Fable 5, GPT-5.6 luna/sol/terra,
Gemini 3.5 and 3.6 Flash, Grok 4.5 / 4.6, Kimi K3, MiniMax M3, GLM 5.2 /
5.3, the Qwen3.7 and Qwen3.8 lines, Hy3 and the -free variants.

Costs come from https://api.orcarouter.ai/api/pricing, whose ratios
convert as input = model_ratio * 2 and output = input *
completion_ratio, verified against prompt_per_million in
https://api.orcarouter.ai/v1/models.

reasoning_options mirror each model's lab entry. Anthropic entries
use low/medium/high, the three levels the gateway documents mapping
onto Claude thinking budgets.

https://docs.orcarouter.ai/advanced/reasoning

* fix(orcarouter): refresh pricing and limits from /api/pricing

Prices derived from the ratio catalog as input = model_ratio * 2,
output = input * completion_ratio and cache_* = input * cache_ratio,
verified against prompt_per_million in /v1/models.

I treated the endpoint as authoritative for prices.

* fix(orcarouter): correct tiered base prices

Five entries had the >200K/272K tier's input and output copied into
the base [cost] row, so the high band was applied from the first
token. cache_read was left at the correct base rate.

gpt-5.4, gpt-5.4-pro, gemini-2.5-pro, gemini-3.1-pro-preview and
gemini-3.1-pro-preview-customtools now match providers/openai and
providers/google for both the base row and the tier.

* fix(orcarouter): author real reasoning controls

Replaces reasoning_options = [] on 41 models that do expose caller
control. Entries are derived from each model's first-party provider
entry and confirmed against supported_parameters in /api/pricing.

Also corrects OpenAI effort sets: gpt-5-pro accepts only high, and
the .2/.4/.5-pro models are medium/high/xhigh rather than
low/medium/high.

Leaves [] for: deepseek-reasoner (effort is a documented no-op), the
Gemma models, and the chat-latest aliases.

Keeps the Gemini 2.5 effort controls rather than swapping them for
the native thinkingBudget, since the gateway maps reasoning_effort
onto thinkingConfig.

Adds the leading toggle wire-path comments mentioned as required.

* fix(orcarouter): use the lab control surface for qwen3.7-flash

The effort list none|high came from relay peers. The first-party CN
host providers/alibaba-cn and OpenRouter both expose enable_thinking
plus thinking_budget.

Also makes the file consistent with qwen3.7-plus and qwen3.7-max in
this provider, which already use toggle plus budget_tokens.

* fix(orcarouter): drop minimal from gemini-flash-latest effort

The alias tracks Gemini 3.7 Flash, whose effort set is low, medium
and high. minimal belongs to gemini-flash-lite-latest, which tracks
3.5-flash-lite.

Note that the host prices this alias below Googles own flash-latest
rate, which /api/pricing and /v1/models both report as 0.5/3.

* fix(orcarouter): declare interleaved on dated DeepSeek entries

deepseek-v4-flash-0731 and deepseek-v4-pro-0813 omitted
[interleaved]. Neither has a first-party provider entry to inherit
it from, unlike their undated siblings, so both were generated
without the reasoning_content side channel their model returns.

* docs(orcarouter): note deepseek-reasoner cache rate

Its rate of 0.147/0.295 comes straight from prompt_per_million and
completion_per_million in /v1/models, and its cache_ratio of
0.19047619 differs from the 0.136054422 that deepseek-chat and
deepseek-v4-flash carry. That is why cache_read stays at 0.028 while
the siblings moved to 0.02.

* docs(orcarouter): note claude-opus-5 cache write rate

The host bills cache write at 2x input for this model, against 1.25x
for every other Claude entry it serves. Confirmed by
create_cache_ratio = 2 in /api/pricing and by the public model page,
which lists Cache write / 1M $10.00.

* fix(orcarouter): correct Gemini audio input rates

gemini-flash-latest carried input_audio = 1.5, derived from the
superseded 1.5 input price. /api/pricing publishes no audio_ratio for
that alias, so the field is dropped.

Adds the audio rates to: gemini-3.1-flash-lite 0.5, gemini-3.5-flash 3
and gemini-3.1-pro-preview 2.

Notes that gemini-3.6-flash is billed at 2x Googles own rate, which
/api/pricing and /v1/models both confirm.

* docs(orcarouter): explain the Fusion cost omission

The Fusion slugs bill as the sum of the panel legs actually run, at
upstream cost with no markup, so there is no per-token rate to
publish. AGENTS.md allows omitting [cost] for that case, but the
omission was indistinguishable from missing data.

* fix(orcarouter): drop invented toggle on qwen3-vl thinking SKU

This is a dedicated thinking checkpoint paired with
qwen3-vl-235b-a22b-instruct, and PR documentation treats separate
-thinking/instruct ids as not constituting a toggle.

The toggle came from merge-gateway, which is alone in declaring one:
eight peers use [] and nano-gpt uses budget-only with the same 1024
to 81920 bounds. Keeps budget_tokens, which OrcaRouter advertises as
thinking_budget.
---
 .../models/anthropic/claude-fable-5.toml      |  8 +++++++
 .../models/anthropic/claude-haiku-4.5.toml    |  2 +-
 .../models/anthropic/claude-opus-4.1.toml     |  8 -------
 .../models/anthropic/claude-opus-4.5.toml     |  2 +-
 .../models/anthropic/claude-opus-4.8.toml     |  8 +++++++
 .../models/anthropic/claude-opus-4.toml       |  8 -------
 .../models/anthropic/claude-opus-5.toml       | 10 ++++++++
 .../models/anthropic/claude-sonnet-4.5.toml   |  5 +++-
 .../models/anthropic/claude-sonnet-4.toml     |  8 -------
 .../models/anthropic/claude-sonnet-5.toml     |  8 +++++++
 .../models/deepseek/deepseek-chat.toml        |  6 ++---
 .../models/deepseek/deepseek-reasoner.toml    |  6 +++--
 .../deepseek/deepseek-v4-flash-0731.toml      | 14 +++++++++++
 .../deepseek/deepseek-v4-flash-free.toml      | 14 +++++++++++
 .../deepseek-v4-flash-vision-exp.toml         | 14 +++++++++++
 .../models/deepseek/deepseek-v4-flash.toml    | 12 ++++++----
 .../models/deepseek/deepseek-v4-pro-0813.toml | 14 +++++++++++
 .../models/deepseek/deepseek-v4-pro.toml      | 12 ++++++----
 .../models/google/gemini-2.5-pro.toml         |  4 ++--
 .../models/google/gemini-3-flash-preview.toml |  2 +-
 .../models/google/gemini-3-pro-preview.toml   | 13 ----------
 .../google/gemini-3.1-flash-lite-preview.toml |  2 +-
 .../models/google/gemini-3.1-flash-lite.toml  |  8 +++++++
 .../gemini-3.1-pro-preview-customtools.toml   |  4 ++--
 .../models/google/gemini-3.1-pro-preview.toml |  5 ++--
 .../models/google/gemini-3.5-flash-lite.toml  |  7 ++++++
 .../models/google/gemini-3.5-flash.toml       |  9 +++++++
 .../models/google/gemini-3.6-flash.toml       |  9 +++++++
 .../models/google/gemini-flash-latest.toml    | 12 +++++-----
 .../gemini-robotics-er-1.6-preview.toml       | 10 ++++++++
 .../models/google/gemma-4-26b-a4b-it.toml     |  1 +
 .../models/google/gemma-4-31b-it.toml         |  1 +
 .../orcarouter/models/grok/grok-4.3.toml      |  2 +-
 .../orcarouter/models/grok/grok-4.5.toml      |  7 ++++++
 .../orcarouter/models/grok/grok-4.6.toml      |  7 ++++++
 .../orcarouter/models/kimi/kimi-k2.5.toml     |  7 +++++-
 .../orcarouter/models/kimi/kimi-k2.6.toml     |  6 ++++-
 .../models/kimi/kimi-k2.7-code.toml           | 10 ++++++++
 providers/orcarouter/models/kimi/kimi-k3.toml | 14 +++++++++++
 .../models/meta/muse-spark-1.1.toml           |  7 ++++++
 .../models/meta/muse-spark-1.2.toml           |  7 ++++++
 .../minimax/minimax-m2.5-highspeed.toml       |  2 +-
 .../orcarouter/models/minimax/minimax-m3.toml |  8 +++++++
 .../models/openai/gpt-5-chat-latest.toml      |  3 +++
 .../orcarouter/models/openai/gpt-5-codex.toml |  7 ------
 .../orcarouter/models/openai/gpt-5-mini.toml  |  2 +-
 .../orcarouter/models/openai/gpt-5-nano.toml  |  2 +-
 .../orcarouter/models/openai/gpt-5-pro.toml   |  2 +-
 .../models/openai/gpt-5.1-codex-max.toml      |  7 ------
 .../models/openai/gpt-5.1-codex-mini.toml     |  2 +-
 .../models/openai/gpt-5.1-codex.toml          |  2 +-
 .../orcarouter/models/openai/gpt-5.1.toml     |  2 +-
 .../models/openai/gpt-5.2-chat-latest.toml    |  2 +-
 .../models/openai/gpt-5.2-codex.toml          |  2 +-
 .../orcarouter/models/openai/gpt-5.2-pro.toml |  2 +-
 .../orcarouter/models/openai/gpt-5.2.toml     |  2 +-
 .../models/openai/gpt-5.3-chat-latest.toml    |  6 -----
 .../models/openai/gpt-5.3-codex.toml          |  2 +-
 .../models/openai/gpt-5.4-mini.toml           |  2 +-
 .../models/openai/gpt-5.4-nano.toml           |  2 +-
 .../orcarouter/models/openai/gpt-5.4-pro.toml |  6 ++---
 .../orcarouter/models/openai/gpt-5.4.toml     |  6 ++---
 .../orcarouter/models/openai/gpt-5.5-pro.toml |  5 +++-
 .../orcarouter/models/openai/gpt-5.5.toml     |  2 +-
 .../models/openai/gpt-5.6-luna.toml           |  8 +++++++
 .../orcarouter/models/openai/gpt-5.6-sol.toml |  8 +++++++
 .../models/openai/gpt-5.6-terra.toml          |  8 +++++++
 providers/orcarouter/models/openai/gpt-5.toml |  2 +-
 .../models/openai/gpt-oss-120b.toml           |  6 +++++
 .../orcarouter/models/orcarouter/free.toml    | 24 +++++++++++++++++++
 .../models/orcarouter/fusion-flash.toml       | 22 +++++++++++++++++
 .../models/orcarouter/fusion-mini.toml        | 22 +++++++++++++++++
 .../orcarouter/models/orcarouter/fusion.toml  | 22 +++++++++++++++++
 .../qwen/qwen3-vl-235b-a22b-instruct.toml     |  5 ++++
 .../qwen/qwen3-vl-235b-a22b-thinking.toml     | 10 ++++++++
 .../models/qwen/qwen3.5-122b-a10b.toml        |  6 ++++-
 .../orcarouter/models/qwen/qwen3.5-27b.toml   |  6 ++++-
 .../models/qwen/qwen3.5-35b-a3b.toml          |  6 ++++-
 .../models/qwen/qwen3.5-397b-a17b.toml        |  6 ++++-
 .../orcarouter/models/qwen/qwen3.5-flash.toml | 10 ++++++++
 .../orcarouter/models/qwen/qwen3.5-plus.toml  |  6 ++++-
 .../models/qwen/qwen3.6-35b-a3b.toml          |  6 ++++-
 .../orcarouter/models/qwen/qwen3.6-flash.toml | 10 ++++++++
 .../orcarouter/models/qwen/qwen3.6-plus.toml  |  6 ++++-
 .../orcarouter/models/qwen/qwen3.7-flash.toml | 12 ++++++++++
 .../orcarouter/models/qwen/qwen3.7-max.toml   | 12 ++++++++++
 .../orcarouter/models/qwen/qwen3.7-plus.toml  | 11 +++++++++
 .../models/qwen/qwen3.8-27b-free.toml         | 10 ++++++++
 .../orcarouter/models/qwen/qwen3.8-27b.toml   |  6 +++++
 .../orcarouter/models/qwen/qwen3.8-max.toml   | 16 +++++++++++++
 .../orcarouter/models/tencent/hy3-free.toml   |  7 ++++++
 providers/orcarouter/models/tencent/hy3.toml  |  7 ++++++
 .../orcarouter/models/z-ai/glm-4.5-air.toml   |  3 ++-
 providers/orcarouter/models/z-ai/glm-4.5.toml |  3 ++-
 providers/orcarouter/models/z-ai/glm-4.6.toml |  3 ++-
 providers/orcarouter/models/z-ai/glm-4.7.toml |  3 ++-
 providers/orcarouter/models/z-ai/glm-5.1.toml |  3 ++-
 providers/orcarouter/models/z-ai/glm-5.2.toml | 10 ++++++++
 providers/orcarouter/models/z-ai/glm-5.3.toml | 10 ++++++++
 providers/orcarouter/models/z-ai/glm-5.toml   |  5 ++--
 100 files changed, 584 insertions(+), 127 deletions(-)
 create mode 100644 providers/orcarouter/models/anthropic/claude-fable-5.toml
 delete mode 100644 providers/orcarouter/models/anthropic/claude-opus-4.1.toml
 create mode 100644 providers/orcarouter/models/anthropic/claude-opus-4.8.toml
 delete mode 100644 providers/orcarouter/models/anthropic/claude-opus-4.toml
 create mode 100644 providers/orcarouter/models/anthropic/claude-opus-5.toml
 delete mode 100644 providers/orcarouter/models/anthropic/claude-sonnet-4.toml
 create mode 100644 providers/orcarouter/models/anthropic/claude-sonnet-5.toml
 create mode 100644 providers/orcarouter/models/deepseek/deepseek-v4-flash-0731.toml
 create mode 100644 providers/orcarouter/models/deepseek/deepseek-v4-flash-free.toml
 create mode 100644 providers/orcarouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
 create mode 100644 providers/orcarouter/models/deepseek/deepseek-v4-pro-0813.toml
 delete mode 100644 providers/orcarouter/models/google/gemini-3-pro-preview.toml
 create mode 100644 providers/orcarouter/models/google/gemini-3.1-flash-lite.toml
 create mode 100644 providers/orcarouter/models/google/gemini-3.5-flash-lite.toml
 create mode 100644 providers/orcarouter/models/google/gemini-3.5-flash.toml
 create mode 100644 providers/orcarouter/models/google/gemini-3.6-flash.toml
 create mode 100644 providers/orcarouter/models/google/gemini-robotics-er-1.6-preview.toml
 create mode 100644 providers/orcarouter/models/grok/grok-4.5.toml
 create mode 100644 providers/orcarouter/models/grok/grok-4.6.toml
 create mode 100644 providers/orcarouter/models/kimi/kimi-k2.7-code.toml
 create mode 100644 providers/orcarouter/models/kimi/kimi-k3.toml
 create mode 100644 providers/orcarouter/models/meta/muse-spark-1.1.toml
 create mode 100644 providers/orcarouter/models/meta/muse-spark-1.2.toml
 create mode 100644 providers/orcarouter/models/minimax/minimax-m3.toml
 delete mode 100644 providers/orcarouter/models/openai/gpt-5-codex.toml
 delete mode 100644 providers/orcarouter/models/openai/gpt-5.1-codex-max.toml
 delete mode 100644 providers/orcarouter/models/openai/gpt-5.3-chat-latest.toml
 create mode 100644 providers/orcarouter/models/openai/gpt-5.6-luna.toml
 create mode 100644 providers/orcarouter/models/openai/gpt-5.6-sol.toml
 create mode 100644 providers/orcarouter/models/openai/gpt-5.6-terra.toml
 create mode 100644 providers/orcarouter/models/openai/gpt-oss-120b.toml
 create mode 100644 providers/orcarouter/models/orcarouter/free.toml
 create mode 100644 providers/orcarouter/models/orcarouter/fusion-flash.toml
 create mode 100644 providers/orcarouter/models/orcarouter/fusion-mini.toml
 create mode 100644 providers/orcarouter/models/orcarouter/fusion.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-instruct.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-thinking.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.5-flash.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.6-flash.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.7-flash.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.7-max.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.7-plus.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.8-27b.toml
 create mode 100644 providers/orcarouter/models/qwen/qwen3.8-max.toml
 create mode 100644 providers/orcarouter/models/tencent/hy3-free.toml
 create mode 100644 providers/orcarouter/models/tencent/hy3.toml
 create mode 100644 providers/orcarouter/models/z-ai/glm-5.2.toml
 create mode 100644 providers/orcarouter/models/z-ai/glm-5.3.toml

diff --git a/providers/orcarouter/models/anthropic/claude-fable-5.toml b/providers/orcarouter/models/anthropic/claude-fable-5.toml
new file mode 100644
index 00000000000..8ca46d228ca
--- /dev/null
+++ b/providers/orcarouter/models/anthropic/claude-fable-5.toml
@@ -0,0 +1,8 @@
+base_model = "anthropic/claude-fable-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
diff --git a/providers/orcarouter/models/anthropic/claude-haiku-4.5.toml b/providers/orcarouter/models/anthropic/claude-haiku-4.5.toml
index 57d8d4b851c..88980951c82 100644
--- a/providers/orcarouter/models/anthropic/claude-haiku-4.5.toml
+++ b/providers/orcarouter/models/anthropic/claude-haiku-4.5.toml
@@ -1,5 +1,5 @@
 base_model = "anthropic/claude-haiku-4-5"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 1
diff --git a/providers/orcarouter/models/anthropic/claude-opus-4.1.toml b/providers/orcarouter/models/anthropic/claude-opus-4.1.toml
deleted file mode 100644
index d3b28dfa9d1..00000000000
--- a/providers/orcarouter/models/anthropic/claude-opus-4.1.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "anthropic/claude-opus-4-1"
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-cache_write = 18.75
diff --git a/providers/orcarouter/models/anthropic/claude-opus-4.5.toml b/providers/orcarouter/models/anthropic/claude-opus-4.5.toml
index cf12800a02f..cd3b01382ff 100644
--- a/providers/orcarouter/models/anthropic/claude-opus-4.5.toml
+++ b/providers/orcarouter/models/anthropic/claude-opus-4.5.toml
@@ -1,5 +1,5 @@
 base_model = "anthropic/claude-opus-4-5"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 5
diff --git a/providers/orcarouter/models/anthropic/claude-opus-4.8.toml b/providers/orcarouter/models/anthropic/claude-opus-4.8.toml
new file mode 100644
index 00000000000..2ce24092ea7
--- /dev/null
+++ b/providers/orcarouter/models/anthropic/claude-opus-4.8.toml
@@ -0,0 +1,8 @@
+base_model = "anthropic/claude-opus-4-8"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/orcarouter/models/anthropic/claude-opus-4.toml b/providers/orcarouter/models/anthropic/claude-opus-4.toml
deleted file mode 100644
index e030d0de629..00000000000
--- a/providers/orcarouter/models/anthropic/claude-opus-4.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-cache_write = 18.75
diff --git a/providers/orcarouter/models/anthropic/claude-opus-5.toml b/providers/orcarouter/models/anthropic/claude-opus-5.toml
new file mode 100644
index 00000000000..d71233ef011
--- /dev/null
+++ b/providers/orcarouter/models/anthropic/claude-opus-5.toml
@@ -0,0 +1,10 @@
+# Host bills cache write at 2x input, not the usual 1.25x.
+# Model page states: Cache read / 1M $0.500, Cache write / 1M $10.00.
+base_model = "anthropic/claude-opus-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 10
diff --git a/providers/orcarouter/models/anthropic/claude-sonnet-4.5.toml b/providers/orcarouter/models/anthropic/claude-sonnet-4.5.toml
index b3c98aed8bc..5e0d41e9751 100644
--- a/providers/orcarouter/models/anthropic/claude-sonnet-4.5.toml
+++ b/providers/orcarouter/models/anthropic/claude-sonnet-4.5.toml
@@ -1,8 +1,11 @@
 base_model = "anthropic/claude-sonnet-4-5"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 3
 output = 15
 cache_read = 0.3
 cache_write = 3.75
+
+[limit]
+context = 1_000_000
diff --git a/providers/orcarouter/models/anthropic/claude-sonnet-4.toml b/providers/orcarouter/models/anthropic/claude-sonnet-4.toml
deleted file mode 100644
index 6bd3ab6d6fe..00000000000
--- a/providers/orcarouter/models/anthropic/claude-sonnet-4.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
diff --git a/providers/orcarouter/models/anthropic/claude-sonnet-5.toml b/providers/orcarouter/models/anthropic/claude-sonnet-5.toml
new file mode 100644
index 00000000000..aaf6e7badd0
--- /dev/null
+++ b/providers/orcarouter/models/anthropic/claude-sonnet-5.toml
@@ -0,0 +1,8 @@
+base_model = "anthropic/claude-sonnet-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/orcarouter/models/deepseek/deepseek-chat.toml b/providers/orcarouter/models/deepseek/deepseek-chat.toml
index 695d36ceb11..ed5e8a032ad 100644
--- a/providers/orcarouter/models/deepseek/deepseek-chat.toml
+++ b/providers/orcarouter/models/deepseek/deepseek-chat.toml
@@ -1,6 +1,6 @@
 base_model = "deepseek/deepseek-chat"
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.147
+output = 0.295
+cache_read = 0.02
diff --git a/providers/orcarouter/models/deepseek/deepseek-reasoner.toml b/providers/orcarouter/models/deepseek/deepseek-reasoner.toml
index 9d828bee1fb..51835b382ba 100644
--- a/providers/orcarouter/models/deepseek/deepseek-reasoner.toml
+++ b/providers/orcarouter/models/deepseek/deepseek-reasoner.toml
@@ -1,3 +1,5 @@
+# Host rate per /v1/models: 0.147/0.295, matching the V4 Flash line.
+# cache_ratio 0.19047619 gives 0.028, not the 0.02 those models use.
 base_model = "deepseek/deepseek-reasoner"
 reasoning_options = []
 
@@ -5,6 +7,6 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.435
-output = 0.87
+input = 0.147
+output = 0.295
 cache_read = 0.028
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/orcarouter/models/deepseek/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..c020036fca0
--- /dev/null
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking true|false
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.147
+output = 0.295
+cache_read = 0.02
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-flash-free.toml b/providers/orcarouter/models/deepseek/deepseek-v4-flash-free.toml
new file mode 100644
index 00000000000..37dcc13d6d3
--- /dev/null
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-flash-free.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking true|false
+base_model = "deepseek/deepseek-v4-flash"
+name = "DeepSeek V4 Flash (free)"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/orcarouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..2a00b7d5b35
--- /dev/null
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking true|false
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.147
+output = 0.295
+cache_read = 0.02
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-flash.toml b/providers/orcarouter/models/deepseek/deepseek-v4-flash.toml
index 6f7b5474ebc..74dfbf6a010 100644
--- a/providers/orcarouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-flash.toml
@@ -1,10 +1,14 @@
+# Toggle: thinking true|false
 base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.19
-output = 0.37
-cache_read = 0.0028
+input = 0.147
+output = 0.295
+cache_read = 0.02
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/orcarouter/models/deepseek/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..41b6665954e
--- /dev/null
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking true|false
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.442
+output = 0.884
+cache_read = 0.06
diff --git a/providers/orcarouter/models/deepseek/deepseek-v4-pro.toml b/providers/orcarouter/models/deepseek/deepseek-v4-pro.toml
index 050cbe08668..080e7d9a9df 100644
--- a/providers/orcarouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/orcarouter/models/deepseek/deepseek-v4-pro.toml
@@ -1,10 +1,14 @@
+# Toggle: thinking true|false
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["high", "max"] },
+]
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.56
-output = 1.12
-cache_read = 0.003625
+input = 0.442
+output = 0.884
+cache_read = 0.06
diff --git a/providers/orcarouter/models/google/gemini-2.5-pro.toml b/providers/orcarouter/models/google/gemini-2.5-pro.toml
index 5b10310e718..e4196aa2919 100644
--- a/providers/orcarouter/models/google/gemini-2.5-pro.toml
+++ b/providers/orcarouter/models/google/gemini-2.5-pro.toml
@@ -2,8 +2,8 @@ base_model = "google/gemini-2.5-pro"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 2.5
-output = 15
+input = 1.25
+output = 10
 cache_read = 0.125
 
 [[cost.tiers]]
diff --git a/providers/orcarouter/models/google/gemini-3-flash-preview.toml b/providers/orcarouter/models/google/gemini-3-flash-preview.toml
index a2fac1cec45..56a17f39209 100644
--- a/providers/orcarouter/models/google/gemini-3-flash-preview.toml
+++ b/providers/orcarouter/models/google/gemini-3-flash-preview.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3-flash-preview"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
 
 [cost]
 input = 0.5
diff --git a/providers/orcarouter/models/google/gemini-3-pro-preview.toml b/providers/orcarouter/models/google/gemini-3-pro-preview.toml
deleted file mode 100644
index e8612555a17..00000000000
--- a/providers/orcarouter/models/google/gemini-3-pro-preview.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "google/gemini-3-pro-preview"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-
-[cost]
-input = 4
-output = 18
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 18
-cache_read = 0.4
diff --git a/providers/orcarouter/models/google/gemini-3.1-flash-lite-preview.toml b/providers/orcarouter/models/google/gemini-3.1-flash-lite-preview.toml
index 2e8165f38ec..e862d7dcb19 100644
--- a/providers/orcarouter/models/google/gemini-3.1-flash-lite-preview.toml
+++ b/providers/orcarouter/models/google/gemini-3.1-flash-lite-preview.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.1-flash-lite-preview"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
 
 [cost]
 input = 0.25
diff --git a/providers/orcarouter/models/google/gemini-3.1-flash-lite.toml b/providers/orcarouter/models/google/gemini-3.1-flash-lite.toml
new file mode 100644
index 00000000000..d9d9b5d7ddc
--- /dev/null
+++ b/providers/orcarouter/models/google/gemini-3.1-flash-lite.toml
@@ -0,0 +1,8 @@
+base_model = "google/gemini-3.1-flash-lite"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+input_audio = 0.5
diff --git a/providers/orcarouter/models/google/gemini-3.1-pro-preview-customtools.toml b/providers/orcarouter/models/google/gemini-3.1-pro-preview-customtools.toml
index 1acf47cc5c5..6669e542fc9 100644
--- a/providers/orcarouter/models/google/gemini-3.1-pro-preview-customtools.toml
+++ b/providers/orcarouter/models/google/gemini-3.1-pro-preview-customtools.toml
@@ -2,8 +2,8 @@ base_model = "google/gemini-3.1-pro-preview-customtools"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 4
-output = 18
+input = 2
+output = 12
 cache_read = 0.2
 
 [[cost.tiers]]
diff --git a/providers/orcarouter/models/google/gemini-3.1-pro-preview.toml b/providers/orcarouter/models/google/gemini-3.1-pro-preview.toml
index d7b7c58587c..4f19ecfb352 100644
--- a/providers/orcarouter/models/google/gemini-3.1-pro-preview.toml
+++ b/providers/orcarouter/models/google/gemini-3.1-pro-preview.toml
@@ -2,9 +2,10 @@ base_model = "google/gemini-3.1-pro-preview"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 4
-output = 18
+input = 2
+output = 12
 cache_read = 0.2
+input_audio = 2
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
diff --git a/providers/orcarouter/models/google/gemini-3.5-flash-lite.toml b/providers/orcarouter/models/google/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..f1f4096223f
--- /dev/null
+++ b/providers/orcarouter/models/google/gemini-3.5-flash-lite.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemini-3.5-flash-lite"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.3
+output = 2.5
+cache_read = 0.03
diff --git a/providers/orcarouter/models/google/gemini-3.5-flash.toml b/providers/orcarouter/models/google/gemini-3.5-flash.toml
new file mode 100644
index 00000000000..bc7cc0a7294
--- /dev/null
+++ b/providers/orcarouter/models/google/gemini-3.5-flash.toml
@@ -0,0 +1,9 @@
+base_model = "google/gemini-3.5-flash"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 1.5
+output = 9
+cache_read = 0.15
+cache_write = 0.08333
+input_audio = 3
diff --git a/providers/orcarouter/models/google/gemini-3.6-flash.toml b/providers/orcarouter/models/google/gemini-3.6-flash.toml
new file mode 100644
index 00000000000..09bd1d9a756
--- /dev/null
+++ b/providers/orcarouter/models/google/gemini-3.6-flash.toml
@@ -0,0 +1,9 @@
+# Host bills this at 2x Google's own rate (0.75/3.75); confirmed by
+# /api/pricing and prompt_per_million in /v1/models.
+base_model = "google/gemini-3.6-flash"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 1.5
+output = 7.5
+cache_read = 0.15
diff --git a/providers/orcarouter/models/google/gemini-flash-latest.toml b/providers/orcarouter/models/google/gemini-flash-latest.toml
index 9b124656bb4..402f49ad91e 100644
--- a/providers/orcarouter/models/google/gemini-flash-latest.toml
+++ b/providers/orcarouter/models/google/gemini-flash-latest.toml
@@ -1,9 +1,9 @@
-# Alias for the current Gemini Flash release (gemini-3.5-flash).
+# Host prices this alias below Google's own flash-latest rate.
+# /api/pricing publishes no audio_ratio for it, so no input_audio.
 base_model = "google/gemini-flash-latest"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 1.5
-output = 9
-cache_read = 0.15
-input_audio = 1.5
+input = 0.5
+output = 3
+cache_read = 0.1
diff --git a/providers/orcarouter/models/google/gemini-robotics-er-1.6-preview.toml b/providers/orcarouter/models/google/gemini-robotics-er-1.6-preview.toml
new file mode 100644
index 00000000000..8dcd261ce75
--- /dev/null
+++ b/providers/orcarouter/models/google/gemini-robotics-er-1.6-preview.toml
@@ -0,0 +1,10 @@
+# Toggle/budget: generationConfig.thinkingConfig
+base_model = "google/gemini-robotics-er-1.6-preview"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 0 },
+]
+
+[cost]
+input = 1
+output = 5
diff --git a/providers/orcarouter/models/google/gemma-4-26b-a4b-it.toml b/providers/orcarouter/models/google/gemma-4-26b-a4b-it.toml
index 0b91d5e1ef8..4d987768ba0 100644
--- a/providers/orcarouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/orcarouter/models/google/gemma-4-26b-a4b-it.toml
@@ -4,3 +4,4 @@ reasoning_options = []
 [cost]
 input = 0.06
 output = 0.33
+cache_read = 0.0075
diff --git a/providers/orcarouter/models/google/gemma-4-31b-it.toml b/providers/orcarouter/models/google/gemma-4-31b-it.toml
index 46fde24a2b9..f7ba4aaeba5 100644
--- a/providers/orcarouter/models/google/gemma-4-31b-it.toml
+++ b/providers/orcarouter/models/google/gemma-4-31b-it.toml
@@ -4,3 +4,4 @@ reasoning_options = []
 [cost]
 input = 0.13
 output = 0.38
+cache_read = 0.02
diff --git a/providers/orcarouter/models/grok/grok-4.3.toml b/providers/orcarouter/models/grok/grok-4.3.toml
index faeaba94b59..5a02d4a8ea0 100644
--- a/providers/orcarouter/models/grok/grok-4.3.toml
+++ b/providers/orcarouter/models/grok/grok-4.3.toml
@@ -1,5 +1,5 @@
 base_model = "xai/grok-4.3"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 
 [cost]
 input = 1.25
diff --git a/providers/orcarouter/models/grok/grok-4.5.toml b/providers/orcarouter/models/grok/grok-4.5.toml
new file mode 100644
index 00000000000..c755185fba7
--- /dev/null
+++ b/providers/orcarouter/models/grok/grok-4.5.toml
@@ -0,0 +1,7 @@
+base_model = "xai/grok-4.5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
diff --git a/providers/orcarouter/models/grok/grok-4.6.toml b/providers/orcarouter/models/grok/grok-4.6.toml
new file mode 100644
index 00000000000..5264268a96f
--- /dev/null
+++ b/providers/orcarouter/models/grok/grok-4.6.toml
@@ -0,0 +1,7 @@
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
diff --git a/providers/orcarouter/models/kimi/kimi-k2.5.toml b/providers/orcarouter/models/kimi/kimi-k2.5.toml
index 19871ed4cc3..d33354c9039 100644
--- a/providers/orcarouter/models/kimi/kimi-k2.5.toml
+++ b/providers/orcarouter/models/kimi/kimi-k2.5.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "moonshotai/kimi-k2.5"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
@@ -8,3 +9,7 @@ field = "reasoning_content"
 input = 0.6
 output = 3
 cache_read = 0.1
+cache_write = 0
+
+[limit]
+output = 32_768
diff --git a/providers/orcarouter/models/kimi/kimi-k2.6.toml b/providers/orcarouter/models/kimi/kimi-k2.6.toml
index 7e038adc8d2..8e343d088cb 100644
--- a/providers/orcarouter/models/kimi/kimi-k2.6.toml
+++ b/providers/orcarouter/models/kimi/kimi-k2.6.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "moonshotai/kimi-k2.6"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
@@ -8,3 +9,6 @@ field = "reasoning_content"
 input = 0.95
 output = 4
 cache_read = 0.16
+
+[limit]
+output = 32_768
diff --git a/providers/orcarouter/models/kimi/kimi-k2.7-code.toml b/providers/orcarouter/models/kimi/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..59c001905bc
--- /dev/null
+++ b/providers/orcarouter/models/kimi/kimi-k2.7-code.toml
@@ -0,0 +1,10 @@
+base_model = "moonshotai/kimi-k2.7-code"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/orcarouter/models/kimi/kimi-k3.toml b/providers/orcarouter/models/kimi/kimi-k3.toml
new file mode 100644
index 00000000000..7f76968dcfd
--- /dev/null
+++ b/providers/orcarouter/models/kimi/kimi-k3.toml
@@ -0,0 +1,14 @@
+# Toggle: thinking true|false
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 3.3
+output = 16.5
+cache_read = 0.33
diff --git a/providers/orcarouter/models/meta/muse-spark-1.1.toml b/providers/orcarouter/models/meta/muse-spark-1.1.toml
new file mode 100644
index 00000000000..e810b05129d
--- /dev/null
+++ b/providers/orcarouter/models/meta/muse-spark-1.1.toml
@@ -0,0 +1,7 @@
+base_model = "meta/muse-spark-1.1"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
diff --git a/providers/orcarouter/models/meta/muse-spark-1.2.toml b/providers/orcarouter/models/meta/muse-spark-1.2.toml
new file mode 100644
index 00000000000..d53a66ded75
--- /dev/null
+++ b/providers/orcarouter/models/meta/muse-spark-1.2.toml
@@ -0,0 +1,7 @@
+base_model = "meta/muse-spark-1.2"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
diff --git a/providers/orcarouter/models/minimax/minimax-m2.5-highspeed.toml b/providers/orcarouter/models/minimax/minimax-m2.5-highspeed.toml
index d3be98c0fab..64df7841b29 100644
--- a/providers/orcarouter/models/minimax/minimax-m2.5-highspeed.toml
+++ b/providers/orcarouter/models/minimax/minimax-m2.5-highspeed.toml
@@ -4,5 +4,5 @@ reasoning_options = []
 [cost]
 input = 0.6
 output = 2.4
-cache_read = 0.06
+cache_read = 0.03
 cache_write = 0.375
diff --git a/providers/orcarouter/models/minimax/minimax-m3.toml b/providers/orcarouter/models/minimax/minimax-m3.toml
new file mode 100644
index 00000000000..5475d5a734b
--- /dev/null
+++ b/providers/orcarouter/models/minimax/minimax-m3.toml
@@ -0,0 +1,8 @@
+# Toggle: thinking true|false
+base_model = "minimax/MiniMax-M3"
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
diff --git a/providers/orcarouter/models/openai/gpt-5-chat-latest.toml b/providers/orcarouter/models/openai/gpt-5-chat-latest.toml
index 4dd858d1af8..7359fa1d4a0 100644
--- a/providers/orcarouter/models/openai/gpt-5-chat-latest.toml
+++ b/providers/orcarouter/models/openai/gpt-5-chat-latest.toml
@@ -5,3 +5,6 @@ reasoning_options = []
 input = 1.25
 output = 10
 cache_read = 0.125
+
+[limit]
+output = 100_000
diff --git a/providers/orcarouter/models/openai/gpt-5-codex.toml b/providers/orcarouter/models/openai/gpt-5-codex.toml
deleted file mode 100644
index 614bb89b104..00000000000
--- a/providers/orcarouter/models/openai/gpt-5-codex.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "openai/gpt-5-codex"
-reasoning_options = []
-
-[cost]
-input = 1.25
-output = 10
-cache_read = 0.125
diff --git a/providers/orcarouter/models/openai/gpt-5-mini.toml b/providers/orcarouter/models/openai/gpt-5-mini.toml
index ec2e0406779..5754d47df3f 100644
--- a/providers/orcarouter/models/openai/gpt-5-mini.toml
+++ b/providers/orcarouter/models/openai/gpt-5-mini.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5-mini"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
 
 [cost]
 input = 0.25
diff --git a/providers/orcarouter/models/openai/gpt-5-nano.toml b/providers/orcarouter/models/openai/gpt-5-nano.toml
index 35e4f5d9e48..b96eddfd6f0 100644
--- a/providers/orcarouter/models/openai/gpt-5-nano.toml
+++ b/providers/orcarouter/models/openai/gpt-5-nano.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5-nano"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
 
 [cost]
 input = 0.05
diff --git a/providers/orcarouter/models/openai/gpt-5-pro.toml b/providers/orcarouter/models/openai/gpt-5-pro.toml
index d9e22584955..59f10fdd949 100644
--- a/providers/orcarouter/models/openai/gpt-5-pro.toml
+++ b/providers/orcarouter/models/openai/gpt-5-pro.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5-pro"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["high"] }]
 
 [cost]
 input = 15
diff --git a/providers/orcarouter/models/openai/gpt-5.1-codex-max.toml b/providers/orcarouter/models/openai/gpt-5.1-codex-max.toml
deleted file mode 100644
index a137a421adb..00000000000
--- a/providers/orcarouter/models/openai/gpt-5.1-codex-max.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "openai/gpt-5.1-codex-max"
-reasoning_options = []
-
-[cost]
-input = 1.25
-output = 10
-cache_read = 0.125
diff --git a/providers/orcarouter/models/openai/gpt-5.1-codex-mini.toml b/providers/orcarouter/models/openai/gpt-5.1-codex-mini.toml
index ba981225f1a..3cb2b12eeba 100644
--- a/providers/orcarouter/models/openai/gpt-5.1-codex-mini.toml
+++ b/providers/orcarouter/models/openai/gpt-5.1-codex-mini.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.1-codex-mini"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 0.25
diff --git a/providers/orcarouter/models/openai/gpt-5.1-codex.toml b/providers/orcarouter/models/openai/gpt-5.1-codex.toml
index 6e672919a05..17e3fc0fbc3 100644
--- a/providers/orcarouter/models/openai/gpt-5.1-codex.toml
+++ b/providers/orcarouter/models/openai/gpt-5.1-codex.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.1-codex"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 1.25
diff --git a/providers/orcarouter/models/openai/gpt-5.1.toml b/providers/orcarouter/models/openai/gpt-5.1.toml
index 7ab9851a4ab..489f83fffcb 100644
--- a/providers/orcarouter/models/openai/gpt-5.1.toml
+++ b/providers/orcarouter/models/openai/gpt-5.1.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.1"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 
 [cost]
 input = 1.25
diff --git a/providers/orcarouter/models/openai/gpt-5.2-chat-latest.toml b/providers/orcarouter/models/openai/gpt-5.2-chat-latest.toml
index bbad0e8faf5..e8664270145 100644
--- a/providers/orcarouter/models/openai/gpt-5.2-chat-latest.toml
+++ b/providers/orcarouter/models/openai/gpt-5.2-chat-latest.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.2-chat-latest"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["medium"] }]
 
 [cost]
 input = 1.75
diff --git a/providers/orcarouter/models/openai/gpt-5.2-codex.toml b/providers/orcarouter/models/openai/gpt-5.2-codex.toml
index a03382c9d90..c084afebb9b 100644
--- a/providers/orcarouter/models/openai/gpt-5.2-codex.toml
+++ b/providers/orcarouter/models/openai/gpt-5.2-codex.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.2-codex"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 1.75
diff --git a/providers/orcarouter/models/openai/gpt-5.2-pro.toml b/providers/orcarouter/models/openai/gpt-5.2-pro.toml
index 860e9e002de..86169206ac8 100644
--- a/providers/orcarouter/models/openai/gpt-5.2-pro.toml
+++ b/providers/orcarouter/models/openai/gpt-5.2-pro.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.2-pro"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
 
 [cost]
 input = 21
diff --git a/providers/orcarouter/models/openai/gpt-5.2.toml b/providers/orcarouter/models/openai/gpt-5.2.toml
index 4f38ef3d4dc..a2a20d0e325 100644
--- a/providers/orcarouter/models/openai/gpt-5.2.toml
+++ b/providers/orcarouter/models/openai/gpt-5.2.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.2"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 1.75
diff --git a/providers/orcarouter/models/openai/gpt-5.3-chat-latest.toml b/providers/orcarouter/models/openai/gpt-5.3-chat-latest.toml
deleted file mode 100644
index 197284ef5f0..00000000000
--- a/providers/orcarouter/models/openai/gpt-5.3-chat-latest.toml
+++ /dev/null
@@ -1,6 +0,0 @@
-base_model = "openai/gpt-5.3-chat-latest"
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175
diff --git a/providers/orcarouter/models/openai/gpt-5.3-codex.toml b/providers/orcarouter/models/openai/gpt-5.3-codex.toml
index 1d3369a77e8..949c3ed11a7 100644
--- a/providers/orcarouter/models/openai/gpt-5.3-codex.toml
+++ b/providers/orcarouter/models/openai/gpt-5.3-codex.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.3-codex"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 1.75
diff --git a/providers/orcarouter/models/openai/gpt-5.4-mini.toml b/providers/orcarouter/models/openai/gpt-5.4-mini.toml
index 030517c703c..56a278d15d0 100644
--- a/providers/orcarouter/models/openai/gpt-5.4-mini.toml
+++ b/providers/orcarouter/models/openai/gpt-5.4-mini.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.4-mini"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 0.75
diff --git a/providers/orcarouter/models/openai/gpt-5.4-nano.toml b/providers/orcarouter/models/openai/gpt-5.4-nano.toml
index a4366b00e40..3872b109a25 100644
--- a/providers/orcarouter/models/openai/gpt-5.4-nano.toml
+++ b/providers/orcarouter/models/openai/gpt-5.4-nano.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.4-nano"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 0.2
diff --git a/providers/orcarouter/models/openai/gpt-5.4-pro.toml b/providers/orcarouter/models/openai/gpt-5.4-pro.toml
index 38eb4124bf8..11330748034 100644
--- a/providers/orcarouter/models/openai/gpt-5.4-pro.toml
+++ b/providers/orcarouter/models/openai/gpt-5.4-pro.toml
@@ -1,9 +1,9 @@
 base_model = "openai/gpt-5.4-pro"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
 
 [cost]
-input = 60
-output = 270
+input = 30
+output = 180
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
diff --git a/providers/orcarouter/models/openai/gpt-5.4.toml b/providers/orcarouter/models/openai/gpt-5.4.toml
index 344da90f2e0..76ccd302ea0 100644
--- a/providers/orcarouter/models/openai/gpt-5.4.toml
+++ b/providers/orcarouter/models/openai/gpt-5.4.toml
@@ -1,9 +1,9 @@
 base_model = "openai/gpt-5.4"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
-input = 5
-output = 22.5
+input = 2.5
+output = 15
 cache_read = 0.25
 
 [[cost.tiers]]
diff --git a/providers/orcarouter/models/openai/gpt-5.5-pro.toml b/providers/orcarouter/models/openai/gpt-5.5-pro.toml
index e5630efdaec..616e9a17ee1 100644
--- a/providers/orcarouter/models/openai/gpt-5.5-pro.toml
+++ b/providers/orcarouter/models/openai/gpt-5.5-pro.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.5-pro"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
 
 [cost]
 input = 30
@@ -9,3 +9,6 @@ output = 180
 tier = { type = "context", size = 272_000 }
 input = 60
 output = 270
+
+[limit]
+output = 100_000
diff --git a/providers/orcarouter/models/openai/gpt-5.5.toml b/providers/orcarouter/models/openai/gpt-5.5.toml
index 6499ab01a51..9b565dd3a2e 100644
--- a/providers/orcarouter/models/openai/gpt-5.5.toml
+++ b/providers/orcarouter/models/openai/gpt-5.5.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5.5"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 5
diff --git a/providers/orcarouter/models/openai/gpt-5.6-luna.toml b/providers/orcarouter/models/openai/gpt-5.6-luna.toml
new file mode 100644
index 00000000000..a9090b0f1cd
--- /dev/null
+++ b/providers/orcarouter/models/openai/gpt-5.6-luna.toml
@@ -0,0 +1,8 @@
+base_model = "openai/gpt-5.6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/orcarouter/models/openai/gpt-5.6-sol.toml b/providers/orcarouter/models/openai/gpt-5.6-sol.toml
new file mode 100644
index 00000000000..770c9799aad
--- /dev/null
+++ b/providers/orcarouter/models/openai/gpt-5.6-sol.toml
@@ -0,0 +1,8 @@
+base_model = "openai/gpt-5.6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
diff --git a/providers/orcarouter/models/openai/gpt-5.6-terra.toml b/providers/orcarouter/models/openai/gpt-5.6-terra.toml
new file mode 100644
index 00000000000..0b76158fb01
--- /dev/null
+++ b/providers/orcarouter/models/openai/gpt-5.6-terra.toml
@@ -0,0 +1,8 @@
+base_model = "openai/gpt-5.6-terra"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/orcarouter/models/openai/gpt-5.toml b/providers/orcarouter/models/openai/gpt-5.toml
index 4c848769ca7..104575502bb 100644
--- a/providers/orcarouter/models/openai/gpt-5.toml
+++ b/providers/orcarouter/models/openai/gpt-5.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-5"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
 
 [cost]
 input = 1.25
diff --git a/providers/orcarouter/models/openai/gpt-oss-120b.toml b/providers/orcarouter/models/openai/gpt-oss-120b.toml
new file mode 100644
index 00000000000..ac1ceee7702
--- /dev/null
+++ b/providers/orcarouter/models/openai/gpt-oss-120b.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-oss-120b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.03
+output = 0.17
diff --git a/providers/orcarouter/models/orcarouter/free.toml b/providers/orcarouter/models/orcarouter/free.toml
new file mode 100644
index 00000000000..19924b3e0ef
--- /dev/null
+++ b/providers/orcarouter/models/orcarouter/free.toml
@@ -0,0 +1,24 @@
+name = "OrcaRouter Free"
+description = "Built-in router over the free tier that scores each request's difficulty and sends light work to the smaller free model and harder work to the stronger one. Priced at zero and never falls back to a paid model."
+family = "auto"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+reasoning_options = []
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 65_536
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/orcarouter/models/orcarouter/fusion-flash.toml b/providers/orcarouter/models/orcarouter/fusion-flash.toml
new file mode 100644
index 00000000000..4491dcdb44c
--- /dev/null
+++ b/providers/orcarouter/models/orcarouter/fusion-flash.toml
@@ -0,0 +1,22 @@
+# No [cost]: billed as the sum of the panel legs actually run, at
+# upstream cost with zero markup, so there is no per-token price.
+name = "OrcaRouter Fusion Flash"
+description = "Budget Fusion panel that runs Gemini 3.5 Flash, MiniMax M2.7 and GLM 5.1 in parallel on hard requests, then has a Claude Opus 4.8 judge return the strongest single answer verbatim. Cost-sensitive fan-out over a 200K window."
+family = "model-router"
+release_date = "2026-06-15"
+last_updated = "2026-06-15"
+attachment = false
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/orcarouter/models/orcarouter/fusion-mini.toml b/providers/orcarouter/models/orcarouter/fusion-mini.toml
new file mode 100644
index 00000000000..6e12edd0899
--- /dev/null
+++ b/providers/orcarouter/models/orcarouter/fusion-mini.toml
@@ -0,0 +1,22 @@
+# No [cost]: billed as the sum of the panel legs actually run, at
+# upstream cost with zero markup, so there is no per-token price.
+name = "OrcaRouter Fusion Mini"
+description = "Leaner two-model Fusion panel that runs Claude Opus 4.8 and GPT-5.5 in parallel on hard requests, then has a Claude Opus 4.8 judge return the strongest single answer verbatim. Easy requests fall through to a cheaper default and bill as one call."
+family = "model-router"
+release_date = "2026-06-15"
+last_updated = "2026-06-15"
+attachment = true
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/orcarouter/models/orcarouter/fusion.toml b/providers/orcarouter/models/orcarouter/fusion.toml
new file mode 100644
index 00000000000..b9dd4814a43
--- /dev/null
+++ b/providers/orcarouter/models/orcarouter/fusion.toml
@@ -0,0 +1,22 @@
+# No [cost]: billed as the sum of the panel legs actually run, at
+# upstream cost with zero markup, so there is no per-token price.
+name = "OrcaRouter Fusion"
+description = "Curated fan-out router that runs Claude Opus 4.8, GPT-5.5 and Gemini 3.1 Pro in parallel on hard requests, then has a Claude Opus 4.8 judge return the strongest single answer verbatim. Easy requests fall through to a cheaper default and bill as one call."
+family = "model-router"
+release_date = "2026-06-15"
+last_updated = "2026-06-15"
+attachment = true
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-instruct.toml b/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-instruct.toml
new file mode 100644
index 00000000000..e3c8680b66e
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-instruct.toml
@@ -0,0 +1,5 @@
+base_model = "alibaba/qwen3-vl-235b-a22b-instruct"
+
+[cost]
+input = 0.4
+output = 1.6
diff --git a/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-thinking.toml b/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-thinking.toml
new file mode 100644
index 00000000000..99c063e1aae
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3-vl-235b-a22b-thinking.toml
@@ -0,0 +1,10 @@
+# Budget: thinking_budget. Dedicated thinking SKU, so no off switch.
+base_model = "alibaba/qwen3-vl-235b-a22b-thinking"
+reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 81_920 }]
+
+[cost]
+input = 0.4
+output = 4
+
+[limit]
+output = 40_960
diff --git a/providers/orcarouter/models/qwen/qwen3.5-122b-a10b.toml b/providers/orcarouter/models/qwen/qwen3.5-122b-a10b.toml
index 1eb4bd3c71f..90e0b506cb0 100644
--- a/providers/orcarouter/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/orcarouter/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.5-122b-a10b"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.115
diff --git a/providers/orcarouter/models/qwen/qwen3.5-27b.toml b/providers/orcarouter/models/qwen/qwen3.5-27b.toml
index c029b6e3bd1..09945904753 100644
--- a/providers/orcarouter/models/qwen/qwen3.5-27b.toml
+++ b/providers/orcarouter/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.5-27b"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.086
diff --git a/providers/orcarouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/orcarouter/models/qwen/qwen3.5-35b-a3b.toml
index b3505ad727b..9a706594451 100644
--- a/providers/orcarouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/orcarouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.5-35b-a3b"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.057
diff --git a/providers/orcarouter/models/qwen/qwen3.5-397b-a17b.toml b/providers/orcarouter/models/qwen/qwen3.5-397b-a17b.toml
index 359409c9f84..1b3f128da2f 100644
--- a/providers/orcarouter/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/orcarouter/models/qwen/qwen3.5-397b-a17b.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.5-397b-a17b"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.172
diff --git a/providers/orcarouter/models/qwen/qwen3.5-flash.toml b/providers/orcarouter/models/qwen/qwen3.5-flash.toml
new file mode 100644
index 00000000000..45e1c566934
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.5-flash.toml
@@ -0,0 +1,10 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.5-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 1, max = 81_920 },
+]
+
+[cost]
+input = 0.1
+output = 0.4
diff --git a/providers/orcarouter/models/qwen/qwen3.5-plus.toml b/providers/orcarouter/models/qwen/qwen3.5-plus.toml
index 723cc50ec34..69aba29ee3f 100644
--- a/providers/orcarouter/models/qwen/qwen3.5-plus.toml
+++ b/providers/orcarouter/models/qwen/qwen3.5-plus.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.5-plus"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.115
diff --git a/providers/orcarouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/orcarouter/models/qwen/qwen3.6-35b-a3b.toml
index 4f30ca3b51d..55a66762c0c 100644
--- a/providers/orcarouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/orcarouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.6-35b-a3b"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.248
diff --git a/providers/orcarouter/models/qwen/qwen3.6-flash.toml b/providers/orcarouter/models/qwen/qwen3.6-flash.toml
new file mode 100644
index 00000000000..542b5a094d4
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.6-flash.toml
@@ -0,0 +1,10 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.6-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
+
+[cost]
+input = 0.25
+output = 1.5
diff --git a/providers/orcarouter/models/qwen/qwen3.6-plus.toml b/providers/orcarouter/models/qwen/qwen3.6-plus.toml
index 308cef96883..d1d76da4a45 100644
--- a/providers/orcarouter/models/qwen/qwen3.6-plus.toml
+++ b/providers/orcarouter/models/qwen/qwen3.6-plus.toml
@@ -1,5 +1,9 @@
+# Toggle: enable_thinking; budget: thinking_budget
 base_model = "alibaba/qwen3.6-plus"
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
 
 [cost]
 input = 0.5
diff --git a/providers/orcarouter/models/qwen/qwen3.7-flash.toml b/providers/orcarouter/models/qwen/qwen3.7-flash.toml
new file mode 100644
index 00000000000..06976af6ab0
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.7-flash.toml
@@ -0,0 +1,12 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.7-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
+
+[cost]
+input = 0.03
+output = 0.13
+cache_read = 0.006
+cache_write = 0.038
diff --git a/providers/orcarouter/models/qwen/qwen3.7-max.toml b/providers/orcarouter/models/qwen/qwen3.7-max.toml
new file mode 100644
index 00000000000..a71ff398d5d
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.7-max.toml
@@ -0,0 +1,12 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.7-max"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
+
+[cost]
+input = 1.25
+output = 3.75
+cache_read = 0.25
+cache_write = 1.563
diff --git a/providers/orcarouter/models/qwen/qwen3.7-plus.toml b/providers/orcarouter/models/qwen/qwen3.7-plus.toml
new file mode 100644
index 00000000000..2156f471fa1
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.7-plus.toml
@@ -0,0 +1,11 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.7-plus"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens" },
+]
+
+[cost]
+input = 0.35
+output = 1.42
+cache_read = 0.071
diff --git a/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml b/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
new file mode 100644
index 00000000000..72bf87a72bb
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
@@ -0,0 +1,10 @@
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8 27B (free)"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 65_536
diff --git a/providers/orcarouter/models/qwen/qwen3.8-27b.toml b/providers/orcarouter/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..9fdacd23754
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,6 @@
+base_model = "alibaba/qwen3.8-27b"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
+
+[cost]
+input = 0.33
+output = 2.4
diff --git a/providers/orcarouter/models/qwen/qwen3.8-max.toml b/providers/orcarouter/models/qwen/qwen3.8-max.toml
new file mode 100644
index 00000000000..f2a983c1bd1
--- /dev/null
+++ b/providers/orcarouter/models/qwen/qwen3.8-max.toml
@@ -0,0 +1,16 @@
+# Toggle: enable_thinking; budget: thinking_budget
+base_model = "alibaba/qwen3.8-max"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", min = 0, max = 262_144 },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+cache_write = 2.5
diff --git a/providers/orcarouter/models/tencent/hy3-free.toml b/providers/orcarouter/models/tencent/hy3-free.toml
new file mode 100644
index 00000000000..5c4f2f54d2d
--- /dev/null
+++ b/providers/orcarouter/models/tencent/hy3-free.toml
@@ -0,0 +1,7 @@
+base_model = "tencent/hy3"
+name = "Hy3 (free)"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high"] }]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/orcarouter/models/tencent/hy3.toml b/providers/orcarouter/models/tencent/hy3.toml
new file mode 100644
index 00000000000..d82ff03bcf2
--- /dev/null
+++ b/providers/orcarouter/models/tencent/hy3.toml
@@ -0,0 +1,7 @@
+base_model = "tencent/hy3"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high"] }]
+
+[cost]
+input = 0.18
+output = 0.59
+cache_read = 0.059
diff --git a/providers/orcarouter/models/z-ai/glm-4.5-air.toml b/providers/orcarouter/models/z-ai/glm-4.5-air.toml
index 4324692a5be..b99b52d9d03 100644
--- a/providers/orcarouter/models/z-ai/glm-4.5-air.toml
+++ b/providers/orcarouter/models/z-ai/glm-4.5-air.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-4.5-air"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.2
diff --git a/providers/orcarouter/models/z-ai/glm-4.5.toml b/providers/orcarouter/models/z-ai/glm-4.5.toml
index ab4cdeb9737..81bf4ed1436 100644
--- a/providers/orcarouter/models/z-ai/glm-4.5.toml
+++ b/providers/orcarouter/models/z-ai/glm-4.5.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-4.5"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.6
diff --git a/providers/orcarouter/models/z-ai/glm-4.6.toml b/providers/orcarouter/models/z-ai/glm-4.6.toml
index 329b6bbd5cf..103d2fc70fd 100644
--- a/providers/orcarouter/models/z-ai/glm-4.6.toml
+++ b/providers/orcarouter/models/z-ai/glm-4.6.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-4.6"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.6
diff --git a/providers/orcarouter/models/z-ai/glm-4.7.toml b/providers/orcarouter/models/z-ai/glm-4.7.toml
index 2e6a1258941..e65ed707921 100644
--- a/providers/orcarouter/models/z-ai/glm-4.7.toml
+++ b/providers/orcarouter/models/z-ai/glm-4.7.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-4.7"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/orcarouter/models/z-ai/glm-5.1.toml b/providers/orcarouter/models/z-ai/glm-5.1.toml
index 691da7cfdf2..da8828bb9ab 100644
--- a/providers/orcarouter/models/z-ai/glm-5.1.toml
+++ b/providers/orcarouter/models/z-ai/glm-5.1.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-5.1"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/orcarouter/models/z-ai/glm-5.2.toml b/providers/orcarouter/models/z-ai/glm-5.2.toml
new file mode 100644
index 00000000000..939962684e8
--- /dev/null
+++ b/providers/orcarouter/models/z-ai/glm-5.2.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/orcarouter/models/z-ai/glm-5.3.toml b/providers/orcarouter/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..7fbbffd3aff
--- /dev/null
+++ b/providers/orcarouter/models/z-ai/glm-5.3.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.26
+output = 3.96
+cache_read = 0.234
diff --git a/providers/orcarouter/models/z-ai/glm-5.toml b/providers/orcarouter/models/z-ai/glm-5.toml
index 046f58af413..67d7aef87d6 100644
--- a/providers/orcarouter/models/z-ai/glm-5.toml
+++ b/providers/orcarouter/models/z-ai/glm-5.toml
@@ -1,5 +1,6 @@
+# Toggle: thinking true|false
 base_model = "zhipuai/glm-5"
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
@@ -7,5 +8,5 @@ field = "reasoning_content"
 [cost]
 input = 1
 output = 3.2
-cache_read = 0.2
+cache_read = 0.26
 cache_write = 0

From cc460f27fb811f06ea6a0bdc61c00fc9b980bd33 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Fri, 28 Aug 2026 08:27:04 +0800
Subject: [PATCH 0092/2012] fix(volcengine): correct the effort levels on Seed
 character (#5645)

Driving every model from the live api.json with exactly what it declares turned
up that five of this entry's six effort values are rejected outright:

  reasoning_effort=low   -> 400 "Invalid combination of reasoning_effort and
                                 thinking type: low + disabled"
  same for medium, high, xhigh, max; only none and minimal are accepted alone

The cause is that this model defaults thinking OFF, so a graded effort has to be
paired with thinking.type=enabled. The existing comment noted the default but the
values list still implied the levels worked on their own.

With thinking enabled they are accepted, but not all are separable. n=4 runs per
level on a prompt hard enough to consume the budget:

  minimal 0 0 0 0                 low  1347/1258/1166/1149
  high    1709/2696/1695/1513     max  2304/1881/1985/1480

max overlaps high rather than exceeding it, and medium/xhigh fall in the same
band, so the effective set is minimal|low|high.

Worth noting an easy prompt hides this completely: on a one-step question every
level lands at 67-79 reasoning tokens.

Verified with the repo's validator locally (bun packages/core/script/validate.ts,
exit 0) and confirmed the generated api.json now lists the three levels.

Change-Id: Ie8ad1ec73fe304c65fe3cc10b95015b7ec0c9937
---
 .../models/doubao-seed-character-260628.toml       | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/providers/volcengine/models/doubao-seed-character-260628.toml b/providers/volcengine/models/doubao-seed-character-260628.toml
index 2bebe7a5463..83be4ae59b9 100644
--- a/providers/volcengine/models/doubao-seed-character-260628.toml
+++ b/providers/volcengine/models/doubao-seed-character-260628.toml
@@ -2,8 +2,16 @@
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
 # Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = minimal|low|medium|high|xhigh|max (no none: this
-# model defaults thinking off, so thinking.type is the real switch here)
+# Effort: reasoning_effort = minimal|low|high, and it must be paired with
+# thinking.type=enabled. This model defaults thinking OFF, so a graded effort sent
+# on its own is rejected: "Invalid combination of reasoning_effort and thinking
+# type: low + disabled". 'none' is likewise unavailable; 'minimal' is the off value.
+# Only three levels are separable. With thinking enabled, n=4 runs per level on a
+# prompt hard enough to consume the budget:
+#   minimal 0 0 0 0                 low  1347/1258/1166/1149
+#   high    1709/2696/1695/1513     max  2304/1881/1985/1480
+# max overlaps high rather than exceeding it, and medium/xhigh sit inside the same
+# band, so they are dropped.
 base_model = "bytedance-seed/seed-character"
 
 [[reasoning_options]]
@@ -11,7 +19,7 @@ type = "toggle"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["minimal", "low", "high"]
 
 [interleaved]
 field = "reasoning_content"

From 991dd4ce306909e08ad27fcb636ea503539f5742 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 19:28:25 -0500
Subject: [PATCH 0093/2012] chore(sync): update Weights & Biases model catalog
 (#5665)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/moonshotai/Kimi-K3.toml  |  4 +++-
 .../wandb/models/zai-org/GLM-5.3-Flash.toml     | 17 +++++++++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)
 create mode 100644 providers/wandb/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/wandb/models/moonshotai/Kimi-K3.toml b/providers/wandb/models/moonshotai/Kimi-K3.toml
index e336060fb28..68fa428979c 100644
--- a/providers/wandb/models/moonshotai/Kimi-K3.toml
+++ b/providers/wandb/models/moonshotai/Kimi-K3.toml
@@ -1,7 +1,9 @@
 base_model = "moonshotai/kimi-k3"
 description = "Kimi K3 is a 2.8T-parameter multimodal MoE model with 104B active parameters built for long-horizon coding and agentic workflows."
 temperature = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 3
diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..0edd2acb8c5
--- /dev/null
+++ b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]

From 7e490708c8bb287023e2428c399dff203dd67b16 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Fri, 28 Aug 2026 08:30:28 +0800
Subject: [PATCH 0094/2012] fix(volcengine): correct the output limit on Seed
 2.0 lite and mini (#5644)

Both entries inherit output = 32_000 from models/bytedance-seed, but this host
accepts four times that. Probed the boundary on /api/v3/chat/completions:

  lite  max_tokens=131072 -> 200,  132000 -> 400 "max_tokens ... not valid"
  mini  max_tokens=131072 -> 200,  132000 -> 400

131_072 is also what openrouter, kilo and nano-gpt already declare for the same
base model, so the lab default is simply stale rather than host-specific. Added
a per-host [limit] override, which is the pattern those three use.

Left context at the inherited 256_000: a 180k-token prompt was accepted, which
does not discriminate between 256_000 and the 262_144 the other hosts claim, so
there is no measurement to justify changing it.

Checked the neighbours rather than assuming the family shares a ceiling:
doubao-seed-1-6-flash-250828 and -vision-250815 reject max_tokens=131072 and
genuinely cap at 32_000, so they are left alone.

Verified with the repo's validator locally (bun packages/core/script/validate.ts,
exit 0) and confirmed the generated limit reads 131072 for these two and stays
32000 for the 1.6 entries.

Change-Id: I87cca991db79bd2aa586d75d44c1e26c4580fac9
---
 .../volcengine/models/doubao-seed-2-0-lite-260428.toml     | 7 +++++++
 .../volcengine/models/doubao-seed-2-0-mini-260428.toml     | 7 +++++++
 2 files changed, 14 insertions(+)

diff --git a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
index effe10d6c07..f75f626cf34 100644
--- a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
@@ -4,8 +4,15 @@
 # Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
 # none/minimal disable thinking, so no separate toggle: sending effort alone
 # already enables it, and thinking.type=disabled + a graded effort is rejected.
+# Output limit: the lab default is 32_000, but this host accepts up to 131_072
+# (verified 2026-08-27: max_tokens=131072 -> 200, 132000 -> 400 invalid), matching
+# what openrouter/kilo/nano-gpt already declare for this model. Overridden below.
 base_model = "bytedance-seed/seed-2.0-lite"
 
+[limit]
+context = 256_000
+output = 131_072
+
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
diff --git a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
index a2774f2ddd9..dafad8fa4bf 100644
--- a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
@@ -4,8 +4,15 @@
 # Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
 # none/minimal disable thinking, so no separate toggle: sending effort alone
 # already enables it, and thinking.type=disabled + a graded effort is rejected.
+# Output limit: the lab default is 32_000, but this host accepts up to 131_072
+# (verified 2026-08-27: max_tokens=131072 -> 200, 132000 -> 400 invalid), matching
+# what openrouter/kilo/nano-gpt already declare for this model. Overridden below.
 base_model = "bytedance-seed/seed-2.0-mini"
 
+[limit]
+context = 256_000
+output = 131_072
+
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]

From 242ba90e29fbd4bd753661b53744bf835f476fd2 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Fri, 28 Aug 2026 08:31:26 +0800
Subject: [PATCH 0095/2012] fix(volcengine): declare reasoning on Seed 1.6
 flash and vision (#5639)

* fix(volcengine): declare reasoning on Seed 1.6 flash and vision

Both models are served as reasoning=false on models.dev today, but they reason
by default: a bare request with no reasoning parameters returns 283 and 166
reasoning tokens respectively, with reasoning_content populated. The entries were
missing a reasoning_options block, and the bytedance-seed lab entries do not
declare one either, so nothing supplied it.

Declared as a toggle, which is what actually works here:
  flash   thinking.type=enabled -> 278 reasoning tokens, disabled -> 0
  vision  thinking.type=enabled -> 162 reasoning tokens, disabled -> 0

No graded effort on either. reasoning_effort is accepted but not separable, and
'none' does not switch thinking off, measured n=4 per level:
  flash   none 762/479/627/646  minimal 715/513/760/717  low 942/561/660/596
          high 599/502/586/837  max 575/863/714/598
  vision  none 895/927/670/736  low 550/901/775/598
          high 578/738/787/1031 max 852/813/1011/699

Found while auditing the live api.json against the real API rather than from a
report, so downstream consumers have been reading the wrong flag since #5513.

Change-Id: I6e2129eded9b5d26768e29afb9c0a5aa4de10999

* fix(volcengine): set reasoning = true explicitly on the two Seed 1.6 entries

CI caught that reasoning_options alone isn't enough: the validator rejects
"Cannot set reasoning_options when reasoning is false". The boolean is not
derived from the options block, and neither the volcengine entries nor the
bytedance-seed lab entries they inherit from declare it, so it defaulted to
false. Set it explicitly, the way peers such as providers/302ai do.

Verified with the repo's own validator locally (bun packages/core/script/
validate.ts, exit 0) rather than relying on CI to check it.

Change-Id: Ia6ebc6cf2e1c5a78d2454c23247f26dbb7bbdcbd
---
 .../models/doubao-seed-1-6-flash-250828.toml      | 15 +++++++++++++++
 .../models/doubao-seed-1-6-vision-250815.toml     | 15 +++++++++++++++
 2 files changed, 30 insertions(+)

diff --git a/providers/volcengine/models/doubao-seed-1-6-flash-250828.toml b/providers/volcengine/models/doubao-seed-1-6-flash-250828.toml
index 045c026b4dd..1f612a39a73 100644
--- a/providers/volcengine/models/doubao-seed-1-6-flash-250828.toml
+++ b/providers/volcengine/models/doubao-seed-1-6-flash-250828.toml
@@ -1,7 +1,22 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-1-6-flash-250828`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
+# Toggle: thinking.type = enabled|disabled, POST /api/v3/chat/completions
+# (verified 2026-08-27: enabled -> 278 reasoning tokens, disabled -> 0). This model
+# reasons by default -- a bare request with no reasoning params returns 283 reasoning
+# tokens -- so the absence of a reasoning declaration here was wrong.
+# No graded effort: reasoning_effort is accepted but not separable. Over n=4 runs per
+# level, none 762/479/627/646, minimal 715/513/760/717, low 942/561/660/596,
+# high 599/502/586/837, max 575/863/714/598 -- fully overlapping, and 'none' does not
+# even switch thinking off here, so thinking.type is the only real control.
 base_model = "bytedance-seed/seed-1-6-flash"
+reasoning = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.02227
diff --git a/providers/volcengine/models/doubao-seed-1-6-vision-250815.toml b/providers/volcengine/models/doubao-seed-1-6-vision-250815.toml
index e720b45e8a2..981343233eb 100644
--- a/providers/volcengine/models/doubao-seed-1-6-vision-250815.toml
+++ b/providers/volcengine/models/doubao-seed-1-6-vision-250815.toml
@@ -1,7 +1,22 @@
 # Model ID verified against POST /api/v3/chat/completions (2026-08-26): `doubao-seed-1-6-vision-250815`
 # CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
 # CNY list price source: https://www.volcengine.com/docs/82379/1544106
+# Toggle: thinking.type = enabled|disabled, POST /api/v3/chat/completions
+# (verified 2026-08-27: enabled -> 162 reasoning tokens, disabled -> 0). This model
+# reasons by default -- a bare request with no reasoning params returns 166 reasoning
+# tokens -- so the absence of a reasoning declaration here was wrong.
+# No graded effort: reasoning_effort is accepted but not separable. Over n=4 runs per
+# level, none 895/927/670/736, low 550/901/775/598, high 578/738/787/1031,
+# max 852/813/1011/699 -- fully overlapping, and 'none' does not switch thinking off,
+# so thinking.type is the only real control.
 base_model = "bytedance-seed/seed-1-6-vision"
+reasoning = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.11875

From 2c9e70853665d2a95bfd406f8af3ad7a696f85b5 Mon Sep 17 00:00:00 2001
From: Andrew Szeto <github@jabagawee.com>
Date: Thu, 27 Aug 2026 17:31:57 -0700
Subject: [PATCH 0096/2012] fix(sync): reject synced model paths with
 case-collisions (#5615)

Previous PRs #5318 and #5517 fixed an issue where model entries collided on
case-insensitive filesystems (like the default one for MacOS), but they only
fixed it specifically for the Eden AI case. This commit adds a check to handle
it more generally.
---
 packages/core/src/sync/index.ts | 11 +++++--
 packages/core/test/sync.test.ts | 55 +++++++++++++++++++++++++++++++++
 2 files changed, 64 insertions(+), 2 deletions(-)

diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index c5f116f377a..888773f3dd5 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -236,6 +236,7 @@ export async function syncProvider<SourceModel>(
   let { modelMetadata } = existingState;
   const sourceModels = provider.parseModels(await provider.fetchModels());
   const desired = new Map<string, { model: z.infer<typeof SyncedAuthoredModel>; content: string }>();
+  const caseNormalizedDesiredPaths = new Map<string, string>();
   const desiredMetadata = new Map<string, { model: z.infer<typeof ModelMetadata>; content: string }>();
   const skippedRemote: string[] = [];
 
@@ -260,9 +261,15 @@ export async function syncProvider<SourceModel>(
       continue;
     }
 
-    if (desired.has(relativePath)) {
-      throw new Error(`Duplicate synced model path: ${provider.id}/${relativePath}`);
+    const collidingPath = caseNormalizedDesiredPaths.get(relativePath.toLowerCase());
+    if (collidingPath !== undefined) {
+      throw new Error(
+        collidingPath === relativePath
+          ? `Duplicate synced model path: ${provider.id}/${relativePath}`
+          : `Synced model paths differ only in case: ${provider.id}/${collidingPath} and ${provider.id}/${relativePath}`,
+      );
     }
+    caseNormalizedDesiredPaths.set(relativePath.toLowerCase(), relativePath);
 
     if (translated.metadata !== undefined) {
       const parsedMetadata = ModelMetadata.safeParse({
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 692aea87134..987dd262877 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -4731,3 +4731,58 @@ function openRouterModel(overrides: Partial<OpenRouterModel> = {}): OpenRouterMo
     ...overrides,
   };
 }
+
+function caseFoldProvider(modelsDir: string, ids: string[]): SyncProvider<string> {
+  return {
+    id: "case-fold-test",
+    name: "Case fold test",
+    modelsDir,
+    async fetchModels() {
+      return ids;
+    },
+    parseModels(raw) {
+      return raw as string[];
+    },
+    translateModel(id) {
+      return {
+        id,
+        model: {
+          name: id,
+          description: "Case-fold guard test model.",
+          release_date: "2026-08-14",
+          last_updated: "2026-08-14",
+          attachment: false,
+          reasoning: false,
+          tool_call: false,
+          open_weights: false,
+          cost: { input: 1, output: 2 },
+          limit: { context: 8_192, output: 4_096 },
+          modalities: { input: ["text"], output: ["text"] },
+        },
+      };
+    },
+  };
+}
+
+test("rejects synced model paths that differ only in case", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "models-dev-case-fold-"));
+  const modelsDir = path.join(root, "providers", "case-fold-test", "models");
+  await mkdir(modelsDir, { recursive: true });
+
+  try {
+    await expect(
+      syncProvider(caseFoldProvider(modelsDir, ["Alpha", "alpha"])),
+    ).rejects.toThrow(/differ only in case/u);
+
+    await expect(
+      syncProvider(caseFoldProvider(modelsDir, ["beta", "beta"])),
+    ).rejects.toThrow(/Duplicate synced model path/u);
+
+    const clean = await syncProvider(
+      caseFoldProvider(modelsDir, ["Gamma", "delta"]),
+    );
+    expect(clean).toMatchObject({ created: 2, updated: 0, deleted: 0 });
+  } finally {
+    await rm(root, { recursive: true, force: true });
+  }
+});

From b6e99de447313e3c902fa156776f75788daa6861 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Fri, 28 Aug 2026 07:32:13 +0700
Subject: [PATCH 0097/2012] feat(nvidia): add deepseek-v4-pro-0813 provider
 entry (#5624)

- base_model deepseek/deepseek-v4-pro-0813 with override-only provider fields
- reasoning_effort none|high|max per NIM chat schema, interleaved reasoning_content
- cost 0/0 trial tier (no separate list rate on build.nvidia.com card)
- validated via live endpoint tests (top-level reasoning_effort none/high/max, chat_template_kwargs thinking)

Refs: https://build.nvidia.com/deepseek-ai/deepseek-v4-pro-0813
Precedent: #5288 (flash-0731)

Co-authored-by: 4ebuRushka <4eburushka@users.noreply.github.com>
---
 .../deepseek-ai/deepseek-v4-pro-0813.toml       | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/nvidia/models/deepseek-ai/deepseek-v4-pro-0813.toml

diff --git a/providers/nvidia/models/deepseek-ai/deepseek-v4-pro-0813.toml b/providers/nvidia/models/deepseek-ai/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..c9b6566a09d
--- /dev/null
+++ b/providers/nvidia/models/deepseek-ai/deepseek-v4-pro-0813.toml
@@ -0,0 +1,17 @@
+# NIM Chat schema: `reasoning_effort = none|high|max`; `none` disables thinking.
+# https://build.nvidia.com/deepseek-ai/deepseek-v4-pro-0813
+#
+# Pricing: hosted on NVIDIA's API trial tier and currently free — no separate
+# list rate is published for this ID. Source: this model's catalog card,
+# https://build.nvidia.com/deepseek-ai/deepseek-v4-pro-0813 (governed by the
+# NVIDIA API Trial Terms of Service).
+base_model = "deepseek/deepseek-v4-pro-0813"
+
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0

From ab95d3c3dda70eb6fe05852730c839af025614bc Mon Sep 17 00:00:00 2001
From: Zikri <161694337+Zikri809@users.noreply.github.com>
Date: Fri, 28 Aug 2026 08:33:16 +0800
Subject: [PATCH 0098/2012] Add missing Kenari models (#5612)

* Add missing Kenari models

* Remove Kenari models with unresolved reasoning controls
---
 providers/kenari/models/claude-opus-5.toml           |  9 +++++++++
 providers/kenari/models/claude-sonnet-4-6.toml       |  9 +++++++++
 providers/kenari/models/gemini-3-1-flash-tts.toml    |  5 +++++
 providers/kenari/models/gemini-3-1-pro.toml          |  9 +++++++++
 providers/kenari/models/gemini-3-5-flash.toml        |  9 +++++++++
 providers/kenari/models/gemini-3-6-flash.toml        |  9 +++++++++
 providers/kenari/models/gemini-3-7-flash.toml        |  9 +++++++++
 providers/kenari/models/glm-5-3-flash.toml           |  9 +++++++++
 providers/kenari/models/glm-5-3.toml                 |  9 +++++++++
 providers/kenari/models/grok-4-6.toml                |  9 +++++++++
 providers/kenari/models/grok-imagine-image-2-0.toml  |  5 +++++
 providers/kenari/models/hy3.toml                     |  9 +++++++++
 providers/kenari/models/hy3:free.toml                | 10 ++++++++++
 providers/kenari/models/minimax-m2-7-highspeed.toml  |  6 ++++++
 providers/kenari/models/minimax-m2-7.toml            |  6 ++++++
 providers/kenari/models/mistral-large:free.toml      |  6 ++++++
 providers/kenari/models/mistral-medium-3-5:free.toml | 10 ++++++++++
 providers/kenari/models/qwen3-8-max.toml             |  9 +++++++++
 providers/kenari/models/step-3-7-flash.toml          |  9 +++++++++
 providers/kenari/models/step-3-7-flash:free.toml     | 10 ++++++++++
 providers/kenari/models/whisper-large-v3-turbo.toml  |  5 +++++
 21 files changed, 171 insertions(+)
 create mode 100644 providers/kenari/models/claude-opus-5.toml
 create mode 100644 providers/kenari/models/claude-sonnet-4-6.toml
 create mode 100644 providers/kenari/models/gemini-3-1-flash-tts.toml
 create mode 100644 providers/kenari/models/gemini-3-1-pro.toml
 create mode 100644 providers/kenari/models/gemini-3-5-flash.toml
 create mode 100644 providers/kenari/models/gemini-3-6-flash.toml
 create mode 100644 providers/kenari/models/gemini-3-7-flash.toml
 create mode 100644 providers/kenari/models/glm-5-3-flash.toml
 create mode 100644 providers/kenari/models/glm-5-3.toml
 create mode 100644 providers/kenari/models/grok-4-6.toml
 create mode 100644 providers/kenari/models/grok-imagine-image-2-0.toml
 create mode 100644 providers/kenari/models/hy3.toml
 create mode 100644 providers/kenari/models/hy3:free.toml
 create mode 100644 providers/kenari/models/minimax-m2-7-highspeed.toml
 create mode 100644 providers/kenari/models/minimax-m2-7.toml
 create mode 100644 providers/kenari/models/mistral-large:free.toml
 create mode 100644 providers/kenari/models/mistral-medium-3-5:free.toml
 create mode 100644 providers/kenari/models/qwen3-8-max.toml
 create mode 100644 providers/kenari/models/step-3-7-flash.toml
 create mode 100644 providers/kenari/models/step-3-7-flash:free.toml
 create mode 100644 providers/kenari/models/whisper-large-v3-turbo.toml

diff --git a/providers/kenari/models/claude-opus-5.toml b/providers/kenari/models/claude-opus-5.toml
new file mode 100644
index 00000000000..162ff1c21df
--- /dev/null
+++ b/providers/kenari/models/claude-opus-5.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-5"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/claude-sonnet-4-6.toml b/providers/kenari/models/claude-sonnet-4-6.toml
new file mode 100644
index 00000000000..a35b82de825
--- /dev/null
+++ b/providers/kenari/models/claude-sonnet-4-6.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-sonnet-4-6"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/gemini-3-1-flash-tts.toml b/providers/kenari/models/gemini-3-1-flash-tts.toml
new file mode 100644
index 00000000000..3ce0ea388d5
--- /dev/null
+++ b/providers/kenari/models/gemini-3-1-flash-tts.toml
@@ -0,0 +1,5 @@
+base_model = "google/gemini-3.1-flash-tts-preview"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/gemini-3-1-pro.toml b/providers/kenari/models/gemini-3-1-pro.toml
new file mode 100644
index 00000000000..fc8387d3d92
--- /dev/null
+++ b/providers/kenari/models/gemini-3-1-pro.toml
@@ -0,0 +1,9 @@
+base_model = "google/gemini-3.1-pro-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/gemini-3-5-flash.toml b/providers/kenari/models/gemini-3-5-flash.toml
new file mode 100644
index 00000000000..32d60ab7cf4
--- /dev/null
+++ b/providers/kenari/models/gemini-3-5-flash.toml
@@ -0,0 +1,9 @@
+base_model = "google/gemini-3.5-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/gemini-3-6-flash.toml b/providers/kenari/models/gemini-3-6-flash.toml
new file mode 100644
index 00000000000..76b7a0c3dfa
--- /dev/null
+++ b/providers/kenari/models/gemini-3-6-flash.toml
@@ -0,0 +1,9 @@
+base_model = "google/gemini-3.6-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/gemini-3-7-flash.toml b/providers/kenari/models/gemini-3-7-flash.toml
new file mode 100644
index 00000000000..da58f5c67fb
--- /dev/null
+++ b/providers/kenari/models/gemini-3-7-flash.toml
@@ -0,0 +1,9 @@
+base_model = "google/gemini-3.7-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/glm-5-3-flash.toml b/providers/kenari/models/glm-5-3-flash.toml
new file mode 100644
index 00000000000..98ace03c3eb
--- /dev/null
+++ b/providers/kenari/models/glm-5-3-flash.toml
@@ -0,0 +1,9 @@
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/glm-5-3.toml b/providers/kenari/models/glm-5-3.toml
new file mode 100644
index 00000000000..9b34f5dc786
--- /dev/null
+++ b/providers/kenari/models/glm-5-3.toml
@@ -0,0 +1,9 @@
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/grok-4-6.toml b/providers/kenari/models/grok-4-6.toml
new file mode 100644
index 00000000000..a8dfff08ee9
--- /dev/null
+++ b/providers/kenari/models/grok-4-6.toml
@@ -0,0 +1,9 @@
+base_model = "xai/grok-4.6"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/grok-imagine-image-2-0.toml b/providers/kenari/models/grok-imagine-image-2-0.toml
new file mode 100644
index 00000000000..28f9c82aadc
--- /dev/null
+++ b/providers/kenari/models/grok-imagine-image-2-0.toml
@@ -0,0 +1,5 @@
+base_model = "xai/grok-imagine-image-2.0"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/hy3.toml b/providers/kenari/models/hy3.toml
new file mode 100644
index 00000000000..aec067fe7d2
--- /dev/null
+++ b/providers/kenari/models/hy3.toml
@@ -0,0 +1,9 @@
+base_model = "tencent/hy3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/hy3:free.toml b/providers/kenari/models/hy3:free.toml
new file mode 100644
index 00000000000..60e6b14beb9
--- /dev/null
+++ b/providers/kenari/models/hy3:free.toml
@@ -0,0 +1,10 @@
+base_model = "tencent/hy3"
+name = "Hy3 (Free)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/minimax-m2-7-highspeed.toml b/providers/kenari/models/minimax-m2-7-highspeed.toml
new file mode 100644
index 00000000000..d4d61741ca8
--- /dev/null
+++ b/providers/kenari/models/minimax-m2-7-highspeed.toml
@@ -0,0 +1,6 @@
+base_model = "minimax/MiniMax-M2.7-highspeed"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/minimax-m2-7.toml b/providers/kenari/models/minimax-m2-7.toml
new file mode 100644
index 00000000000..04bde4fc751
--- /dev/null
+++ b/providers/kenari/models/minimax-m2-7.toml
@@ -0,0 +1,6 @@
+base_model = "minimax/MiniMax-M2.7"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/mistral-large:free.toml b/providers/kenari/models/mistral-large:free.toml
new file mode 100644
index 00000000000..49a8f426fd5
--- /dev/null
+++ b/providers/kenari/models/mistral-large:free.toml
@@ -0,0 +1,6 @@
+base_model = "mistral/mistral-large-latest"
+name = "Mistral Large (Free)"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/mistral-medium-3-5:free.toml b/providers/kenari/models/mistral-medium-3-5:free.toml
new file mode 100644
index 00000000000..9f2bf1039bd
--- /dev/null
+++ b/providers/kenari/models/mistral-medium-3-5:free.toml
@@ -0,0 +1,10 @@
+base_model = "mistral/mistral-medium-2604"
+name = "Mistral Medium 3.5 (Free)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/qwen3-8-max.toml b/providers/kenari/models/qwen3-8-max.toml
new file mode 100644
index 00000000000..88171d4a690
--- /dev/null
+++ b/providers/kenari/models/qwen3-8-max.toml
@@ -0,0 +1,9 @@
+base_model = "alibaba/qwen3.8-max"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/step-3-7-flash.toml b/providers/kenari/models/step-3-7-flash.toml
new file mode 100644
index 00000000000..778be2e4008
--- /dev/null
+++ b/providers/kenari/models/step-3-7-flash.toml
@@ -0,0 +1,9 @@
+base_model = "stepfun/step-3.7-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/step-3-7-flash:free.toml b/providers/kenari/models/step-3-7-flash:free.toml
new file mode 100644
index 00000000000..5f90458c8df
--- /dev/null
+++ b/providers/kenari/models/step-3-7-flash:free.toml
@@ -0,0 +1,10 @@
+base_model = "stepfun/step-3.7-flash"
+name = "Step 3.7 Flash (Free)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/kenari/models/whisper-large-v3-turbo.toml b/providers/kenari/models/whisper-large-v3-turbo.toml
new file mode 100644
index 00000000000..0150da262d9
--- /dev/null
+++ b/providers/kenari/models/whisper-large-v3-turbo.toml
@@ -0,0 +1,5 @@
+base_model = "openai/whisper-large-v3-turbo"
+
+[cost]
+input = 0
+output = 0

From 2aaf0276b5cd656db75e4fe714b8aebd52894a82 Mon Sep 17 00:00:00 2001
From: Ahmad Shahzad <107808273+shzdehmd@users.noreply.github.com>
Date: Fri, 28 Aug 2026 06:08:12 +0500
Subject: [PATCH 0099/2012] chore(fireworks-ai): remove deprecated serverless
 models (08/27/2026) (#5683)

Remove models decommissioned from serverless endpoints on 2026-08-27:
- MiniMax M2.7
- GPT OSS 20B
- Kimi K2.6 Fast
- Kimi K2.6 Turbo
- Kimi K2.7 Code Fast
- DeepSeek V4 Pro
---
 .../fireworks/models/deepseek-v4-pro.toml     | 18 -----------
 .../fireworks/models/gpt-oss-20b.toml         | 29 -----------------
 .../fireworks/models/minimax-m2p7.toml        | 32 -------------------
 .../fireworks/routers/kimi-k2p6-fast.toml     | 31 ------------------
 .../fireworks/routers/kimi-k2p6-turbo.toml    | 31 ------------------
 .../routers/kimi-k2p7-code-fast.toml          | 31 ------------------
 6 files changed, 172 deletions(-)
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-20b.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-fast.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-turbo.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p7-code-fast.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
deleted file mode 100644
index eda5cdf4d64..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "deepseek/deepseek-v4-pro"
-# Fireworks' published schema does not document this model's toggle or "max"
-# effort mapping. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.74
-output = 3.48
-cache_read = 0.145
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-20b.toml b/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-20b.toml
deleted file mode 100644
index e125b71c646..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-20b.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-name = "GPT OSS 20B"
-description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-# $.reasoning_effort = "low" | "medium" | "high".
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-family = "gpt-oss"
-release_date = "2025-08-05"
-last_updated = "2025-08-05"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.07
-output = 0.30
-cache_read = 0.035
-
-[limit]
-context = 131_072
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
deleted file mode 100644
index 2fc97e864b0..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
+++ /dev/null
@@ -1,32 +0,0 @@
-name = "MiniMax-M2.7"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-# $.reasoning_effort = "low" | "medium" | "high".
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-family = "minimax"
-release_date = "2026-04-12"
-last_updated = "2026-04-12"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.30
-output = 1.20
-cache_read = 0.06
-
-[limit]
-context = 196_608
-output = 196_608
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-fast.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-fast.toml
deleted file mode 100644
index 52ddcd2c874..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-fast.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Kimi K2.6 Fast"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
-# No router-specific toggle mapping or explicit disabled value is documented.
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-family = "kimi-thinking"
-release_date = "2026-04-17"
-last_updated = "2026-06-05"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-cache_read = 0.30
-input = 2.00
-output = 8.00
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-turbo.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-turbo.toml
deleted file mode 100644
index 87e3f6b32bb..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p6-turbo.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Kimi K2.6 Turbo"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
-# No router-specific toggle mapping or explicit disabled value is documented.
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-family = "kimi-thinking"
-release_date = "2026-04-17"
-last_updated = "2026-04-17"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-cache_read = 0.30
-input = 2.00
-output = 8.00
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p7-code-fast.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p7-code-fast.toml
deleted file mode 100644
index 0cbc6520216..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p7-code-fast.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Kimi K2.7 Code Fast"
-description = "Kimi coding model for software agents, refactors, and repository reasoning"
-# No router-specific toggle mapping or explicit disabled value is documented.
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-family = "kimi-k2"
-release_date = "2026-06-12"
-last_updated = "2026-06-16"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-cache_read = 0.38
-input = 1.90
-output = 8.00
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

From 4d4e34c8042675fb26123719988726297c7ca04e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 01:56:53 +0000
Subject: [PATCH 0100/2012] chore(sync): update OpenRouter model catalog
 (#5689)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml          | 6 +++---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 9 +++++----
 .../openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml  | 2 +-
 .../openrouter/models/nvidia/nemotron-3.5-lightning.toml | 8 ++++----
 providers/openrouter/models/qwen/qwen3.6-27b.toml        | 7 +++----
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml  | 1 +
 providers/openrouter/models/qwen/qwen3.8-27b.toml        | 7 ++++---
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 10 files changed, 30 insertions(+), 28 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 1273df92943..df31c0bbb32 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.01
+input = 0.06
+output = 0.12
+cache_read = 0.012
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 8076cd981ad..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0749
-output = 0.1498
-cache_read = 0.01498
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 309ffa254af..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-pro-0813"
 
 [[reasoning_options]]
@@ -8,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
-cache_read = 0.0374
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 943_717
diff --git a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
index 2a0c70fda72..9f459a478b0 100644
--- a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
+++ b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
@@ -16,7 +16,7 @@ output = 2.96
 cache_read = 0.15
 
 [limit]
-context = 256_000
+context = 262_144
 output = 80_000
 
 [modalities]
diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
index 3b251679339..757af4e5d09 100644
--- a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
@@ -7,9 +7,9 @@ description = "Nemotron model for efficient reasoning, coding, and specialized A
 type = "toggle"
 
 [cost]
-input = 0.08
-output = 0.2
-cache_read = 0.04
+input = 0.1
+output = 0.25
+cache_read = 0.05
 
 [limit]
-output = 131_072
+output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index d35b93066de..68d7d650159 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,12 +6,11 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.6
-output = 3.6
-cache_read = 0.12
+input = 0.32
+output = 3.2
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index 74b5d1276f6..abacdd68c88 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -18,3 +18,4 @@ cache_read = 0.25
 
 [limit]
 context = 1_048_576
+output = 262_144
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index f443508a413..c1caf217614 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.8-27b"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 
@@ -9,10 +11,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.425
+input = 0.4
 output = 2.55
-cache_read = 0.085
-cache_write = 0.53125
+cache_read = 0.05
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 1cdebb561570e0ae55202f94c37facb685db329b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 01:57:02 +0000
Subject: [PATCH 0101/2012] chore(sync): update Kilo model catalog (#5687)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml    | 3 +--
 providers/kilo/models/meta-llama/llama-4-maverick.toml      | 2 +-
 providers/kilo/models/nvidia/nemotron-3.5-lightning.toml    | 2 +-
 providers/kilo/models/qwen/qwen3.6-27b.toml                 | 2 +-
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml           | 3 ++-
 providers/kilo/models/qwen/qwen3.8-27b.toml                 | 1 -
 providers/kilo/models/tencent/hy3.toml                      | 6 +++---
 providers/kilo/models/z-ai/glm-5.3-flash.toml               | 6 +++---
 9 files changed, 15 insertions(+), 16 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..e79a3173757 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 7f2ba8568dd..778f98b65d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_048_575
-output = 943_717
+context = 1_048_576
diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 3db8291511b..5443201b6e8 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [cost]
 input = 0.2
-output = 0.696
+output = 0.8
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index f6b9541a819..28181cd4e48 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -11,4 +11,4 @@ output = 0.2
 cache_read = 0.04
 
 [limit]
-output = 131_072
+output = 235_929
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index 7aaf82ee885..fdb3246f594 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -10,7 +10,7 @@ input = 0.45
 output = 2.7
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index 2e2c290259f..c79bdfce589 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -15,4 +15,5 @@ cache_read = 0.25
 cache_write = 2.5
 
 [limit]
-context = 1_048_576
+context = 1_000_000
+output = 262_144
diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 820f69fb194..9b18fb98b33 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,5 +12,4 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
-context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 731fb78b785..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 0ed9d5f31a9..0104ea56ade 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_048_576

From 41d4cd606b858fc3c9d2a4849c9f508033416908 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 01:57:04 +0000
Subject: [PATCH 0102/2012] chore(sync): update NanoGPT model catalog (#5688)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 6fa95040b26f7f0a6c57c78216c2e6111da4b771 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 01:57:06 +0000
Subject: [PATCH 0103/2012] chore(sync): update Merge Gateway model catalog
 (#5690)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 3 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 16d15cbb411..c680bd0f847 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +13,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index faa70b7c1e0..3ef83d21d9d 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 35e52350151..f30ae74cdb3 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From f36210a4bf1cbeb2331ff5d6340c24553477d9f1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 01:57:24 +0000
Subject: [PATCH 0104/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5685)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/deepseek/deepseek-v4-pro.toml | 10 +++-------
 1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/providers/vercel/models/deepseek/deepseek-v4-pro.toml b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
index 917089fcfb7..d1c7fbe6694 100644
--- a/providers/vercel/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
@@ -8,10 +8,6 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.74
-output = 3.48
-cache_read = 0.14
-
-[limit]
-context = 1_048_600
-output = 1_048_600
+input = 0.66
+output = 1.98
+cache_read = 0.022

From e6b26ff747b884afd23d3d19c4df4bb320c2e230 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:16:47 -0500
Subject: [PATCH 0105/2012] chore(sync): update DigitalOcean model catalog
 (#5686)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../digitalocean/models/deepseek-3.2.toml      |  6 +++---
 .../digitalocean/models/deepseek-4-flash.toml  |  6 +++---
 .../models/deepseek-v4-flash-0731.toml         |  6 +++---
 .../digitalocean/models/deepseek-v4-pro.toml   |  6 +++---
 providers/digitalocean/models/glm-5.2.toml     |  6 +++---
 .../digitalocean/models/glm-5.3-flash.toml     | 18 ++++++++++++++++++
 providers/digitalocean/models/kimi-k3.toml     |  6 +++---
 .../digitalocean/models/llama-4-maverick.toml  |  4 ++--
 .../digitalocean/models/mimo-v2.5-pro.toml     |  6 +++---
 .../models/openai-gpt-5.6-sol.toml             | 12 ++++++------
 .../models/openai-gpt-oss-120b.toml            |  4 ++--
 11 files changed, 49 insertions(+), 31 deletions(-)
 create mode 100644 providers/digitalocean/models/glm-5.3-flash.toml

diff --git a/providers/digitalocean/models/deepseek-3.2.toml b/providers/digitalocean/models/deepseek-3.2.toml
index 86a0601e001..d8d92fa42fa 100644
--- a/providers/digitalocean/models/deepseek-3.2.toml
+++ b/providers/digitalocean/models/deepseek-3.2.toml
@@ -18,9 +18,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 0.8
-cache_read = 0.075
+input = 0.5
+output = 1.6
+cache_read = 0.15
 
 [limit]
 context = 163_840
diff --git a/providers/digitalocean/models/deepseek-4-flash.toml b/providers/digitalocean/models/deepseek-4-flash.toml
index 2e033308f91..143346a359b 100644
--- a/providers/digitalocean/models/deepseek-4-flash.toml
+++ b/providers/digitalocean/models/deepseek-4-flash.toml
@@ -10,9 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0.0679
-output = 0.168
-cache_read = 0.0168
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-flash-0731.toml b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
index 74ce3b0ae91..2e9e64c5632 100644
--- a/providers/digitalocean/models/deepseek-v4-flash-0731.toml
+++ b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.08
-output = 0.252
-cache_read = 0.0252
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-pro.toml b/providers/digitalocean/models/deepseek-v4-pro.toml
index 8a4edb98729..b4ebe0207c1 100644
--- a/providers/digitalocean/models/deepseek-v4-pro.toml
+++ b/providers/digitalocean/models/deepseek-v4-pro.toml
@@ -19,9 +19,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.174
+input = 1.74
+output = 3.48
+cache_read = 0.348
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/glm-5.2.toml b/providers/digitalocean/models/glm-5.2.toml
index 3f01602c179..eb3b3d1b78d 100644
--- a/providers/digitalocean/models/glm-5.2.toml
+++ b/providers/digitalocean/models/glm-5.2.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["medium", "high", "xhigh"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.105
+input = 1.4
+output = 4.4
+cache_read = 0.21
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/glm-5.3-flash.toml b/providers/digitalocean/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..f0bf95a9448
--- /dev/null
+++ b/providers/digitalocean/models/glm-5.3-flash.toml
@@ -0,0 +1,18 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM5.3 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["medium", "high", "xhigh"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/digitalocean/models/kimi-k3.toml b/providers/digitalocean/models/kimi-k3.toml
index b71a3e29e2e..62d0f7b7404 100644
--- a/providers/digitalocean/models/kimi-k3.toml
+++ b/providers/digitalocean/models/kimi-k3.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.85
-output = 14.25
-cache_read = 0.285
+input = 3
+output = 15
+cache_read = 0.3
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/digitalocean/models/llama-4-maverick.toml b/providers/digitalocean/models/llama-4-maverick.toml
index a72af75a264..095b2da6674 100644
--- a/providers/digitalocean/models/llama-4-maverick.toml
+++ b/providers/digitalocean/models/llama-4-maverick.toml
@@ -12,8 +12,8 @@ knowledge = "2024-08"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.696
+input = 0.25
+output = 0.87
 
 [limit]
 context = 128_000
diff --git a/providers/digitalocean/models/mimo-v2.5-pro.toml b/providers/digitalocean/models/mimo-v2.5-pro.toml
index 20f6f6db3ea..365235fb39f 100644
--- a/providers/digitalocean/models/mimo-v2.5-pro.toml
+++ b/providers/digitalocean/models/mimo-v2.5-pro.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.4
-output = 1.5
-cache_read = 0.08
+input = 0.8
+output = 3
+cache_read = 0.16
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/openai-gpt-5.6-sol.toml b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
index e14ff09a4ee..3c9af7df850 100644
--- a/providers/digitalocean/models/openai-gpt-5.6-sol.toml
+++ b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
@@ -6,15 +6,15 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 4
-output = 20
-cache_read = 0.4
+input = 5
+output = 30
+cache_read = 0.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 8
-output = 30
-cache_read = 0.8
+input = 10
+output = 45
+cache_read = 1
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/digitalocean/models/openai-gpt-oss-120b.toml b/providers/digitalocean/models/openai-gpt-oss-120b.toml
index fc45d3aac8b..4af4fef37ad 100644
--- a/providers/digitalocean/models/openai-gpt-oss-120b.toml
+++ b/providers/digitalocean/models/openai-gpt-oss-120b.toml
@@ -19,8 +19,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.055
-output = 0.385
+input = 0.1
+output = 0.7
 cache_read = 0.02
 
 [limit]

From 7a724909f311dc88f500d505ddea9c7f489a509b Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:17:03 -0500
Subject: [PATCH 0106/2012] fix: [missing-model] ofox: z-ai/glm-5.3-flash
 (#5611)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/z-ai/glm-5.3-flash.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/ofox/models/z-ai/glm-5.3-flash.toml

diff --git a/providers/ofox/models/z-ai/glm-5.3-flash.toml b/providers/ofox/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..ead3405dc30
--- /dev/null
+++ b/providers/ofox/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,15 @@
+# Sources:
+# - https://ofox.ai/en/models/z-ai/glm-5.3-flash (Ofox: $0.075/$0.25/M, cache read $0.015/M at -50%; 1M context; 131K output; always-on thinking with low/high/max effort)
+# - https://docs.z.ai/guides/overview/pricing (Z.AI list/promo: GLM-5.3-Flash $0.075/$0.25/M, cached input $0.015/M; 50% promo ends 2026-09-09 24:00 UTC+8)
+# - https://docs.z.ai/guides/llm/glm-5.3-flash (thinking.type only enabled; cannot disable; reasoning_effort recommended max)
+# Effort: reasoning_effort = low|high|max (thinking always on)
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015

From 4daf03dcfe812456014af4d736b9f91887189a90 Mon Sep 17 00:00:00 2001
From: mickalchen <mickalchen@tencent.com>
Date: Fri, 28 Aug 2026 11:54:23 +0800
Subject: [PATCH 0107/2012] add Hy4 preview on Tencent TokenHub and Token Plan

Keep the catalog entries on an unmerged branch so they are not loaded by models.dev until the model is ready to publish.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 models/tencent/hy4-preview.toml               | 18 +++++++++++++++++
 .../models/hy4-preview.toml                   | 20 +++++++++++++++++++
 .../tencent-tokenhub/models/hy4-preview.toml  | 20 +++++++++++++++++++
 3 files changed, 58 insertions(+)
 create mode 100644 models/tencent/hy4-preview.toml
 create mode 100644 providers/tencent-token-plan/models/hy4-preview.toml
 create mode 100644 providers/tencent-tokenhub/models/hy4-preview.toml

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..19f5c9a36ec
--- /dev/null
+++ b/models/tencent/hy4-preview.toml
@@ -0,0 +1,18 @@
+name = "Hy4 preview"
+description = "A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities."
+family = "Hy"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_024_000
+output = 64_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/tencent-token-plan/models/hy4-preview.toml b/providers/tencent-token-plan/models/hy4-preview.toml
new file mode 100644
index 00000000000..2a75bc2426a
--- /dev/null
+++ b/providers/tencent-token-plan/models/hy4-preview.toml
@@ -0,0 +1,20 @@
+# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
+# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
+# Catalog is USD / million tokens (×1000).
+# Toggle: thinking.type = enabled|disabled (default disabled)
+# Effort: reasoning_effort = none|high
+# https://cloud.tencent.com/document/product/1823/130055
+# https://cloud.tencent.com/document/product/1823/131208
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
diff --git a/providers/tencent-tokenhub/models/hy4-preview.toml b/providers/tencent-tokenhub/models/hy4-preview.toml
new file mode 100644
index 00000000000..2a75bc2426a
--- /dev/null
+++ b/providers/tencent-tokenhub/models/hy4-preview.toml
@@ -0,0 +1,20 @@
+# China-site list CNY → USD: list_price / tax_rate * FX (accessed 2026-08-26).
+# Source USD/1k tokens: input 0.000834, output 0.002501, cache hit 0.000042.
+# Catalog is USD / million tokens (×1000).
+# Toggle: thinking.type = enabled|disabled (default disabled)
+# Effort: reasoning_effort = none|high
+# https://cloud.tencent.com/document/product/1823/130055
+# https://cloud.tencent.com/document/product/1823/131208
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042

From f2c217d47c92237a42d2d6e261f7329e366e203d Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 27 Aug 2026 22:55:22 -0500
Subject: [PATCH 0108/2012] =?UTF-8?q?fix:=20=E9=98=BF=E9=87=8C=E5=B7=B4?=
 =?UTF-8?q?=E5=B7=B4tokenplan=E9=87=8C=E5=BA=94=E6=9C=89qwen3.8-flash=20(#?=
 =?UTF-8?q?5657)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/qwen3.8-flash.toml                 | 25 +++++++++++++++++++
 .../models/qwen3.8-flash.toml                 | 24 ++++++++++++++++++
 2 files changed, 49 insertions(+)
 create mode 100644 providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
 create mode 100644 providers/alibaba-token-plan/models/qwen3.8-flash.toml

diff --git a/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml b/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..7f93b2a617c
--- /dev/null
+++ b/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
@@ -0,0 +1,25 @@
+# Sources (accessed 2026-08-27):
+# https://help.aliyun.com/zh/model-studio/token-plan-personal-overview
+# https://help.aliyun.com/en/model-studio/token-plan-personal-overview
+# https://www.qwencloud.com/models/qwen3.8-flash
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# https://docs.qwencloud.com/api-reference/chat/openai-chat
+# Toggle: enable_thinking true|false (hybrid)
+# Budget: thinking_budget — model page lists Max Reasoning 262K
+# reasoning_effort is documented for qwen3.8-max / qwen3.8-max-preview only, not flash.
+# Reasoning streams in reasoning_content.
+# Cost is Credits-subscription (no public USD per-token rate on Token Plan).
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", max = 262_144 },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
diff --git a/providers/alibaba-token-plan/models/qwen3.8-flash.toml b/providers/alibaba-token-plan/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..9d3107caa65
--- /dev/null
+++ b/providers/alibaba-token-plan/models/qwen3.8-flash.toml
@@ -0,0 +1,24 @@
+# Sources (accessed 2026-08-27):
+# https://docs.qwencloud.com/token-plan/personal/token-plan-personal-overview
+# https://www.qwencloud.com/models/qwen3.8-flash
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# https://docs.qwencloud.com/api-reference/chat/openai-chat
+# Toggle: enable_thinking true|false (hybrid)
+# Budget: thinking_budget — model page lists Max Reasoning 262K
+# reasoning_effort is documented for qwen3.8-max / qwen3.8-max-preview only, not flash.
+# Reasoning streams in reasoning_content.
+# Cost is Credits-subscription (no public USD per-token rate on Token Plan).
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", max = 262_144 },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From a6705529c698e573343cd849ba09881bc116ba61 Mon Sep 17 00:00:00 2001
From: mickalchen <mickalchen@tencent.com>
Date: Fri, 28 Aug 2026 12:44:15 +0800
Subject: [PATCH 0109/2012] mark Hy4 preview as open weights

Hy4 preview ships public weights; align the lab entry with Hy3 preview.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 models/tencent/hy4-preview.toml | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
index 19f5c9a36ec..136a2dc83a4 100644
--- a/models/tencent/hy4-preview.toml
+++ b/models/tencent/hy4-preview.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_024_000
@@ -16,3 +16,7 @@ output = 64_000
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/tencent/Hy4-preview"

From 9f588170d298057203d27cec438d520cd01f861e Mon Sep 17 00:00:00 2001
From: mickalchen <mickalchen@tencent.com>
Date: Fri, 28 Aug 2026 12:54:44 +0800
Subject: [PATCH 0110/2012] drop Hy4 preview Hugging Face weights link

The weights page is not published yet; keep open_weights without a URL.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 models/tencent/hy4-preview.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/models/tencent/hy4-preview.toml b/models/tencent/hy4-preview.toml
index 136a2dc83a4..50bce9915d3 100644
--- a/models/tencent/hy4-preview.toml
+++ b/models/tencent/hy4-preview.toml
@@ -16,7 +16,3 @@ output = 64_000
 [modalities]
 input = ["text"]
 output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/tencent/Hy4-preview"

From b253ad0e2d6f62876c3e88d234a19dd7db171655 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 04:57:59 +0000
Subject: [PATCH 0111/2012] chore(sync): update NanoGPT model catalog (#5697)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 65dfa50bc0d4df3dc970bd741c22118890ecd493 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 04:58:05 +0000
Subject: [PATCH 0112/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5695)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/deepseek/deepseek-v4-pro.toml | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/providers/vercel/models/deepseek/deepseek-v4-pro.toml b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
index d1c7fbe6694..917089fcfb7 100644
--- a/providers/vercel/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
@@ -8,6 +8,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.74
+output = 3.48
+cache_read = 0.14
+
+[limit]
+context = 1_048_600
+output = 1_048_600

From 0af2a5de5235cdca97d70362f971658a655f2a08 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 04:58:08 +0000
Subject: [PATCH 0113/2012] chore(sync): update Kilo model catalog (#5693)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 providers/kilo/models/google/gemma-4-31b-it.toml            | 6 +++---
 providers/kilo/models/qwen/qwen3.8-27b.toml                 | 1 +
 providers/kilo/models/z-ai/glm-5.3-flash.toml               | 6 +++---
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index e79a3173757..6f876cf5199 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/google/gemma-4-31b-it.toml b/providers/kilo/models/google/gemma-4-31b-it.toml
index ccdbcb026b4..14b6cfb6e84 100644
--- a/providers/kilo/models/google/gemma-4-31b-it.toml
+++ b/providers/kilo/models/google/gemma-4-31b-it.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.09
-output = 0.34
-cache_read = 0.05
+input = 0.08
+output = 0.35
+cache_read = 0.01
 
 [limit]
 output = 16_384
diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 9b18fb98b33..820f69fb194 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,4 +12,5 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
+context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 0104ea56ade..0ed9d5f31a9 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 950fbab52a63f15a3efa997091e6f48c0005f4d9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 04:58:16 +0000
Subject: [PATCH 0114/2012] chore(sync): update CrossModel model catalog
 (#5694)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/z-ai/glm-5.3-flash.toml   | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/crossmodel/models/z-ai/glm-5.3-flash.toml

diff --git a/providers/crossmodel/models/z-ai/glm-5.3-flash.toml b/providers/crossmodel/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..b5971512328
--- /dev/null
+++ b/providers/crossmodel/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+cache_write = 0.075
+
+[limit]
+output = 128_000
+
+[modalities]
+input = ["text", "image", "video"]

From c6282e24a57f7ce5f4f57d2805b5a5df3d97b60d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 04:58:23 +0000
Subject: [PATCH 0115/2012] chore(sync): update OpenRouter model catalog
 (#5696)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-27b.toml           | 5 +++--
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index c1caf217614..d6e1375e402 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,9 +11,10 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.4
+input = 0.425
 output = 2.55
-cache_read = 0.05
+cache_read = 0.085
+cache_write = 0.53125
 
 [limit]
 context = 1_000_000

From 7431bd4b82a2df4500fe9c9bd40b33ca3ee4b7f9 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 28 Aug 2026 13:47:04 +0800
Subject: [PATCH 0116/2012] feat(go): add qwen3.8-flash

---
 providers/opencode-go/models/qwen3.8-flash.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/opencode-go/models/qwen3.8-flash.toml

diff --git a/providers/opencode-go/models/qwen3.8-flash.toml b/providers/opencode-go/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..73a89946990
--- /dev/null
+++ b/providers/opencode-go/models/qwen3.8-flash.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-flash"
+temperature = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file

From 4fa188de04db9d7c911e38c00c71f7fcadd01356 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:57:01 +0000
Subject: [PATCH 0117/2012] chore(sync): update Kilo model catalog (#5708)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 providers/kilo/models/qwen/qwen3.6-27b.toml   |  2 +-
 .../kilo/models/tencent/hy4-preview.toml      | 27 +++++++++++++++++++
 3 files changed, 31 insertions(+), 4 deletions(-)
 create mode 100644 providers/kilo/models/tencent/hy4-preview.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..e79a3173757 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index fdb3246f594..7aaf82ee885 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -10,7 +10,7 @@ input = 0.45
 output = 2.7
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy4-preview.toml b/providers/kilo/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..191cb9a173c
--- /dev/null
+++ b/providers/kilo/models/tencent/hy4-preview.toml
@@ -0,0 +1,27 @@
+name = "Tencent: Hy4 preview"
+description = "Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total. It is designed for coding agents, complex tool-use workflows, and productivity tasks that..."
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
+
+[limit]
+context = 1_048_576
+output = 64_000
+
+[modalities]
+input = ["text"]
+output = ["text"]

From e8af69043abcd2f95fcc9feccc3cd199bde954ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:57:04 +0000
Subject: [PATCH 0118/2012] chore(sync): update CrossModel model catalog
 (#5707)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/qwen/qwen3.8-flash.toml      | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/crossmodel/models/qwen/qwen3.8-flash.toml

diff --git a/providers/crossmodel/models/qwen/qwen3.8-flash.toml b/providers/crossmodel/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..96512c56fd5
--- /dev/null
+++ b/providers/crossmodel/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,14 @@
+base_model = "alibaba/qwen3.8-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.13
+output = 0.43
+cache_read = 0.016
+cache_write = 0.2

From 491da01704dc4069d475fe292b3730464b48f23f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:57:08 +0000
Subject: [PATCH 0119/2012] chore(sync): update Eden AI model catalog (#5706)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/ByteDance/Seed-2.0-code.toml      |  2 --
 .../models/deepinfra/ByteDance/Seed-2.0-mini.toml      |  3 +--
 providers/edenai/models/fireworks_ai/gpt-oss-20b.toml  | 10 ----------
 .../edenai/models/mistral/voxtral-small-latest.toml    |  9 +++++++++
 4 files changed, 10 insertions(+), 14 deletions(-)
 delete mode 100644 providers/edenai/models/fireworks_ai/gpt-oss-20b.toml
 create mode 100644 providers/edenai/models/mistral/voxtral-small-latest.toml

diff --git a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
index f0535988d69..6a4e99b9ea7 100644
--- a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
+++ b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
@@ -1,6 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-code"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
index 559b3a3a285..5b3a9d1765c 100644
--- a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
+++ b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
@@ -1,6 +1,5 @@
 base_model = "bytedance-seed/seed-2.0-mini"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/fireworks_ai/gpt-oss-20b.toml b/providers/edenai/models/fireworks_ai/gpt-oss-20b.toml
deleted file mode 100644
index 52e5e9d4321..00000000000
--- a/providers/edenai/models/fireworks_ai/gpt-oss-20b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "openai/gpt-oss-20b"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.07
-output = 0.3
-cache_read = 0.035
diff --git a/providers/edenai/models/mistral/voxtral-small-latest.toml b/providers/edenai/models/mistral/voxtral-small-latest.toml
new file mode 100644
index 00000000000..e092715c287
--- /dev/null
+++ b/providers/edenai/models/mistral/voxtral-small-latest.toml
@@ -0,0 +1,9 @@
+base_model = "mistral/voxtral-small-latest"
+structured_output = true
+
+[cost]
+input = 0.1
+output = 0.4
+
+[limit]
+context = 32_768

From f4ccabc02146f682146e74a5d186211caf885bbd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:57:14 +0000
Subject: [PATCH 0120/2012] chore(sync): update NanoGPT model catalog (#5709)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From a5361abe76fe968042e1dcba900698a77614b625 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 07:57:18 +0000
Subject: [PATCH 0121/2012] chore(sync): update OpenRouter model catalog
 (#5705)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731.toml      |  6 ++---
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++---
 .../openrouter/models/qwen/qwen3.6-27b.toml   |  7 ++---
 .../models/tencent/hy4-preview.toml           | 27 +++++++++++++++++++
 5 files changed, 40 insertions(+), 12 deletions(-)
 create mode 100644 providers/openrouter/models/tencent/hy4-preview.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index df31c0bbb32..06d23ff5e01 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.12
-cache_read = 0.012
+input = 0.07
+output = 0.14
+cache_read = 0.014
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index 68d7d650159..d35b93066de 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,11 +6,12 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.32
-output = 3.2
+input = 0.6
+output = 3.6
+cache_read = 0.12
 
 [limit]
-output = 81_920
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy4-preview.toml b/providers/openrouter/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..9ce403c52ff
--- /dev/null
+++ b/providers/openrouter/models/tencent/hy4-preview.toml
@@ -0,0 +1,27 @@
+name = "Hy4 preview"
+description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
+
+[limit]
+context = 1_048_576
+output = 64_000
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 24b3044efe48bb184dd1ce2279f54299f7730a84 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 28 Aug 2026 19:31:22 +0800
Subject: [PATCH 0122/2012] feat(go): add Hy4 preview

---
 providers/opencode-go/models/hy4-preview.toml | 10 ++++++++++
 1 file changed, 10 insertions(+)
 create mode 100644 providers/opencode-go/models/hy4-preview.toml

diff --git a/providers/opencode-go/models/hy4-preview.toml b/providers/opencode-go/models/hy4-preview.toml
new file mode 100644
index 00000000000..bf1bce1abc0
--- /dev/null
+++ b/providers/opencode-go/models/hy4-preview.toml
@@ -0,0 +1,10 @@
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042

From 585cd4775519027457e74282c06c30c740a23399 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 28 Aug 2026 20:02:25 +0800
Subject: [PATCH 0123/2012] feat(opencode): add Ling 3.0 Flash Fin Free

---
 providers/opencode/models/ling-3.0-flash-fin-free.toml | 10 ++++++++++
 1 file changed, 10 insertions(+)
 create mode 100644 providers/opencode/models/ling-3.0-flash-fin-free.toml

diff --git a/providers/opencode/models/ling-3.0-flash-fin-free.toml b/providers/opencode/models/ling-3.0-flash-fin-free.toml
new file mode 100644
index 00000000000..a111395b0d2
--- /dev/null
+++ b/providers/opencode/models/ling-3.0-flash-fin-free.toml
@@ -0,0 +1,10 @@
+base_model = "inclusionai/ling-3.0-flash-fin"
+name = "Ling 3.0 Flash Fin Free"
+temperature = true
+structured_output = false
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0

From a32093b993b6cc105beb77f7d56dafb9ddb21791 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 12:07:37 +0000
Subject: [PATCH 0124/2012] chore(sync): update LLM Gateway model catalog
 (#5716)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-flash.toml           | 12 ++++++++++++
 .../models/iceberg/gemini-3-flash-preview.toml  | 14 --------------
 .../models/iceberg/gemini-3.1-pro-preview.toml  | 17 -----------------
 .../models/iceberg/gemini-3.6-flash.toml        | 12 ------------
 .../models/iceberg/gemini-3.7-flash.toml        | 12 ------------
 .../models/scx-ai-gp/kimi-k2.7-code.toml        | 15 +++++++++++++++
 .../models/scx-ai-gp/kimi-k3.toml               | 14 ++++++++++++++
 7 files changed, 41 insertions(+), 55 deletions(-)
 create mode 100644 providers/llmgateway-providers/models/alibaba/qwen3.8-flash.toml
 delete mode 100644 providers/llmgateway-providers/models/iceberg/gemini-3-flash-preview.toml
 delete mode 100644 providers/llmgateway-providers/models/iceberg/gemini-3.1-pro-preview.toml
 delete mode 100644 providers/llmgateway-providers/models/iceberg/gemini-3.6-flash.toml
 delete mode 100644 providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml
 create mode 100644 providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
 create mode 100644 providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml

diff --git a/providers/llmgateway-providers/models/alibaba/qwen3.8-flash.toml b/providers/llmgateway-providers/models/alibaba/qwen3.8-flash.toml
new file mode 100644
index 00000000000..ef229ca5be6
--- /dev/null
+++ b/providers/llmgateway-providers/models/alibaba/qwen3.8-flash.toml
@@ -0,0 +1,12 @@
+base_model = "alibaba/qwen3.8-flash"
+name = "Qwen3.8 Flash (Alibaba Cloud)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
diff --git a/providers/llmgateway-providers/models/iceberg/gemini-3-flash-preview.toml b/providers/llmgateway-providers/models/iceberg/gemini-3-flash-preview.toml
deleted file mode 100644
index 4241c3129d0..00000000000
--- a/providers/llmgateway-providers/models/iceberg/gemini-3-flash-preview.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "google/gemini-3-flash-preview"
-name = "Gemini 3 Flash (Preview) (Iceberg)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 0.5
-output = 3
-cache_read = 0.05
-
-[limit]
-output = 65_535
diff --git a/providers/llmgateway-providers/models/iceberg/gemini-3.1-pro-preview.toml b/providers/llmgateway-providers/models/iceberg/gemini-3.1-pro-preview.toml
deleted file mode 100644
index 0da6fb52187..00000000000
--- a/providers/llmgateway-providers/models/iceberg/gemini-3.1-pro-preview.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "google/gemini-3.1-pro-preview"
-name = "Gemini 3.1 Pro (Preview) (Iceberg)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 2
-output = 12
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 18
-cache_read = 0.4
diff --git a/providers/llmgateway-providers/models/iceberg/gemini-3.6-flash.toml b/providers/llmgateway-providers/models/iceberg/gemini-3.6-flash.toml
deleted file mode 100644
index b7a396ebd56..00000000000
--- a/providers/llmgateway-providers/models/iceberg/gemini-3.6-flash.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "google/gemini-3.6-flash"
-name = "Gemini 3.6 Flash (Iceberg)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 0.75
-output = 3.75
-cache_read = 0.075
-cache_write = 0.08333
diff --git a/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml b/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml
deleted file mode 100644
index a6672029187..00000000000
--- a/providers/llmgateway-providers/models/iceberg/gemini-3.7-flash.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "google/gemini-3.7-flash"
-name = "Gemini 3.7 Flash (Iceberg)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 0.75
-output = 3.75
-cache_read = 0.075
-cache_write = 0.08333
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..a52f9e11855
--- /dev/null
+++ b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
@@ -0,0 +1,15 @@
+base_model = "moonshotai/kimi-k2.7-code"
+name = "Kimi K2.7 Code (SCX.ai)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.89
+output = 3.71
+cache_read = 0.18
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml
new file mode 100644
index 00000000000..665f42b2175
--- /dev/null
+++ b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (SCX.ai)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2.83
+output = 14.13
+cache_read = 0.28
+
+[limit]
+output = 1_048_576

From aa1810a8695409a2a37bcd6291f65885f9502d28 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 12:07:48 +0000
Subject: [PATCH 0125/2012] chore(sync): update Inceptron model catalog (#5721)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/inceptron/models/moonshotai/Kimi-K2.6.toml | 4 ++--
 providers/inceptron/models/zai-org/GLM-5.2.toml      | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
index 1d4fb10b7bd..a18dab49113 100644
--- a/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/inceptron/models/moonshotai/Kimi-K2.6.toml
@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.54
+input = 0.53
 output = 3.39
-cache_read = 0.15
+cache_read = 0.17
 cache_write = 0
 
 [modalities]
diff --git a/providers/inceptron/models/zai-org/GLM-5.2.toml b/providers/inceptron/models/zai-org/GLM-5.2.toml
index d2bbfd839b2..52d1dc577ca 100644
--- a/providers/inceptron/models/zai-org/GLM-5.2.toml
+++ b/providers/inceptron/models/zai-org/GLM-5.2.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 0.75
-output = 2.4
-cache_read = 0.17
+input = 0.71
+output = 2.35
+cache_read = 0.12
 cache_write = 0
 
 [limit]

From 6546b4aacf602e06e3d69d92674bddf04fe1053e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 12:07:57 +0000
Subject: [PATCH 0126/2012] chore(sync): update OpenRouter model catalog
 (#5722)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v3.2.toml  | 10 ++++++----
 .../deepseek/deepseek-v4-flash-vision-exp.toml     |  6 +++---
 .../models/deepseek/deepseek-v4-pro-0813.toml      |  6 +++---
 .../openrouter/models/google/gemma-3-27b-it.toml   |  2 +-
 .../openrouter/models/qwen/qwen3.5-122b-a10b.toml  |  8 +++++---
 .../openrouter/models/tencent/hy4-preview.toml     | 14 +-------------
 6 files changed, 19 insertions(+), 27 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v3.2.toml b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
index 5b00049c654..80cd9bc29e1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v3.2.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v3.2"
 description = "DeepSeek chat model for instruction following, coding, and analysis"
 
@@ -5,10 +7,10 @@ description = "DeepSeek chat model for instruction following, coding, and analys
 type = "toggle"
 
 [cost]
-input = 0.26
-output = 0.38
-cache_read = 0.13
+input = 0.269
+output = 0.4
+cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemma-3-27b-it.toml b/providers/openrouter/models/google/gemma-3-27b-it.toml
index c08fc52ae88..f76ae331963 100644
--- a/providers/openrouter/models/google/gemma-3-27b-it.toml
+++ b/providers/openrouter/models/google/gemma-3-27b-it.toml
@@ -17,7 +17,7 @@ output = 0.45
 cache_read = 0.04
 
 [limit]
-context = 262_144
+context = 131_072
 output = 117_964
 
 [modalities]
diff --git a/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml b/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
index 726aba0bf25..c07c101174e 100644
--- a/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,14 +1,16 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.5-122b-a10b"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.26
-output = 2.08
+input = 0.29
+output = 2.4
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy4-preview.toml b/providers/openrouter/models/tencent/hy4-preview.toml
index 9ce403c52ff..e39f8986b6d 100644
--- a/providers/openrouter/models/tencent/hy4-preview.toml
+++ b/providers/openrouter/models/tencent/hy4-preview.toml
@@ -1,13 +1,6 @@
-name = "Hy4 preview"
+base_model = "tencent/hy4-preview"
 description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,8 +13,3 @@ cache_read = 0.042
 
 [limit]
 context = 1_048_576
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 0ef87b690c0c1f2ae42d3ec18bd4ff70dd25765f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 12:08:02 +0000
Subject: [PATCH 0127/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#5719)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/kimi-k2.7-code.toml |  8 ++++----
 providers/llmgateway/models/kimi-k3.toml        |  6 +++---
 providers/llmgateway/models/qwen3.8-flash.toml  | 11 +++++++++++
 3 files changed, 18 insertions(+), 7 deletions(-)
 create mode 100644 providers/llmgateway/models/qwen3.8-flash.toml

diff --git a/providers/llmgateway/models/kimi-k2.7-code.toml b/providers/llmgateway/models/kimi-k2.7-code.toml
index 9fbaa5e5e74..082253a616a 100644
--- a/providers/llmgateway/models/kimi-k2.7-code.toml
+++ b/providers/llmgateway/models/kimi-k2.7-code.toml
@@ -8,9 +8,9 @@ field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.95
-output = 4
-cache_read = 0.19
+input = 0.89
+output = 3.71
+cache_read = 0.18
diff --git a/providers/llmgateway/models/kimi-k3.toml b/providers/llmgateway/models/kimi-k3.toml
index b4adfd00e15..5196be431f3 100644
--- a/providers/llmgateway/models/kimi-k3.toml
+++ b/providers/llmgateway/models/kimi-k3.toml
@@ -16,6 +16,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.83
+output = 14.13
+cache_read = 0.28
diff --git a/providers/llmgateway/models/qwen3.8-flash.toml b/providers/llmgateway/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..78303b8e6ba
--- /dev/null
+++ b/providers/llmgateway/models/qwen3.8-flash.toml
@@ -0,0 +1,11 @@
+base_model = "alibaba/qwen3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From 78bdefe1cf8776427a1c6702116c3139bc700ebb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 12:08:08 +0000
Subject: [PATCH 0128/2012] chore(sync): update Kilo model catalog (#5717)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v3.2.toml  |  2 +-
 .../deepseek/deepseek-v4-flash-vision-exp.toml     |  6 +++---
 providers/kilo/models/qwen/qwen3.5-122b-a10b.toml  |  2 +-
 providers/kilo/models/tencent/hy4-preview.toml     | 14 +-------------
 4 files changed, 6 insertions(+), 18 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v3.2.toml b/providers/kilo/models/deepseek/deepseek-v3.2.toml
index 1af0a0ca64c..a29137a7cc0 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2.toml
@@ -12,4 +12,4 @@ cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536
diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index e79a3173757..6f876cf5199 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml b/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
index 6e5f287283b..375dea8cc3c 100644
--- a/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
@@ -10,7 +10,7 @@ input = 0.26
 output = 2.08
 
 [limit]
-output = 235_929
+output = 81_920
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy4-preview.toml b/providers/kilo/models/tencent/hy4-preview.toml
index 191cb9a173c..664c533e3b8 100644
--- a/providers/kilo/models/tencent/hy4-preview.toml
+++ b/providers/kilo/models/tencent/hy4-preview.toml
@@ -1,13 +1,6 @@
-name = "Tencent: Hy4 preview"
+base_model = "tencent/hy4-preview"
 description = "Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total. It is designed for coding agents, complex tool-use workflows, and productivity tasks that..."
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,8 +13,3 @@ cache_read = 0.042
 
 [limit]
 context = 1_048_576
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

From d1d2d9548f3fb108d49c30d5519dae956c3395b1 Mon Sep 17 00:00:00 2001
From: heimoshuiyu <heimoshuiyu@gmail.com>
Date: Fri, 28 Aug 2026 23:11:33 +0800
Subject: [PATCH 0129/2012] feat(alibaba): add qwen3.8-flash effort levels and
 first-party entries

---
 .../alibaba-cn/models/qwen3.8-flash.toml      | 26 +++++++++++++++++++
 .../models/qwen3.8-flash.toml                 |  6 +++--
 .../models/qwen3.8-flash.toml                 |  6 +++--
 providers/alibaba/models/qwen3.8-flash.toml   | 23 ++++++++++++++++
 4 files changed, 57 insertions(+), 4 deletions(-)
 create mode 100644 providers/alibaba-cn/models/qwen3.8-flash.toml
 create mode 100644 providers/alibaba/models/qwen3.8-flash.toml

diff --git a/providers/alibaba-cn/models/qwen3.8-flash.toml b/providers/alibaba-cn/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..076549e30b2
--- /dev/null
+++ b/providers/alibaba-cn/models/qwen3.8-flash.toml
@@ -0,0 +1,26 @@
+# Sources (accessed 2026-08-28):
+# https://www.qianwenai.com/models/qwen3.8-flash
+# https://help.aliyun.com/zh/model-studio/model-pricing (Beijing: ¥0.8 input / ¥2.7
+# output per 1M tokens, flat 0<Token≤1M band, no tiers)
+# CNY→USD rate: 6.737174, 2026-08-28, source: https://open.er-api.com/v6/latest/USD
+# Cache pricing: cache read 10%, explicit cache create 125% of input
+# (https://help.aliyun.com/zh/model-studio/context-cache)
+# Toggle: enable_thinking true|false (hybrid)
+# Effort: reasoning_effort = low|medium|xhigh (not yet in the API reference; see
+# the Qwen3.8-Flash-Next release blog)
+# Budget: thinking_budget (Max Reasoning 262K)
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", max = 262_144 },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.11875
+output = 0.40073
+cache_read = 0.01187
+cache_write = 0.14844
diff --git a/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml b/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
index 7f93b2a617c..7058146174e 100644
--- a/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
+++ b/providers/alibaba-token-plan-cn/models/qwen3.8-flash.toml
@@ -5,13 +5,15 @@
 # https://docs.qwencloud.com/developer-guides/text-generation/thinking
 # https://docs.qwencloud.com/api-reference/chat/openai-chat
 # Toggle: enable_thinking true|false (hybrid)
-# Budget: thinking_budget — model page lists Max Reasoning 262K
-# reasoning_effort is documented for qwen3.8-max / qwen3.8-max-preview only, not flash.
+# Effort: reasoning_effort = low|medium|xhigh (not yet in the API reference; see
+# the Qwen3.8-Flash-Next release blog)
+# Budget: thinking_budget (Max Reasoning 262K)
 # Reasoning streams in reasoning_content.
 # Cost is Credits-subscription (no public USD per-token rate on Token Plan).
 base_model = "alibaba/qwen3.8-flash"
 reasoning_options = [
   { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
   { type = "budget_tokens", max = 262_144 },
 ]
 
diff --git a/providers/alibaba-token-plan/models/qwen3.8-flash.toml b/providers/alibaba-token-plan/models/qwen3.8-flash.toml
index 9d3107caa65..deab92706c9 100644
--- a/providers/alibaba-token-plan/models/qwen3.8-flash.toml
+++ b/providers/alibaba-token-plan/models/qwen3.8-flash.toml
@@ -4,13 +4,15 @@
 # https://docs.qwencloud.com/developer-guides/text-generation/thinking
 # https://docs.qwencloud.com/api-reference/chat/openai-chat
 # Toggle: enable_thinking true|false (hybrid)
-# Budget: thinking_budget — model page lists Max Reasoning 262K
-# reasoning_effort is documented for qwen3.8-max / qwen3.8-max-preview only, not flash.
+# Effort: reasoning_effort = low|medium|xhigh (not yet in the API reference; see
+# the Qwen3.8-Flash-Next release blog)
+# Budget: thinking_budget (Max Reasoning 262K)
 # Reasoning streams in reasoning_content.
 # Cost is Credits-subscription (no public USD per-token rate on Token Plan).
 base_model = "alibaba/qwen3.8-flash"
 reasoning_options = [
   { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
   { type = "budget_tokens", max = 262_144 },
 ]
 
diff --git a/providers/alibaba/models/qwen3.8-flash.toml b/providers/alibaba/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..3b7328b6818
--- /dev/null
+++ b/providers/alibaba/models/qwen3.8-flash.toml
@@ -0,0 +1,23 @@
+# Sources (accessed 2026-08-28):
+# https://www.qwencloud.com/models/qwen3.8-flash
+# Pricing: $0.15 input / $0.47 output / cache read $0.016 / cache create $0.2
+# per 1M tokens, flat. (Release blog quotes ~$0.16 input; model page prevails.)
+# Toggle: enable_thinking true|false (hybrid)
+# Effort: reasoning_effort = low|medium|xhigh (not yet in the API reference; see
+# the Qwen3.8-Flash-Next release blog)
+# Budget: thinking_budget (Max Reasoning 262K)
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", max = 262_144 },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From f03e0bc80988a626a5588e6109ae3fa59fe529d8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 16:25:01 +0000
Subject: [PATCH 0130/2012] chore(sync): update Eden AI model catalog (#5739)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/deepinfra/tencent/Hy3.toml | 14 ++++++++++++++
 .../edenai/models/google/lyria-3-clip-preview.toml |  9 ---------
 .../ionos/meta-llama/Llama-3.3-70B-Instruct.toml   |  4 ++--
 .../edenai/models/ionos/openai/gpt-oss-120b.toml   |  4 ++--
 providers/edenai/models/qwen/qwen3.8-flash.toml    | 11 +++++++++++
 .../models/scaleway/deepseek-v4-flash-0731.toml    |  4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml |  4 ++--
 .../models/scaleway/llama-3.3-70b-instruct.toml    |  4 ++--
 8 files changed, 35 insertions(+), 19 deletions(-)
 create mode 100644 providers/edenai/models/deepinfra/tencent/Hy3.toml
 delete mode 100644 providers/edenai/models/google/lyria-3-clip-preview.toml
 create mode 100644 providers/edenai/models/qwen/qwen3.8-flash.toml

diff --git a/providers/edenai/models/deepinfra/tencent/Hy3.toml b/providers/edenai/models/deepinfra/tencent/Hy3.toml
new file mode 100644
index 00000000000..6048f17c085
--- /dev/null
+++ b/providers/edenai/models/deepinfra/tencent/Hy3.toml
@@ -0,0 +1,14 @@
+base_model = "tencent/hy3"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.14
+output = 0.58
+cache_read = 0.035
+
+[limit]
+context = 262_144
diff --git a/providers/edenai/models/google/lyria-3-clip-preview.toml b/providers/edenai/models/google/lyria-3-clip-preview.toml
deleted file mode 100644
index 0e5aade5c80..00000000000
--- a/providers/edenai/models/google/lyria-3-clip-preview.toml
+++ /dev/null
@@ -1,9 +0,0 @@
-base_model = "google/lyria-3-clip-preview"
-structured_output = true
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 1_048_576
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 76bc9a6d3e2..8307fda46b6 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.756925
-output = 0.756925
+input = 0.756795
+output = 0.756795
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 690cc4bdc6c..384c2737624 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174675
-output = 0.756925
+input = 0.174645
+output = 0.756795
diff --git a/providers/edenai/models/qwen/qwen3.8-flash.toml b/providers/edenai/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..7b6150ebeeb
--- /dev/null
+++ b/providers/edenai/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,11 @@
+base_model = "alibaba/qwen3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 6ff97d9c80d..ca937b79460 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4658
-output = 0.9316
+input = 0.46572
+output = 0.93144
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 9896df51c69..75b0577f6a1 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174675
-output = 0.6987
+input = 0.174645
+output = 0.69858
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 16fd188db09..b10ebf79d5f 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 1.04805
-output = 1.04805
+input = 1.04787
+output = 1.04787

From 0f110ac7b67aaf5ef11113547745058179581efe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 16:25:03 +0000
Subject: [PATCH 0131/2012] chore(sync): update OpenRouter model catalog
 (#5738)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 8 +++++---
 providers/openrouter/models/openai/gpt-4.1-nano.toml      | 5 ++++-
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 1 -
 providers/openrouter/models/~z-ai/glm-latest.toml         | 4 ++--
 6 files changed, 17 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..e9b629fb28e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.0868
+output = 0.1736
+cache_read = 0.01736
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index bd9f599978c..32c1ed84218 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-pro"
 
 [interleaved]
@@ -11,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.0725
+input = 0.772038
+output = 1.544076
+cache_read = 0.064337
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-4.1-nano.toml b/providers/openrouter/models/openai/gpt-4.1-nano.toml
index fb48681a171..8db2a7fea71 100644
--- a/providers/openrouter/models/openai/gpt-4.1-nano.toml
+++ b/providers/openrouter/models/openai/gpt-4.1-nano.toml
@@ -3,7 +3,10 @@ base_model = "openai/gpt-4.1-nano"
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.025
+cache_read = 0.03
+
+[limit]
+output = 942_818
 
 [modalities]
 input = ["image", "text", "pdf"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index e777e1a5cf0..66d809ab3b9 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index ad65d1c7481..ccfe85e96f1 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = false
 
 [[reasoning_options]]
@@ -21,7 +21,7 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 4b0b4ffb58cb4175cbf47f592f4edb2e8b50f8b2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 16:25:19 +0000
Subject: [PATCH 0132/2012] chore(sync): update Kilo model catalog (#5737)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-4.1-nano.toml |  3 +++
 providers/kilo/models/tencent/hy3.toml         |  6 +++---
 providers/kilo/models/z-ai/glm-5.3-flash.toml  |  6 +++---
 providers/kilo/models/z-ai/glm-5.3.toml        |  1 -
 providers/kilo/models/~z-ai/glm-latest.toml    | 10 +++++-----
 5 files changed, 14 insertions(+), 12 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-4.1-nano.toml b/providers/kilo/models/openai/gpt-4.1-nano.toml
index c35fd0eb40b..e2c5942f03b 100644
--- a/providers/kilo/models/openai/gpt-4.1-nano.toml
+++ b/providers/kilo/models/openai/gpt-4.1-nano.toml
@@ -6,5 +6,8 @@ input = 0.1
 output = 0.4
 cache_read = 0.025
 
+[limit]
+output = 942_818
+
 [modalities]
 input = ["image", "text", "pdf"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 0ed9d5f31a9..0104ea56ade 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index 8192c07e8b6..e9d5bdad891 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3"
 description = "GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It supports text input and output with a 1M-token context window, and improves..."
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index ab46f17462b..8f563849520 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = false
 
 [[reasoning_options]]
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.2
+output = 1.2
+cache_read = 0.6
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 90342446008a3b9668277680525282cae14803e0 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Sat, 29 Aug 2026 00:31:21 +0800
Subject: [PATCH 0133/2012] fix(volcengine): drop redundant context override on
 Seed 2.0 lite/mini (#5698)

Follow-up to #5644, which merged with the review point unaddressed.

The lab entries models/bytedance-seed/seed-2.0-{lite,mini}.toml already
declare limit.context = 256_000. Plain objects deep-merge, so restating an
identical inherited value violates the override-only rule for base_model
files. Only output is a real delta (32_000 in the lab, 131_072 on this
host), so keep output and let context inherit.
---
 providers/volcengine/models/doubao-seed-2-0-lite-260428.toml | 1 -
 providers/volcengine/models/doubao-seed-2-0-mini-260428.toml | 1 -
 2 files changed, 2 deletions(-)

diff --git a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
index f75f626cf34..8ff1d78be26 100644
--- a/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-lite-260428.toml
@@ -10,7 +10,6 @@
 base_model = "bytedance-seed/seed-2.0-lite"
 
 [limit]
-context = 256_000
 output = 131_072
 
 [[reasoning_options]]
diff --git a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
index dafad8fa4bf..ab6bb6b255c 100644
--- a/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
+++ b/providers/volcengine/models/doubao-seed-2-0-mini-260428.toml
@@ -10,7 +10,6 @@
 base_model = "bytedance-seed/seed-2.0-mini"
 
 [limit]
-context = 256_000
 output = 131_072
 
 [[reasoning_options]]

From ec0ec5d5bcaa645f178945827759f1f561034a82 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Sat, 29 Aug 2026 00:31:30 +0800
Subject: [PATCH 0134/2012] feat(bytedance-seed): declare structured_output on
 11 Seed models (#5702)

seed-2.0-code already declares it; the other Seed entries omit the field, so
downstream consumers that derive structured-output support from this catalog
treat them as unsupported.

Verified against POST /api/v3/chat/completions on the Volcengine Ark host, one
call per model, using response_format with a strict json_schema
(additionalProperties: false, both properties required). All 11 returned 200 and
emitted schema-conforming JSON.
---
 models/bytedance-seed/seed-1-6-flash.toml  | 1 +
 models/bytedance-seed/seed-1-6-vision.toml | 1 +
 models/bytedance-seed/seed-1-6.toml        | 1 +
 models/bytedance-seed/seed-1-8.toml        | 1 +
 models/bytedance-seed/seed-2.0-lite.toml   | 1 +
 models/bytedance-seed/seed-2.0-mini.toml   | 1 +
 models/bytedance-seed/seed-2.0-pro.toml    | 1 +
 models/bytedance-seed/seed-2.1-pro.toml    | 1 +
 models/bytedance-seed/seed-2.1-turbo.toml  | 1 +
 models/bytedance-seed/seed-character.toml  | 1 +
 models/bytedance-seed/seed-evolving.toml   | 1 +
 11 files changed, 11 insertions(+)

diff --git a/models/bytedance-seed/seed-1-6-flash.toml b/models/bytedance-seed/seed-1-6-flash.toml
index b8f14e3eb77..ab967994ed4 100644
--- a/models/bytedance-seed/seed-1-6-flash.toml
+++ b/models/bytedance-seed/seed-1-6-flash.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-1-6-vision.toml b/models/bytedance-seed/seed-1-6-vision.toml
index 9ac037c55da..6de1a6c267e 100644
--- a/models/bytedance-seed/seed-1-6-vision.toml
+++ b/models/bytedance-seed/seed-1-6-vision.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-1-6.toml b/models/bytedance-seed/seed-1-6.toml
index daf6eb42abb..898524a8f96 100644
--- a/models/bytedance-seed/seed-1-6.toml
+++ b/models/bytedance-seed/seed-1-6.toml
@@ -10,6 +10,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-1-8.toml b/models/bytedance-seed/seed-1-8.toml
index 1bb3255e4c3..d9d6f93c6c9 100644
--- a/models/bytedance-seed/seed-1-8.toml
+++ b/models/bytedance-seed/seed-1-8.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-2.0-lite.toml b/models/bytedance-seed/seed-2.0-lite.toml
index 4722fa87e28..871c9e83e4b 100644
--- a/models/bytedance-seed/seed-2.0-lite.toml
+++ b/models/bytedance-seed/seed-2.0-lite.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-2.0-mini.toml b/models/bytedance-seed/seed-2.0-mini.toml
index 9f796068539..5b9190b334b 100644
--- a/models/bytedance-seed/seed-2.0-mini.toml
+++ b/models/bytedance-seed/seed-2.0-mini.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-2.0-pro.toml b/models/bytedance-seed/seed-2.0-pro.toml
index a5fd4573f3a..aaf0107c518 100644
--- a/models/bytedance-seed/seed-2.0-pro.toml
+++ b/models/bytedance-seed/seed-2.0-pro.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-2.1-pro.toml b/models/bytedance-seed/seed-2.1-pro.toml
index 3043a721dd4..0f7dea4d8b9 100644
--- a/models/bytedance-seed/seed-2.1-pro.toml
+++ b/models/bytedance-seed/seed-2.1-pro.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-2.1-turbo.toml b/models/bytedance-seed/seed-2.1-turbo.toml
index 587fb4fa0e5..dc66a62af75 100644
--- a/models/bytedance-seed/seed-2.1-turbo.toml
+++ b/models/bytedance-seed/seed-2.1-turbo.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-character.toml b/models/bytedance-seed/seed-character.toml
index 8b67258fcc4..816d06af8d2 100644
--- a/models/bytedance-seed/seed-character.toml
+++ b/models/bytedance-seed/seed-character.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]
diff --git a/models/bytedance-seed/seed-evolving.toml b/models/bytedance-seed/seed-evolving.toml
index f824eeab863..f961aea95c0 100644
--- a/models/bytedance-seed/seed-evolving.toml
+++ b/models/bytedance-seed/seed-evolving.toml
@@ -10,6 +10,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [limit]

From dabd176a00cce65f30e036de552f9753e5ae31de Mon Sep 17 00:00:00 2001
From: Daniele Scasciafratte <mte90net@gmail.com>
Date: Fri, 28 Aug 2026 18:31:38 +0200
Subject: [PATCH 0135/2012] fix(regolo): model updated (#5703)

* fix(regolo): model updated

* fix(regolo): model updated
---
 .../regolo-ai/models/{qwen3.6-27b.toml => qwen3.8-27b.toml} | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)
 rename providers/regolo-ai/models/{qwen3.6-27b.toml => qwen3.8-27b.toml} (70%)

diff --git a/providers/regolo-ai/models/qwen3.6-27b.toml b/providers/regolo-ai/models/qwen3.8-27b.toml
similarity index 70%
rename from providers/regolo-ai/models/qwen3.6-27b.toml
rename to providers/regolo-ai/models/qwen3.8-27b.toml
index e10771de9fe..b2797559ce5 100644
--- a/providers/regolo-ai/models/qwen3.6-27b.toml
+++ b/providers/regolo-ai/models/qwen3.8-27b.toml
@@ -1,6 +1,6 @@
 # Prices listed in EUR on Regolo; converted to USD at 1 EUR = 1.1537 USD (2026-08-05)
 # Toggle: thinking = true|false
-base_model = "alibaba/qwen3.6-27b"
+base_model = "alibaba/qwen3.8-27b"
 
 reasoning_options = [{ type = "toggle" }]
 
@@ -11,7 +11,3 @@ output = 2.42
 [limit]
 context = 120_000
 output = 120_000
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]

From 9b6e8aa4176df871de80f11e613579410c6b4562 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:33:33 -0500
Subject: [PATCH 0136/2012] chore(sync): update NanoGPT model catalog (#5720)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/TEE/gemma-3-27b-it.toml   | 24 -----------------
 .../nano-gpt/models/TEE/glm-5.3-flash.toml    | 18 +++++++++++++
 .../models/TEE/qwen3.5-122b-a10b.toml         | 20 --------------
 .../models/abliterated-model-large.toml       | 27 +++++++++++++++++++
 .../nano-gpt/models/abliterated-model.toml    | 27 +++++++++++++++++++
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 .../google/gemma-4-26b-a4b-it-chimerax.toml   | 25 +++++++++++++++++
 .../google/gemma-4-26b-a4b-it-darksoul.toml   | 25 +++++++++++++++++
 .../google/gemma-4-26b-a4b-it-luminous.toml   | 24 +++++++++++++++++
 .../google/gemma-4-26b-a4b-it-moonlight.toml  | 25 +++++++++++++++++
 .../google/gemma-4-26b-a4b-it-musica.toml     | 25 +++++++++++++++++
 .../gemma-4-26b-a4b-it-opusdistill.toml       | 25 +++++++++++++++++
 .../gemma-4-26b-a4b-it-shadowsiren.toml       | 25 +++++++++++++++++
 .../google/gemma-4-31b-it-darkidol.toml       | 25 +++++++++++++++++
 .../models/google/gemma-4-31b-it-fabled.toml  | 25 +++++++++++++++++
 .../models/google/gemma-4-31b-it-garnet.toml  | 25 +++++++++++++++++
 .../google/gemma-4-31b-it-gembrain.toml       | 25 +++++++++++++++++
 .../google/gemma-4-31b-it-isometry.toml       | 25 +++++++++++++++++
 .../google/gemma-4-31b-it-novelist.toml       | 25 +++++++++++++++++
 .../models/openai/gpt-5.6-luna-pro.toml       |  8 +++---
 .../nano-gpt/models/openai/gpt-5.6-luna.toml  |  8 +++---
 .../models/openai/gpt-5.6-terra-pro.toml      |  8 +++---
 .../nano-gpt/models/openai/gpt-5.6-terra.toml |  8 +++---
 .../models/qwen/qwen3.8-27b-fable.toml        | 25 +++++++++++++++++
 .../nano-gpt/models/tencent/hy4-preview.toml  | 16 +++++++++++
 25 files changed, 456 insertions(+), 63 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/gemma-3-27b-it.toml
 create mode 100644 providers/nano-gpt/models/TEE/glm-5.3-flash.toml
 delete mode 100644 providers/nano-gpt/models/TEE/qwen3.5-122b-a10b.toml
 create mode 100644 providers/nano-gpt/models/abliterated-model-large.toml
 create mode 100644 providers/nano-gpt/models/abliterated-model.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
 create mode 100644 providers/nano-gpt/models/tencent/hy4-preview.toml

diff --git a/providers/nano-gpt/models/TEE/gemma-3-27b-it.toml b/providers/nano-gpt/models/TEE/gemma-3-27b-it.toml
deleted file mode 100644
index 548643348ac..00000000000
--- a/providers/nano-gpt/models/TEE/gemma-3-27b-it.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Gemma 3 27B TEE"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2024-01-01"
-last_updated = "2025-03-10"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.2
-output = 0.8
-cache_read = 0.1
-
-[limit]
-context = 131_072
-input = 131_072
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/TEE/glm-5.3-flash.toml b/providers/nano-gpt/models/TEE/glm-5.3-flash.toml
new file mode 100644
index 00000000000..9a90c2b0c6b
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/glm-5.3-flash.toml
@@ -0,0 +1,18 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash TEE"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+input = 1_048_576
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/nano-gpt/models/TEE/qwen3.5-122b-a10b.toml b/providers/nano-gpt/models/TEE/qwen3.5-122b-a10b.toml
deleted file mode 100644
index 62be236f5b9..00000000000
--- a/providers/nano-gpt/models/TEE/qwen3.5-122b-a10b.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-# Not included in subscription
-base_model = "alibaba/qwen3.5-122b-a10b"
-name = "Qwen3.5 122B A10B TEE"
-attachment = false
-structured_output = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.46
-output = 3.68
-cache_read = 0.23
-
-[limit]
-input = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/abliterated-model-large.toml b/providers/nano-gpt/models/abliterated-model-large.toml
new file mode 100644
index 00000000000..4e36a511c70
--- /dev/null
+++ b/providers/nano-gpt/models/abliterated-model-large.toml
@@ -0,0 +1,27 @@
+name = "Abliterated Model Large"
+description = "Abliteration.ai's large text reasoning model is derived from GLM-5.2 and supports structured output, automatic prompt caching, and a one-million-token context window."
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 5
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 999_990
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/abliterated-model.toml b/providers/nano-gpt/models/abliterated-model.toml
new file mode 100644
index 00000000000..ac86f6f5738
--- /dev/null
+++ b/providers/nano-gpt/models/abliterated-model.toml
@@ -0,0 +1,27 @@
+name = "Abliterated Model"
+description = "Abliteration.ai's multimodal reasoning model supports text and image input, structured output, automatic prompt caching, and a 262K-token context window."
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = true
+reasoning = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 3
+output = 3
+cache_read = 0.3
+
+[limit]
+context = 262_144
+input = 262_144
+output = 262_134
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml
new file mode 100644
index 00000000000..f2c876ddf7b
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml
@@ -0,0 +1,25 @@
+name = "Chimera X"
+description = "Chimera X is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml
new file mode 100644
index 00000000000..cf1dbd3b699
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml
@@ -0,0 +1,25 @@
+name = "Dark Soul"
+description = "Dark Soul is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml
new file mode 100644
index 00000000000..d1eeb0fe13d
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml
@@ -0,0 +1,24 @@
+name = "Luminous Mirror"
+description = "Luminous Mirror is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml
new file mode 100644
index 00000000000..41e520abf2e
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml
@@ -0,0 +1,25 @@
+name = "Moonlight Dusk"
+description = "Moonlight Dusk is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml
new file mode 100644
index 00000000000..5bf3710aaa4
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml
@@ -0,0 +1,25 @@
+name = "Musica"
+description = "Musica is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml
new file mode 100644
index 00000000000..fc4e0eac717
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml
@@ -0,0 +1,25 @@
+name = "Opus Distill"
+description = "Opus Distill is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml
new file mode 100644
index 00000000000..fde75187d86
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml
@@ -0,0 +1,25 @@
+name = "Shadow Siren"
+description = "Shadow Siren is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml
new file mode 100644
index 00000000000..671290b4794
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml
@@ -0,0 +1,25 @@
+name = "DarkIdol"
+description = "DarkIdol is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml
new file mode 100644
index 00000000000..d2b178d01cd
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml
@@ -0,0 +1,25 @@
+name = "Fabled"
+description = "Fabled is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml
new file mode 100644
index 00000000000..cd2a2ade94d
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml
@@ -0,0 +1,25 @@
+name = "Garnet"
+description = "Garnet is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml
new file mode 100644
index 00000000000..dbc4f8512ad
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml
@@ -0,0 +1,25 @@
+name = "Gembrain"
+description = "Gembrain is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml
new file mode 100644
index 00000000000..2c848317d24
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml
@@ -0,0 +1,25 @@
+name = "Isometry"
+description = "Isometry is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml
new file mode 100644
index 00000000000..e2c70ac2f3b
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml
@@ -0,0 +1,25 @@
+name = "Novelist"
+description = "Novelist is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-5.6-luna-pro.toml b/providers/nano-gpt/models/openai/gpt-5.6-luna-pro.toml
index f344b44756f..d119f3ac6af 100644
--- a/providers/nano-gpt/models/openai/gpt-5.6-luna-pro.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.6-luna-pro.toml
@@ -6,10 +6,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.1
-output = 0.6
-cache_read = 0.01
-cache_write = 0.125
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
 
 [limit]
 input = 1_050_000
diff --git a/providers/nano-gpt/models/openai/gpt-5.6-luna.toml b/providers/nano-gpt/models/openai/gpt-5.6-luna.toml
index af5dd2bd12e..614549cd96a 100644
--- a/providers/nano-gpt/models/openai/gpt-5.6-luna.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.6-luna.toml
@@ -6,10 +6,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.1
-output = 0.6
-cache_read = 0.01
-cache_write = 0.125
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
 
 [limit]
 input = 1_050_000
diff --git a/providers/nano-gpt/models/openai/gpt-5.6-terra-pro.toml b/providers/nano-gpt/models/openai/gpt-5.6-terra-pro.toml
index 94be7d6ab03..3f9c760c1bd 100644
--- a/providers/nano-gpt/models/openai/gpt-5.6-terra-pro.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.6-terra-pro.toml
@@ -6,10 +6,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 1
-output = 6
-cache_read = 0.1
-cache_write = 1.25
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
 
 [limit]
 input = 1_050_000
diff --git a/providers/nano-gpt/models/openai/gpt-5.6-terra.toml b/providers/nano-gpt/models/openai/gpt-5.6-terra.toml
index 3b7fabce4b4..e08edd7cf47 100644
--- a/providers/nano-gpt/models/openai/gpt-5.6-terra.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.6-terra.toml
@@ -6,10 +6,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 1
-output = 6
-cache_read = 0.1
-cache_write = 1.25
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
 
 [limit]
 input = 1_050_000
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
new file mode 100644
index 00000000000..8a877ac220d
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.8 27B Fable"
+description = "Qwen 3.8 27B Fable is an open-weight multimodal creative finetune for expressive dialogue, long-form storytelling, character work, and roleplay."
+family = "qwen"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+tool_call = false
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.2
+output = 1.4
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/tencent/hy4-preview.toml b/providers/nano-gpt/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..349cc439215
--- /dev/null
+++ b/providers/nano-gpt/models/tencent/hy4-preview.toml
@@ -0,0 +1,16 @@
+base_model = "tencent/hy4-preview"
+name = "Tencent Hy4 Preview"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
+
+[limit]
+context = 1_048_576
+input = 1_048_576

From 46c46fe13b13626bd0f182b99bbc838e9c48917f Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:34:01 -0500
Subject: [PATCH 0137/2012] fix: bug(huggingface): zai-org/GLM-5.3-Flash
 missing vision modality (#5713)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/huggingface/models/zai-org/GLM-5.3-Flash.toml | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/providers/huggingface/models/zai-org/GLM-5.3-Flash.toml b/providers/huggingface/models/zai-org/GLM-5.3-Flash.toml
index b37f2da2873..c89651d6d77 100644
--- a/providers/huggingface/models/zai-org/GLM-5.3-Flash.toml
+++ b/providers/huggingface/models/zai-org/GLM-5.3-Flash.toml
@@ -1,9 +1,10 @@
 # Effort: reasoning_effort = low|high|max; thinking cannot be disabled.
+# Native multimodal (image input); HF Inference Providers chat-completion supports image_url.
 # https://docs.z.ai/guides/llm/glm-5.3-flash
+# https://huggingface.co/zai-org/GLM-5.3-Flash
 # https://huggingface.co/docs/inference-providers/tasks/chat-completion
 base_model = "zhipuai/glm-5.3-flash"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-attachment = false
 open_weights = true
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
@@ -15,4 +16,4 @@ output = 0.5
 context = 1_048_576
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 78c4edc8f628d198128740494c07b0076480a783 Mon Sep 17 00:00:00 2001
From: Jaber Jaber <103749727+jaberjaber23@users.noreply.github.com>
Date: Fri, 28 Aug 2026 19:34:09 +0300
Subject: [PATCH 0138/2012] fix(runinfra): correct a 2.4T claim about what the
 caller actually reads (#5723)

The entry said a caller sending reasoning_effort "none" learns this model has no
off switch. Re-measured on the buyer path 2026-08-27: they do not. The gateway
authors every rejection sentence a buyer reads and answers an unrecognised
refusal generically, so the caller learns the request was rejected and no more.
The claim was true when written and stopped being true when the gateway replaced
upstream text forwarding, which is exactly the kind of quiet drift worth
recording rather than silently overwriting.

What is still true is unchanged and restated: "none" is refused with 400 rather
than folded, so the request is never silently served and billed at the most
expensive setting.
---
 .../models/Inferact/Qwen3.8-2.4T-A95B-NVFP4.toml   | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/providers/runinfra/models/Inferact/Qwen3.8-2.4T-A95B-NVFP4.toml b/providers/runinfra/models/Inferact/Qwen3.8-2.4T-A95B-NVFP4.toml
index 775a2d7d4d0..5e3f2f67645 100644
--- a/providers/runinfra/models/Inferact/Qwen3.8-2.4T-A95B-NVFP4.toml
+++ b/providers/runinfra/models/Inferact/Qwen3.8-2.4T-A95B-NVFP4.toml
@@ -8,11 +8,15 @@
 # It is not published as a separate rung because it is not a distinct behaviour, only the
 # OpenAI ladder's word for the maximum, and this serve's maximum is "xhigh". Callers that
 # send "high" are served rather than refused.
-# Reasoning cannot be turned off: reasoning_effort = "none" is rejected with HTTP 400
-# "Disabling thinking is not supported." That refusal is deliberate and is NOT folded, so a
-# caller asking to switch reasoning off learns the model has no off switch instead of being
-# silently charged the most expensive setting. enable_thinking / chat_template_kwargs /
-# thinking_budget are accepted but have no effect.
+# Reasoning cannot be turned off: reasoning_effort = "none" is rejected with HTTP 400, and the
+# refusal is deliberate rather than folded, so such a request is never silently served and
+# billed at the most expensive setting.
+# ONE CLAIM THAT SAT HERE WAS CORRECTED 2026-08-27 after re-measuring the buyer path. It used
+# to say the caller learns this model has no off switch. That is not what happens. The gateway
+# authors every rejection sentence a buyer reads and answers an unrecognised refusal with a
+# generic one, so a caller sending "none" learns the request was rejected and no more. The
+# reason is recorded because the claim was true when written and quietly stopped being true.
+# enable_thinking / chat_template_kwargs / thinking_budget are accepted but have no effect.
 # cache_read is the cached input price. JSON mode went live on this host 2026-08-17,
 # verified on the public path: 3 of 3 json_object responses parsed and 3 of 3 strict
 # json_schema responses validated, at 0.35x baseline latency, so structured_output now

From 32aa3c49f0b0a150dcea2cf9a9ba83356d781356 Mon Sep 17 00:00:00 2001
From: Tejush <tejush1998@gmail.com>
Date: Fri, 28 Aug 2026 22:04:39 +0530
Subject: [PATCH 0139/2012] sync crof providers (#5726)

Co-authored-by: tejush <mac@MacBook-Air.local>
---
 .../crof/models/deepseek-v4-flash-0731.toml   |  4 ++--
 .../crof/models/deepseek-v4-pro-0813.toml     | 20 +++++++++++++++++++
 .../models/deepseek-v4-pro-lightning.toml     | 17 ----------------
 providers/crof/models/glm-5.3-flash.toml      | 17 ++++++++++++++++
 providers/crof/models/qwen3.8-27b.toml        |  6 +++---
 5 files changed, 42 insertions(+), 22 deletions(-)
 create mode 100644 providers/crof/models/deepseek-v4-pro-0813.toml
 delete mode 100644 providers/crof/models/deepseek-v4-pro-lightning.toml
 create mode 100644 providers/crof/models/glm-5.3-flash.toml

diff --git a/providers/crof/models/deepseek-v4-flash-0731.toml b/providers/crof/models/deepseek-v4-flash-0731.toml
index 7fedc21cbff..84aaa9ea9f7 100644
--- a/providers/crof/models/deepseek-v4-flash-0731.toml
+++ b/providers/crof/models/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high
 field = "reasoning_content"
 
 [cost]
-input = 0.12
-output = 0.21
+input = 0.08
+output = 0.1
 cache_read = 0.003
 
 [limit]
diff --git a/providers/crof/models/deepseek-v4-pro-0813.toml b/providers/crof/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..e29edc7fa84
--- /dev/null
+++ b/providers/crof/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,20 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro (0813)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.35
+output = 0.8
+cache_read = 0.01
+
+[limit]
+output = 131_072
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
diff --git a/providers/crof/models/deepseek-v4-pro-lightning.toml b/providers/crof/models/deepseek-v4-pro-lightning.toml
deleted file mode 100644
index 9e76144e88c..00000000000
--- a/providers/crof/models/deepseek-v4-pro-lightning.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "deepseek/deepseek-v4-pro"
-name = "DeepSeek V4 Pro Lightning"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.8
-output = 1.6
-cache_read = 0.02
-
-[limit]
-output = 131_072
-
-[provider]
-npm = "@ai-sdk/openai-compatible"
diff --git a/providers/crof/models/glm-5.3-flash.toml b/providers/crof/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..21b559518dc
--- /dev/null
+++ b/providers/crof/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3-Flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.07
+output = 0.22
+cache_read = 0.01
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
diff --git a/providers/crof/models/qwen3.8-27b.toml b/providers/crof/models/qwen3.8-27b.toml
index c2633fa54b0..4c6a544d27c 100644
--- a/providers/crof/models/qwen3.8-27b.toml
+++ b/providers/crof/models/qwen3.8-27b.toml
@@ -2,9 +2,9 @@ base_model = "alibaba/qwen3.8-27b"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 
 [cost]
-input = 0.25
-output = 2.1
-cache_read = 0.06
+input = 0.2
+output = 1.5
+cache_read = 0.03
 
 [limit]
 output = 262_144

From 0192737626c161e7bebc953e0069f17d2f480d6d Mon Sep 17 00:00:00 2001
From: MicroHEROX <140800253+MicroHEROX@users.noreply.github.com>
Date: Sat, 29 Aug 2026 00:35:01 +0800
Subject: [PATCH 0140/2012] feat: add Qwen3.8-Flash-Next to AMD provider
 (#5731)

* test write access

* feat: add Qwen3.8-Flash-Next to AMD provider
---
 models/alibaba/qwen3.8-flash-next.toml       | 36 ++++++++++++++++++++
 providers/amd/models/Qwen3.8-Flash-Next.toml | 25 ++++++++++++++
 2 files changed, 61 insertions(+)
 create mode 100644 models/alibaba/qwen3.8-flash-next.toml
 create mode 100644 providers/amd/models/Qwen3.8-Flash-Next.toml

diff --git a/models/alibaba/qwen3.8-flash-next.toml b/models/alibaba/qwen3.8-flash-next.toml
new file mode 100644
index 00000000000..f3158f07769
--- /dev/null
+++ b/models/alibaba/qwen3.8-flash-next.toml
@@ -0,0 +1,36 @@
+# Sources (accessed 2026-08-28):
+# https://huggingface.co/Qwen/Qwen3.8-Flash-Next
+# https://huggingface.co/api/models/Qwen/Qwen3.8-Flash-Next
+# https://qwen.ai/blog?id=qwen3.8-flash-next
+# Hub lastModified 2026-08-27T05:03:36Z is the open-weight drop (Do not use
+# Hub createdAt, staged countdown page).
+# Experimental preview of the Qwen4 architecture (Qwen4Exp): hybrid
+# Gated DeltaNet + Qwen Sparse Attention, 512 experts (10 routed + 1 shared),
+# 125B total with 6B active plus 51B n-gram embedding and 4B MTP.
+# Thinking always on: reasoning_effort low|medium|xhigh (default xhigh).
+# Native context 262K, extensible up to 1M tokens.
+
+name = "Qwen3.8 Flash Next"
+description = "Open-weight experimental preview of the Qwen4 architecture: hybrid-attention MoE (125B total, 6B active) with vision encoder for coding, agent tasks, and image and video understanding"
+family = "qwen"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+license = "qwen-community-1.0"
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/Qwen/Qwen3.8-Flash-Next"
diff --git a/providers/amd/models/Qwen3.8-Flash-Next.toml b/providers/amd/models/Qwen3.8-Flash-Next.toml
new file mode 100644
index 00000000000..290a819bfdd
--- /dev/null
+++ b/providers/amd/models/Qwen3.8-Flash-Next.toml
@@ -0,0 +1,25 @@
+# Sources (accessed 2026-08-28):
+# https://developer.amd.com.cn/radeon/tokenfactory
+# Pricing from the Token Factory models API: input $0.15, output $0.47,
+# cache read $0.016 per MTok.
+# AMD hosts a text-only vllm-router deployment (input/output modalities are
+# text only), so the multimodal base entry is overridden below.
+# Reasoning control verified live against POST /v1/chat/completions:
+# reasoning_effort is a validated enum low|medium|xhigh with xhigh the default
+# (values outside the enum return 400; no none, no toggle). No
+# reasoning_content is ever returned (stream + non-stream), so there is no
+# toggle and no interleaved channel on this host.
+base_model = "alibaba/qwen3.8-flash-next"
+attachment = false
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 903b835b0d3f950b94cb3fb3a2f631ee108180b5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:39:26 -0500
Subject: [PATCH 0141/2012] chore(sync): update Cortecs model catalog (#5718)

* chore(sync): update Cortecs model catalog

* fix(cortecs): correct flash model metadata and reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/cortecs/models/glm-5.3-flash.toml   | 23 +++++++++++++++++++
 .../cortecs/models/qwen3.8-flash-next.toml    | 22 ++++++++++++++++++
 2 files changed, 45 insertions(+)
 create mode 100644 providers/cortecs/models/glm-5.3-flash.toml
 create mode 100644 providers/cortecs/models/qwen3.8-flash-next.toml

diff --git a/providers/cortecs/models/glm-5.3-flash.toml b/providers/cortecs/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..6836aac9380
--- /dev/null
+++ b/providers/cortecs/models/glm-5.3-flash.toml
@@ -0,0 +1,23 @@
+# Sources (accessed 2026-08-28):
+# https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash
+# https://docs.cortecs.ai/usage/reasoning
+# Effort: reasoning_effort = low|high|max, matching the lab and relay peers.
+# Thinking is always enabled; no toggle or reasoning-token budget.
+# Output limit: https://api.cortecs.ai/v1/models (max_output_tokens, not context_size).
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.201
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 1_048_576
+output = 64_000
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/cortecs/models/qwen3.8-flash-next.toml b/providers/cortecs/models/qwen3.8-flash-next.toml
new file mode 100644
index 00000000000..37bebcf4aa6
--- /dev/null
+++ b/providers/cortecs/models/qwen3.8-flash-next.toml
@@ -0,0 +1,22 @@
+# Sources (accessed 2026-08-28):
+# https://huggingface.co/Qwen/Qwen3.8-Flash-Next
+# https://docs.cortecs.ai/usage/reasoning
+# Effort: reasoning_effort = low|medium|xhigh, matching the lab and Cortecs Qwen3.8 peers.
+# No lab-only toggle or thinking budget claimed for this relay route.
+# Output limit: https://api.cortecs.ai/v1/models (max_output_tokens, not context_size).
+base_model = "alibaba/qwen3.8-flash-next"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.201
+output = 0.5
+cache_read = 0.05
+
+[limit]
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From 8dceb2091fdba90284fed539728cc73dda121bee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:39:52 -0500
Subject: [PATCH 0142/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5736)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add Hy4 preview reasoning effort options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/deepseek/deepseek-v4-pro.toml  | 10 +++-------
 providers/vercel/models/tencent/hy4-preview.toml | 16 ++++++++++++++++
 2 files changed, 19 insertions(+), 7 deletions(-)
 create mode 100644 providers/vercel/models/tencent/hy4-preview.toml

diff --git a/providers/vercel/models/deepseek/deepseek-v4-pro.toml b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
index 917089fcfb7..d1c7fbe6694 100644
--- a/providers/vercel/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4-pro.toml
@@ -8,10 +8,6 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.74
-output = 3.48
-cache_read = 0.14
-
-[limit]
-context = 1_048_600
-output = 1_048_600
+input = 0.66
+output = 1.98
+cache_read = 0.022
diff --git a/providers/vercel/models/tencent/hy4-preview.toml b/providers/vercel/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..a33145904d7
--- /dev/null
+++ b/providers/vercel/models/tencent/hy4-preview.toml
@@ -0,0 +1,16 @@
+# Effort: reasoning.effort = none|high; none disables reasoning (no separate toggle).
+# Vercel routes this model to Tencent; use Tencent's native effort levels.
+# https://vercel.com/ai-gateway/models/hy4-preview
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# https://cloud.tencent.com/document/product/1823/131208
+base_model = "tencent/hy4-preview"
+name = "Tencent Hy4 Preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042

From bd428cd1dc443ffea28ff7ffc5b231ac52a89817 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:40:51 -0500
Subject: [PATCH 0143/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#5733)

* chore(sync): update Cloudflare Workers AI model catalog

* fix(cloudflare): add GLM-5.3 reasoning controls

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3.toml           | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
new file mode 100644
index 00000000000..a4483d01942
--- /dev/null
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
@@ -0,0 +1,20 @@
+# Cloudflare model input schema verified 2026-08-28 for @cf/zai-org/glm-5.3.
+# https://developers.cloudflare.com/api/resources/ai/subresources/models/subresources/schema/methods/get/
+# Toggle: chat_template_kwargs.enable_thinking = true|false (default true).
+# Effort: reasoning_effort = low|medium|high; Cloudflare's declared controls differ from Z.ai's API.
+base_model = "zhipuai/glm-5.3"
+name = "Glm 5.3"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_310_720
+output = 1_310_720

From 1b008fb61982747a6b85e853181420e231f4ead4 Mon Sep 17 00:00:00 2001
From: MickeyXin <49945543+fx247562340@users.noreply.github.com>
Date: Sat, 29 Aug 2026 00:42:02 +0800
Subject: [PATCH 0144/2012] feat(provider): add Vancine OpenAI-compatible
 models (#5711)

Co-authored-by: fx247562340 <fx247562340@users.noreply.github.com>
---
 providers/vancine/logo.svg                    |  6 ++++
 providers/vancine/models/MiniMax-M3.toml      | 22 +++++++++++++
 .../vancine/models/deepseek-v4-flash.toml     | 27 ++++++++++++++++
 providers/vancine/models/deepseek-v4-pro.toml | 27 ++++++++++++++++
 providers/vancine/models/glm-5.3-flash.toml   | 16 ++++++++++
 providers/vancine/models/glm-5.3.toml         | 17 ++++++++++
 providers/vancine/models/kimi-k3.toml         | 24 ++++++++++++++
 providers/vancine/models/qwen3.8-flash.toml   | 25 +++++++++++++++
 providers/vancine/models/qwen3.8-max.toml     | 31 +++++++++++++++++++
 providers/vancine/provider.toml               |  5 +++
 10 files changed, 200 insertions(+)
 create mode 100644 providers/vancine/logo.svg
 create mode 100644 providers/vancine/models/MiniMax-M3.toml
 create mode 100644 providers/vancine/models/deepseek-v4-flash.toml
 create mode 100644 providers/vancine/models/deepseek-v4-pro.toml
 create mode 100644 providers/vancine/models/glm-5.3-flash.toml
 create mode 100644 providers/vancine/models/glm-5.3.toml
 create mode 100644 providers/vancine/models/kimi-k3.toml
 create mode 100644 providers/vancine/models/qwen3.8-flash.toml
 create mode 100644 providers/vancine/models/qwen3.8-max.toml
 create mode 100644 providers/vancine/provider.toml

diff --git a/providers/vancine/logo.svg b/providers/vancine/logo.svg
new file mode 100644
index 00000000000..965979730de
--- /dev/null
+++ b/providers/vancine/logo.svg
@@ -0,0 +1,6 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <!-- Outer ring -->
+  <circle cx="12" cy="12" r="9.5"/>
+  <!-- Inscribed infinity (lemniscate) sharing the same vertical axis -->
+  <path d="M7 9.5 C9 6, 11 6, 12 9 C13 12, 15 15, 17 12.5 C18.5 10.5, 16.5 8, 14.5 9.5 C13 10.7, 11.5 12, 10 13.5 C8.5 15, 7 15, 7 13.5"/>
+</svg>
diff --git a/providers/vancine/models/MiniMax-M3.toml b/providers/vancine/models/MiniMax-M3.toml
new file mode 100644
index 00000000000..80fe26fadfc
--- /dev/null
+++ b/providers/vancine/models/MiniMax-M3.toml
@@ -0,0 +1,22 @@
+# Sources (accessed 2026-06-25):
+# https://platform.minimax.io/docs/api-reference/text-openai-api
+# https://docs.gmicloud.ai/inference-engine/billing/price
+# Lab reasoning controls on the OpenAI-compatible path: $.thinking.type
+# = "enabled" | "adaptive" | "disabled". reasoning_split=true returns the
+# reasoning stream in a separate reasoning_content field.
+# Interleaved field: reasoning_content.
+# Lab cost (token plan): input 0.30 / output 1.20 / cache_read 0.06.
+# Vancine public pricing (/api/pricing) is 0.24 / 0.96 / 0.048 (USD/MTok).
+# No tiered pricing is published on the public Vancine pricing endpoint.
+base_model = "minimax/MiniMax-M3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.24
+output = 0.96
+cache_read = 0.048
diff --git a/providers/vancine/models/deepseek-v4-flash.toml b/providers/vancine/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..29c63395f0b
--- /dev/null
+++ b/providers/vancine/models/deepseek-v4-flash.toml
@@ -0,0 +1,27 @@
+# Sources (accessed 2026-08-02):
+# https://api-docs.deepseek.com/guides/thinking_mode/
+# https://api-docs.deepseek.com/quick_start/pricing/
+# Reasoning HTTP format: thinking.type = "enabled" | "disabled";
+# reasoning_effort = "low" | "high" | "max". Flash maps low->low;
+# xhigh maps to high. Reasoning tokens are billed at the output rate.
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.14 / output 0.28 / reasoning 0.28 / cache_read 0.0028.
+# Vancine public pricing (/api/pricing) is 0.22 / 0.66 / n/a / 0.007 (USD/MTok).
+# Vancine ModelList exposes "deepseek-v4-flash", "deepseek-v4-flash-none",
+# "deepseek-v4-flash-max" (relay/channel/deepseek/constants.go).
+base_model = "deepseek/deepseek-v4-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
diff --git a/providers/vancine/models/deepseek-v4-pro.toml b/providers/vancine/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..d49d1b23083
--- /dev/null
+++ b/providers/vancine/models/deepseek-v4-pro.toml
@@ -0,0 +1,27 @@
+# Sources (accessed 2026-08-12):
+# https://api-docs.deepseek.com/api/create-chat-completion
+# https://api-docs.deepseek.com/guides/thinking_mode/
+# Reasoning HTTP format: thinking.type = "enabled" | "disabled";
+# reasoning_effort = "high" | "max" (Pro does not expose "low"). Reasoning
+# tokens are billed at the output rate.
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.435 / output 0.87 / reasoning 0.87 / cache_read 0.003625.
+# Vancine public pricing (/api/pricing) is 0.66 / 1.98 / n/a / 0.022 (USD/MTok).
+# Vancine ModelList exposes "deepseek-v4-pro", "deepseek-v4-pro-none",
+# "deepseek-v4-pro-max" (relay/channel/deepseek/constants.go).
+base_model = "deepseek/deepseek-v4-pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.66
+output = 1.98
+cache_read = 0.022
diff --git a/providers/vancine/models/glm-5.3-flash.toml b/providers/vancine/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..2ee2b64c754
--- /dev/null
+++ b/providers/vancine/models/glm-5.3-flash.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-08-26):
+# https://docs.bigmodel.cn/cn/guide/capabilities/thinking
+# https://docs.z.ai/guides/overview/pricing
+# Lab cost (no 50% promo): input 0.15 / output 0.50 / cache_read 0.03.
+# Vancine public pricing (/api/pricing) is 0.06 / 0.20 / 0.012 (USD/MTok).
+# Thinking is always enabled, so no toggle.
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.06
+output = 0.20
+cache_read = 0.012
diff --git a/providers/vancine/models/glm-5.3.toml b/providers/vancine/models/glm-5.3.toml
new file mode 100644
index 00000000000..20a952fb8f2
--- /dev/null
+++ b/providers/vancine/models/glm-5.3.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-08-26):
+# https://docs.bigmodel.cn/cn/guide/capabilities/thinking
+# https://docs.z.ai/guides/overview/pricing
+# Lab cost: input 1.40 / output 4.40 / cache_read 0.26 (USD/MTok).
+# Vancine public pricing (/api/pricing) is 1.12 / 3.52 / 0.208 (USD/MTok).
+# Per the audit-reasoning-options skill, the lab effort list is
+# "low" | "high" | "max"; thinking cannot be disabled, so no toggle.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.12
+output = 3.52
+cache_read = 0.208
diff --git a/providers/vancine/models/kimi-k3.toml b/providers/vancine/models/kimi-k3.toml
new file mode 100644
index 00000000000..ab8f3aa37ec
--- /dev/null
+++ b/providers/vancine/models/kimi-k3.toml
@@ -0,0 +1,24 @@
+# Sources (accessed 2026-07-17):
+# https://platform.kimi.ai/docs/pricing/chat-k3
+# https://platform.moonshot.ai/docs/api/chat
+# Lab reasoning controls: thinking.type = "enabled" | "disabled" | "adaptive";
+# in adaptive mode the depth is set by output_config.effort = "low"|"high"|"max".
+# Interleaved field: reasoning_content.
+# Lab cost: input 3.00 / output 15.00 / cache_read 0.30 (USD/MTok).
+# Vancine public pricing (/api/pricing) is 2.40 / 12.00 / 0.24 (USD/MTok).
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.40
+output = 12.00
+cache_read = 0.24
diff --git a/providers/vancine/models/qwen3.8-flash.toml b/providers/vancine/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..ceedd5283e0
--- /dev/null
+++ b/providers/vancine/models/qwen3.8-flash.toml
@@ -0,0 +1,25 @@
+# Sources (accessed 2026-08-27):
+# https://www.qwencloud.com/models/qwen3.8-flash
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Lab reasoning controls: enable_thinking true|false (toggle);
+# thinking_budget up to 262144. reasoning_effort is NOT exposed for flash
+# (qwencloud docs document it for qwen3.8-max / qwen3.8-max-preview only).
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.15 / output 0.47 / cache_read 0.016 / cache_write 0.20.
+# Vancine public pricing (/api/pricing) is 0.12 / 0.38 / 0.013 / n/a (USD/MTok).
+base_model = "alibaba/qwen3.8-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+max = 262_144
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.12
+output = 0.38
+cache_read = 0.013
diff --git a/providers/vancine/models/qwen3.8-max.toml b/providers/vancine/models/qwen3.8-max.toml
new file mode 100644
index 00000000000..bfe7534f1d7
--- /dev/null
+++ b/providers/vancine/models/qwen3.8-max.toml
@@ -0,0 +1,31 @@
+# Sources (accessed 2026-08-04):
+# https://www.qwencloud.com/models/qwen3.8-max
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# https://help.aliyun.com/zh/model-studio/model-pricing
+# Lab reasoning controls: enable_thinking true|false (toggle);
+# reasoning_effort = low|medium|xhigh (default xhigh; high is alias for xhigh);
+# thinking_budget 0..262144 (cannot be combined with reasoning_effort).
+# Interleaved field: reasoning_content.
+# Lab cost: input 2.00 / output 6.00 / cache_read 0.25 / cache_write 2.50.
+# Vancine public pricing (/api/pricing) is 1.60 / 4.80 / 0.20 / n/a (USD/MTok).
+base_model = "alibaba/qwen3.8-max"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 0
+max = 262_144
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.60
+output = 4.80
+cache_read = 0.20
diff --git a/providers/vancine/provider.toml b/providers/vancine/provider.toml
new file mode 100644
index 00000000000..98a1e99f7c5
--- /dev/null
+++ b/providers/vancine/provider.toml
@@ -0,0 +1,5 @@
+name = "Vancine"
+npm = "@ai-sdk/openai-compatible"
+env = ["VANCINE_API_KEY"]
+api = "https://vancine.com/v1"
+doc = "https://vancine.com/docs"

From 9d7504f1889b831a7a4efb08a89228050c5e00d5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:42:28 -0500
Subject: [PATCH 0145/2012] chore(sync): update Deep Infra model catalog
 (#5734)

* chore(sync): update Deep Infra model catalog

* fix(deepinfra): add GLM-5.3 reasoning effort options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../deepinfra/models/zai-org/GLM-5.3.toml      | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/deepinfra/models/zai-org/GLM-5.3.toml

diff --git a/providers/deepinfra/models/zai-org/GLM-5.3.toml b/providers/deepinfra/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..517df8be057
--- /dev/null
+++ b/providers/deepinfra/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-08-28):
+# https://docs.z.ai/guides/llm/glm-5.3
+# https://api.deepinfra.com/models/zai-org/GLM-5.3/schema/openai-chat-completions
+# Effort: reasoning_effort = low|high|max, matching the lab and relay peers.
+# GLM-5.3 always reasons; no toggle or reasoning-token budget.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576

From 232f848e16f5a34aeaf77f3c46a785648c269926 Mon Sep 17 00:00:00 2001
From: rupulsafaya <rupulsafaya@users.noreply.github.com>
Date: Fri, 28 Aug 2026 18:42:50 +0200
Subject: [PATCH 0146/2012] feat(openreason): add OpenReason provider (#5658)

---
 providers/openreason/logo.svg                 |  8 +++++++
 .../deepseek-ai/deepseek-v4-flash-0731.toml   | 19 +++++++++++++++
 .../models/moonshotai/kimi-k2.7-code.toml     | 23 +++++++++++++++++++
 .../models/openai/gpt-oss-120b.toml           | 18 +++++++++++++++
 providers/openreason/provider.toml            |  9 ++++++++
 5 files changed, 77 insertions(+)
 create mode 100644 providers/openreason/logo.svg
 create mode 100644 providers/openreason/models/deepseek-ai/deepseek-v4-flash-0731.toml
 create mode 100644 providers/openreason/models/moonshotai/kimi-k2.7-code.toml
 create mode 100644 providers/openreason/models/openai/gpt-oss-120b.toml
 create mode 100644 providers/openreason/provider.toml

diff --git a/providers/openreason/logo.svg b/providers/openreason/logo.svg
new file mode 100644
index 00000000000..7484a9dd53e
--- /dev/null
+++ b/providers/openreason/logo.svg
@@ -0,0 +1,8 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" role="img" aria-label="OpenReason">
+  <title>OpenReason</title>
+  <!-- Artwork occupies 3..21 in a 24 box, so a 3-unit margin all round. Stroke is 4 units on a
+       mid-stroke path at 5..19 with a 3-unit corner radius. The opening is the 4 units of the top
+       edge between x=16 and x=12. Colour comes from currentColor: one file serves every context. -->
+  <path d="M12 5 L8 5 A3 3 0 0 0 5 8 L5 16 A3 3 0 0 0 8 19 L16 19 A3 3 0 0 0 19 16 L19 8 A3 3 0 0 0 16 5"
+        fill="none" stroke="currentColor" stroke-width="4" stroke-linecap="butt"/>
+</svg>
diff --git a/providers/openreason/models/deepseek-ai/deepseek-v4-flash-0731.toml b/providers/openreason/models/deepseek-ai/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..919917925e7
--- /dev/null
+++ b/providers/openreason/models/deepseek-ai/deepseek-v4-flash-0731.toml
@@ -0,0 +1,19 @@
+# OpenReason routes deepseek-ai/deepseek-v4-flash-0731 through Baseten by
+# default (endpoint pin `@baseten-us`). OpenReason forwards top-level
+# `reasoning_effort` verbatim; the accepted values are
+# `none | minimal | low | medium | high (default) | xhigh | max`, matching
+# Baseten's documented surface. Reasoning output is returned in the
+# `reasoning_content` field.
+# https://openreason.app · https://docs.baseten.co/inference/model-apis/reasoning
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1371
+output = 0.2743
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/openreason/models/moonshotai/kimi-k2.7-code.toml b/providers/openreason/models/moonshotai/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..55f9a8e73b5
--- /dev/null
+++ b/providers/openreason/models/moonshotai/kimi-k2.7-code.toml
@@ -0,0 +1,23 @@
+# OpenReason routes moonshotai/kimi-k2.7-code through Baseten by default
+# (endpoint pin `@baseten-us`). The upstream accepts opt-in via
+# `chat_template_args.enable_thinking = true`; OpenReason forwards the value
+# verbatim. No off behavior, effort values, or reasoning-token budget are
+# documented. Reasoning output is returned in the `reasoning_content` field.
+# The Baseten route respects temperature, so we override the canonical
+# `false` to `true` for this host.
+# https://openreason.app · https://docs.baseten.co/inference/model-apis/reasoning
+base_model = "moonshotai/kimi-k2.7-code"
+temperature = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 1.0022
+output = 4.22
+
+[interleaved]
+field = "reasoning_content"
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/openreason/models/openai/gpt-oss-120b.toml b/providers/openreason/models/openai/gpt-oss-120b.toml
new file mode 100644
index 00000000000..fff56114067
--- /dev/null
+++ b/providers/openreason/models/openai/gpt-oss-120b.toml
@@ -0,0 +1,18 @@
+# OpenReason routes openai/gpt-oss-120b across multiple upstream inference
+# providers (Baseten, Nscale, Nebius, Scaleway). OpenReason forwards top-level
+# `reasoning_effort` verbatim; the published values are the safe intersection
+# accepted at every routed endpoint (`low | medium | high`). Baseten's own
+# endpoint additionally accepts `none | minimal | xhigh | max` — customers
+# pinning `@baseten-us` may send those values, but OpenReason does not
+# advertise them as universally accepted. Reasoning output is returned in
+# the `reasoning_content` field.
+# https://openreason.app
+base_model = "openai/gpt-oss-120b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.1055
+output = 0.422
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/openreason/provider.toml b/providers/openreason/provider.toml
new file mode 100644
index 00000000000..d7d794a0e56
--- /dev/null
+++ b/providers/openreason/provider.toml
@@ -0,0 +1,9 @@
+name = "OpenReason"
+# Sovereignty-aware OpenAI-compatible router across US and EU inference.
+# Endpoints are pinned via the `@<endpoint-id>` suffix on the model slug
+# (e.g. moonshotai/kimi-k2.7-code@baseten-us). See docs for the full list.
+# https://openreason.app
+env = ["OPENREASON_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.openreason.app/v1"
+doc = "https://openreason.app/docs"

From 1e17f994af4454f2825f663f4bb3976070642e5f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 11:43:10 -0500
Subject: [PATCH 0147/2012] chore(sync): update Baseten model catalog (#5735)

* chore(sync): update Baseten model catalog

* fix(baseten): correct GLM 5.3 reasoning metadata

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/baseten/models/zai-org/GLM-5.3.toml | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/baseten/models/zai-org/GLM-5.3.toml

diff --git a/providers/baseten/models/zai-org/GLM-5.3.toml b/providers/baseten/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..3e11707249d
--- /dev/null
+++ b/providers/baseten/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,22 @@
+# Effort: reasoning_effort = low|high|max; reasoning cannot be disabled.
+# https://docs.baseten.co/inference/model-apis/reasoning
+# Pricing: https://www.baseten.co/pricing
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.14
+
+[limit]
+context = 1_048_576
+output = 262_144

From cef528679550129bf42dc18b91dda9747e0434c7 Mon Sep 17 00:00:00 2001
From: chyroc <chyroc@qq.com>
Date: Sat, 29 Aug 2026 00:43:38 +0800
Subject: [PATCH 0148/2012] feat(volcengine-coding-plan): add Volcengine Ark
 Coding Plan (#5534)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(volcengine-coding-plan): add Volcengine Ark Coding Plan

Coding Plan is a subscription tier on its own base URL. Volcengine warns
that calling the pay-as-you-go endpoint with a Coding Plan key does not
draw down the subscription and bills separately, so it needs a distinct
provider entry rather than an alias.

Model names on this plan are the short console labels, not the dated IDs
the pay-as-you-go endpoint requires.

Change-Id: I90b655e95ea358ccc7079093bf2ce9186b2e88c7

* fix(volcengine-coding-plan): use a distinct API key env var

/api/coding/v3 does not check which plan a key belongs to — a pay-as-you-go
key reaches it and bills per token. Sharing ARK_API_KEY with the
pay-as-you-go provider lights up both at once, so a user could pick a
Coding Plan entry and silently pay pay-as-you-go rates. Matches how
alibaba-coding-plan-cn and tencent-coding-plan scope their keys.

Change-Id: I7631ea649b86639aa2ee49777b218dd0b99e35d7

* fix(volcengine-coding-plan): match reasoning_options to measured behaviour

All eight files shared one toggle + four-level declaration. Probing each
model against /api/coding/v3 turned up four distinct shapes:

- deepseek-v4-*, doubao-* (5 models): all seven levels accepted, none and
  minimal return reasoning_tokens = 0, so the toggle was redundant
- minimax-m3: none does NOT disable thinking (24-96 tokens across runs)
  but thinking.type=disabled does, so it keeps the toggle and drops none
- glm-5.3: rejects none, minimal and thinking.type=disabled outright
- kimi-k2.7-code: rejects none and thinking.type=disabled; minimal works

Also adds the wire-path comments each toggle needs per AGENTS.md.

Change-Id: I1e132a9cb514ed7a3200a1ecdd3bc3fd9fc63632

* docs(volcengine-coding-plan): spell out why these five carry no toggle

'none' and 'minimal' both zero out reasoning on these models, so the
effort list already covers the off state and thinking.type would be a
second control for the same thing.

Change-Id: Ie55717c54f0df2d56a65f7fca8316ef8d1fa6655

* docs(volcengine-coding-plan): correct the effort levels in the provider comment

The provider comment said reasoning_effort was minimal, low, medium or high,
which contradicted every model file in this directory and undersold the
endpoint. Re-tested against /api/coding/v3 on 2026-08-27: the levels are
per-model, and each file already declares the right set.

- Doubao entries accept none through max
- glm-5.3 rejects none and minimal with "not supported by this model"
- minimax-m3 accepts none but still spends reasoning tokens on it, so it is
  omitted there rather than declared

A bogus level is rejected with InvalidParameter, so these are enforced sets
rather than ignored strings.

Change-Id: I1e2c1096cea32dbff449eb74d0faebc7006c9a0c

* fix(volcengine-coding-plan): declare effective effort, not accepted strings

The review was right that these three relay entries dumped the accepted enum
instead of the levels that do something. Measured each on /api/coding/v3 with
n=3 runs per level on a hard arithmetic prompt, reading reasoning_content
length because these models report reasoning_tokens = 0:

- glm-5.3: medians 218/518/635/810/754 for low/medium/high/xhigh/max, with a
  589-1854 spread inside 'max' alone. Non-monotonic, so five tiers are not
  separable. Now ["low", "high", "max"], matching providers/zai and the
  zai/zhipuai coding plans.
- kimi-k2.7-code: minimal 707/697/917, high 823/892/755, max 712/861/690 --
  fully overlapping. Now reasoning_options = [], matching the lab
  providers/moonshotai.
- minimax-m3: minimal 1161/1033/805, high 1195/741/840, max 957/709/1088 --
  fully overlapping. Effort dropped, toggle kept, matching first-party MiniMax.

deepseek-v4-flash/pro keep their seven levels: that census is genuinely split
and the most common peer form is the same seven, so there is no baseline to
align to there.

Change-Id: Icf52878829133c067e8ca6ab6a409fdf8042db55

* docs(volcengine-coding-plan): record the n=5 glm-5.3 effort measurements

Reran at n=5 on a prompt hard enough to consume the thinking budget, which
separates the levels the earlier n=3 sample could not:

  low    0 0 0 0 0        medium 0 0 0 0 0     -> medium aliases low
  high   169 149 0 188 0
  xhigh  2840 1943 3338 4655 6279
  max    2363 4567 2962 4606 2327              -> xhigh aliases max

So ["low", "high", "max"] is now backed by measurement rather than only by
matching the peers.

Also notes that Volcengine's mapping table covers glm-5-2-260617 but not 5.3,
and that the 5-2 row disagrees with 5.3's observed behaviour (it maps
low/medium to high; 5.3 returns no reasoning on either), so the table is not
used as the source here.

Change-Id: I82f78b228abe74f85da0cd8d018f37767036e8fc

* fix(volcengine-coding-plan): set effort from measured effect, not accepted strings

Second review pass was right that the remaining relay entries still dumped the
accepted enum. Measured each on /api/coding/v3 with a prompt hard enough to
consume the thinking budget, n=4-5 per level:

- deepseek-v4-flash / -pro: none/minimal are hard zeros, but the graded values
  overlap completely -- on Flash 'high' (121-349) lands below 'low' (216-873).
  Not separable, so only the toggle is declared, and thinking.type does work
  here (enabled -> 57/79 reasoning tokens, disabled -> 0), matching the lab.
- doubao-seed-2.1-turbo / 2.0-lite: this family does grade. Turbo low 578-1148,
  high 1405-1780, max 1224-2166, with medium overlapping low and xhigh spanning
  high..max. Narrowed to none|low|high|max.
- doubao-seed-evolving: none 0, low ~1200-1600, high 2727-4065, but max
  1965-2674 lands *below* high, so max is dropped -- declaring it would point
  callers at a level that thinks less than the one beneath it. none|low|high.
- minimax-m3: restored the explicit wire comment (thinking.type = enabled|disabled
  plus endpoint) that my previous commit dropped.

Worth recording: an easy prompt hides all of this. On a one-step question every
doubao-seed-2.0-lite level lands at 57-73 tokens, which is how the full enum
looked defensible in the first place.

* fix(volcengine-coding-plan): drop max from the remaining Doubao entries

The review caught me applying my own standard inconsistently: I dropped 'max'
from doubao-seed-evolving for overlapping 'high', then kept it on 2.1-turbo and
2.0-lite whose numbers overlap just as much. The 2.0-lite comment even said "max
sits inside the high band" while still declaring max.

Re-measured 2.1-turbo at n=6 to settle it:
  high 1406-2402 (median 1928)   max 1321-2537 (median 1785)
max's median lands below high's, the ranges overlap fully, and the means are
within 2%. Not a deeper tier. On 2.0-lite max's floor (867) is likewise below
high's (896).

Both are now none|low|high, matching evolving.

Also rewrote the provider-level comment, which still described the old shapes.
It now lists the per-model sets as actually declared, with the reason each was
narrowed.

Change-Id: I7c5aa02ba99bda36f64ac992aac583993d023f95

* docs(volcengine-coding-plan): correct the claim about model naming

The comment said this endpoint takes the short console labels "not the dated IDs
the pay-as-you-go endpoint requires". The second half is wrong: /api/coding/v3
accepts both forms. Verified doubao-seed-2-1-turbo-260628,
doubao-seed-2-0-lite-260428 and glm-5-3-260930 all resolve there, alongside their
short labels.

The asymmetry runs the other way: /api/v3 accepts only the dated form and 404s on
a short label. Reworded to state that, and to say the short labels are used
because they match the Coding Plan console rather than because they are required.

Found by driving the catalog against the live endpoint the way a client reading
it would, which is the same pass that produced #5645.

Change-Id: I3a4943d415772135d71774970269410a27969193

---------

Co-authored-by: chenyunpeng.1024 <chenyunpeng.1024@bytedance.com>
---
 providers/volcengine-coding-plan/logo.svg     |  3 ++
 .../models/deepseek-v4-flash.toml             | 23 ++++++++++
 .../models/deepseek-v4-pro.toml               | 22 ++++++++++
 .../models/doubao-seed-2.0-lite.toml          | 26 ++++++++++++
 .../models/doubao-seed-2.1-turbo.toml         | 28 +++++++++++++
 .../models/doubao-seed-evolving.toml          | 26 ++++++++++++
 .../models/glm-5.3.toml                       | 30 +++++++++++++
 .../models/kimi-k2.7-code.toml                | 19 +++++++++
 .../models/minimax-m3.toml                    | 21 ++++++++++
 .../volcengine-coding-plan/provider.toml      | 42 +++++++++++++++++++
 10 files changed, 240 insertions(+)
 create mode 100644 providers/volcengine-coding-plan/logo.svg
 create mode 100644 providers/volcengine-coding-plan/models/deepseek-v4-flash.toml
 create mode 100644 providers/volcengine-coding-plan/models/deepseek-v4-pro.toml
 create mode 100644 providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml
 create mode 100644 providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml
 create mode 100644 providers/volcengine-coding-plan/models/doubao-seed-evolving.toml
 create mode 100644 providers/volcengine-coding-plan/models/glm-5.3.toml
 create mode 100644 providers/volcengine-coding-plan/models/kimi-k2.7-code.toml
 create mode 100644 providers/volcengine-coding-plan/models/minimax-m3.toml
 create mode 100644 providers/volcengine-coding-plan/provider.toml

diff --git a/providers/volcengine-coding-plan/logo.svg b/providers/volcengine-coding-plan/logo.svg
new file mode 100644
index 00000000000..3a4780bc570
--- /dev/null
+++ b/providers/volcengine-coding-plan/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M2.5 20.5L8.5 5.5C8.8 4.8 9.7 4.8 10 5.5L12 9.5L14 5.5C14.3 4.8 15.2 4.8 15.5 5.5L21.5 20.5C21.8 21.2 21.3 22 20.5 22H3.5C2.7 22 2.2 21.2 2.5 20.5ZM12 11.5L9.8 17H14.2L12 11.5Z"/>
+</svg>
diff --git a/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml b/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..706a72594b5
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml
@@ -0,0 +1,23 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Toggle: thinking.type = enabled|disabled, POST /api/coding/v3/chat/completions
+# (verified: enabled -> reasoning_tokens 57, disabled -> 0).
+# No graded effort: the same body accepts reasoning_effort, but acceptance is not
+# effect. Over n=5 runs per level on a prompt hard enough to consume the budget,
+# none and minimal were hard zeros (the off switch) while the graded values fully
+# overlapped -- low 659/873/216/487/845, medium 271/1548/556/296/1245,
+# high 121/123/132/349/262, xhigh 928/753/268/480/110, max 811/435/570/910/623.
+# 'high' sits below 'low', so the levels are not separable and only the toggle is
+# declared, matching the toggle the lab (providers/deepseek) exposes.
+base_model = "deepseek/deepseek-v4-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml b/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..68e7561f92e
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml
@@ -0,0 +1,22 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Toggle: thinking.type = enabled|disabled, POST /api/coding/v3/chat/completions
+# (verified: enabled -> reasoning_tokens 79, disabled -> 0).
+# No graded effort: the same body accepts reasoning_effort, but acceptance is not
+# effect. Over n=5 runs per level on a prompt hard enough to consume the budget,
+# none and minimal were hard zeros (the off switch) while the graded values fully
+# overlapped -- low 200/517/616/638/313, high 466/673/604/1075/415,
+# max 389/204/240/1535/688. The levels are not separable, so only the toggle is
+# declared, matching the toggle the lab (providers/deepseek) exposes.
+base_model = "deepseek/deepseek-v4-pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml b/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml
new file mode 100644
index 00000000000..36b1f4ec98b
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml
@@ -0,0 +1,26 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Effort: reasoning_effort = none|low|high, POST /api/coding/v3/chat/completions.
+# The wire also accepts minimal|medium|xhigh|max, but they are not separable.
+# Over n=4 runs per level on a prompt hard enough to consume the budget:
+#   none 0 0 0 0    low 610/556/639/583    medium 655/773/777/567
+#   high 1328/1618/896/1043               max 1369/1042/1565/867
+# medium overlaps low, and max sits inside the high band (its floor of 867 is below
+# high's 896), so max is not a deeper tier and is not declared. Only none|low|high
+# hold up. Note an easy prompt hides all of this: on a one-step question every
+# level lands at 57-73 tokens.
+# 'none' is kept rather than a toggle because thinking.type=disabled sent together
+# with a graded effort is rejected as an invalid combination here.
+base_model = "bytedance-seed/seed-2.0-lite"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml b/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml
new file mode 100644
index 00000000000..a1af2dce727
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml
@@ -0,0 +1,28 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Effort: reasoning_effort = none|low|high, POST /api/coding/v3/chat/completions.
+# The wire also accepts minimal|medium|xhigh|max, but they are not separable.
+# Over n=5 runs per level on a prompt hard enough to consume the budget:
+#   none  0 0 0 0 0            minimal 0 0 0 0 0     -> same off switch
+#   low   578-1148             medium  946-1248      -> medium overlaps low
+#   high  1405-1780            xhigh   1008-2085     -> xhigh spans high
+# 'max' was re-measured at n=6 against high on the same prompt:
+#   high 1406-2402 (median 1928)   max 1321-2537 (median 1785)
+# max's median lands *below* high's and the ranges overlap fully, so it is not a
+# deeper tier -- declaring it would advertise a level that thinks no more than
+# high. Only none|low|high hold up.
+# 'none' is kept rather than a toggle because thinking.type=disabled sent together
+# with a graded effort is rejected as an invalid combination here.
+base_model = "bytedance-seed/seed-2.1-turbo"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/doubao-seed-evolving.toml b/providers/volcengine-coding-plan/models/doubao-seed-evolving.toml
new file mode 100644
index 00000000000..ea0b56680c0
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/doubao-seed-evolving.toml
@@ -0,0 +1,26 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Effort: reasoning_effort = none|low|high, POST /api/coding/v3/chat/completions.
+# The wire also accepts minimal|medium|xhigh|max, but they are not separable.
+# Over n=4 runs per level on a prompt hard enough to consume the budget:
+#   none 0 0 0 0            low  1609/1242/1194/1195
+#   medium 1672/1597/2218/1534                        -> overlaps low
+#   high 4065/2727/2988/3863
+#   max  2665/2674/2159/1965                          -> lands *below* high
+# 'max' is not a deeper tier on this model, so declaring it would mislead callers
+# into paying for a level that thinks less than 'high'. Only none|low|high hold up.
+# 'none' is kept rather than a toggle because thinking.type=disabled sent together
+# with a graded effort is rejected as an invalid combination here.
+base_model = "bytedance-seed/seed-evolving"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/glm-5.3.toml b/providers/volcengine-coding-plan/models/glm-5.3.toml
new file mode 100644
index 00000000000..8d75263dbe7
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/glm-5.3.toml
@@ -0,0 +1,30 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Effort: POST /api/coding/v3/chat/completions accepts low|medium|high|xhigh|max,
+# but only the three effective levels are declared. This model reports
+# reasoning_tokens = 0 on every level, so the measure is reasoning_content length;
+# over n=5 runs per level on a prompt hard enough to consume the thinking budget:
+#   low    0 0 0 0 0          medium 0 0 0 0 0            -> medium aliases low
+#   high   169 149 0 188 0
+#   xhigh  2840 1943 3338 4655 6279  max 2363 4567 2962 4606 2327
+#                                                          -> xhigh aliases max
+# Three distinct tiers, matching the lab (providers/zai) and the coding-plan peers.
+# Volcengine's mapping table (82379/1449737) covers glm-5-2-260617 but not 5.3, and
+# that row disagrees with what 5.3 does here (it maps low/medium to high, whereas
+# 5.3 returns no reasoning at all on both), so these levels are set from the
+# measurements above rather than from that table.
+# No toggle: both reasoning_effort=none|minimal and thinking.type=disabled are
+# rejected with 'not supported by this model'.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml b/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..d345d44bd0e
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml
@@ -0,0 +1,19 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# reasoning_options = [] matches the lab (providers/moonshotai) and the other
+# relays: thinking is always on with no caller control. POST /api/coding/v3 does
+# accept reasoning_effort here, but acceptance is not effect -- over n=3 runs per
+# level on a hard arithmetic prompt, reasoning_content length came back
+# minimal 707/697/917, high 823/892/755, max 712/861/690, i.e. fully overlapping,
+# so no graded control is declared.
+base_model = "moonshotai/kimi-k2.7-code"
+
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/minimax-m3.toml b/providers/volcengine-coding-plan/models/minimax-m3.toml
new file mode 100644
index 00000000000..60ca6e93c0d
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/minimax-m3.toml
@@ -0,0 +1,21 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-08-26). Subscription tier: no per-token price.
+# Toggle: thinking.type = enabled|disabled, POST /api/coding/v3/chat/completions
+# Toggle only, matching first-party MiniMax and minimax-coding-plan.
+# The same body also accepts reasoning_effort, but acceptance is not effect:
+# over n=3 runs per level on a hard arithmetic prompt, reasoning_content length was
+# minimal 1161/1033/805, high 1195/741/840, max 957/709/1088 -- fully overlapping,
+# so no graded control is declared. 'none' is likewise accepted without disabling
+# thinking (reasoning_tokens stays non-zero), so thinking.type is the real switch.
+base_model = "minimax/MiniMax-M3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/provider.toml b/providers/volcengine-coding-plan/provider.toml
new file mode 100644
index 00000000000..7420f6ed561
--- /dev/null
+++ b/providers/volcengine-coding-plan/provider.toml
@@ -0,0 +1,42 @@
+name = "Volcengine Ark Coding Plan"
+env = ["ARK_CODING_PLAN_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+# Coding Plan is a subscription tier with its own base URL. Volcengine warns that
+# calling the pay-as-you-go https://ark.cn-beijing.volces.com/api/v3 with a Coding
+# Plan key does NOT draw down the subscription and bills separately, so the two
+# providers must stay distinct.
+# The env var is deliberately not ARK_API_KEY: /api/coding/v3 does not check which
+# plan a key belongs to, so a pay-as-you-go key reaches it and bills per token.
+# Sharing ARK_API_KEY would light up both providers at once and let a user pick a
+# Coding Plan entry while silently paying pay-as-you-go rates.
+# Reasoning HTTP format (accessed 2026-08-26):
+# OpenAI Chat: POST /api/coding/v3/chat/completions; `thinking.type` = enabled or
+# disabled (`auto` is rejected on most models); `reasoning_effort` is accepted on
+# every model but only some values have effect, so each file declares just the
+# levels it honours, measured on this endpoint (n=4-6 per level, 2026-08-27):
+#   doubao-seed-2.1-turbo / 2.0-lite / evolving -> none|low|high
+#       (medium overlaps low, xhigh and max overlap high rather than exceeding it)
+#   glm-5.3                                     -> low|high|max
+#       (rejects none and minimal outright; medium aliases low, xhigh aliases max)
+#   deepseek-v4-flash / -pro                    -> toggle only
+#       (graded values fully overlap; on Flash 'high' lands below 'low')
+#   kimi-k2.7-code                              -> no controls
+#   minimax-m3                                  -> toggle only
+#       (tolerates none but still spends reasoning tokens on it, so it is not an
+#        off switch there; thinking.type is)
+# Plaintext reasoning is in `choices[].message.reasoning_content`
+# and the encrypted block in `choices[].message.encrypted_content`.
+# Anthropic: base https://ark.cn-beijing.volces.com/api/coding, POST /v1/messages
+# with `thinking.type` and `output_config.effort`.
+# Model names here are the short console labels (doubao-seed-2.1-turbo). This
+# endpoint is the more permissive of the two: it accepts the dated IDs as well
+# (verified 2026-08-27 - doubao-seed-2-1-turbo-260628, doubao-seed-2-0-lite-260428
+# and glm-5-3-260930 all resolve here), whereas the pay-as-you-go /api/v3 accepts
+# only the dated form and 404s on a short label. The short labels are used because
+# they are what the Coding Plan console shows. `ark-code-latest` follows whichever
+# model is selected in the console.
+# Sources:
+# https://www.volcengine.com/docs/82379/1928261
+# https://www.volcengine.com/docs/82379/2546386
+doc = "https://www.volcengine.com/docs/82379/1928261"
+api = "https://ark.cn-beijing.volces.com/api/coding/v3"

From 559ab0c1194f474bb8c2dd3d3cde412ca3023215 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 14:04:32 -0500
Subject: [PATCH 0149/2012] fix(opencode-go): add Qwen3.8 reasoning efforts
 (#5741)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../opencode-go/models/qwen3.8-flash.toml     | 10 +++++++-
 providers/opencode-go/models/qwen3.8-max.toml | 24 ++++++++-----------
 2 files changed, 19 insertions(+), 15 deletions(-)

diff --git a/providers/opencode-go/models/qwen3.8-flash.toml b/providers/opencode-go/models/qwen3.8-flash.toml
index 73a89946990..ebfbab557c9 100644
--- a/providers/opencode-go/models/qwen3.8-flash.toml
+++ b/providers/opencode-go/models/qwen3.8-flash.toml
@@ -1,9 +1,17 @@
+# https://opencode.ai/docs/go/#endpoints
+# https://help.aliyun.com/en/model-studio/anthropic-api-messages
+# Toggle (Messages): thinking.type = enabled|disabled
+# Effort (Messages): output_config.effort = low|medium|xhigh
 base_model = "alibaba/qwen3.8-flash"
 temperature = true
 
 [[reasoning_options]]
 type = "toggle"
 
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
 [[reasoning_options]]
 type = "budget_tokens"
 
@@ -14,4 +22,4 @@ cache_read = 0.016
 cache_write = 0.2
 
 [provider]
-npm = "@ai-sdk/anthropic"
\ No newline at end of file
+npm = "@ai-sdk/anthropic"
diff --git a/providers/opencode-go/models/qwen3.8-max.toml b/providers/opencode-go/models/qwen3.8-max.toml
index 2b3305cff9d..efb0a63581f 100644
--- a/providers/opencode-go/models/qwen3.8-max.toml
+++ b/providers/opencode-go/models/qwen3.8-max.toml
@@ -1,15 +1,16 @@
-name = "Qwen3.8 Max"
+# https://opencode.ai/docs/go/#endpoints
+# https://help.aliyun.com/en/model-studio/anthropic-api-messages
+# Toggle (Messages): thinking.type = enabled|disabled
+# Effort (Messages): output_config.effort = low|medium|xhigh
+base_model = "alibaba/qwen3.8-max"
 description = "2.4-trillion-parameter multimodal flagship for coding, professional work, and long-horizon agentic workflows"
 family = "qwen3.8-max"
-release_date = "2026-08-03"
-last_updated = "2026-08-03"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", max = 262_144 }]
-temperature = true
-tool_call = true
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", max = 262_144 },
+]
 structured_output = true
-open_weights = false
 
 [cost]
 input = 2
@@ -17,10 +18,5 @@ output = 6
 cache_read = 0.25
 cache_write = 2.5
 
-[limit]
-context = 1_000_000
-output = 131_072
-
 [modalities]
 input = ["text", "image", "video"]
-output = ["text"]

From dd54159871bacee0141b25b3a694e146be794bf0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:29 +0000
Subject: [PATCH 0150/2012] chore(sync): update Deep Infra model catalog
 (#5753)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/ByteDance/Seed-2.0-mini.toml | 1 -
 providers/deepinfra/models/ByteDance/Seed-2.0-pro.toml  | 1 -
 2 files changed, 2 deletions(-)

diff --git a/providers/deepinfra/models/ByteDance/Seed-2.0-mini.toml b/providers/deepinfra/models/ByteDance/Seed-2.0-mini.toml
index e148079c5f4..c857f3ae913 100644
--- a/providers/deepinfra/models/ByteDance/Seed-2.0-mini.toml
+++ b/providers/deepinfra/models/ByteDance/Seed-2.0-mini.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-mini"
-structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/deepinfra/models/ByteDance/Seed-2.0-pro.toml b/providers/deepinfra/models/ByteDance/Seed-2.0-pro.toml
index be94b676a00..c807fd75189 100644
--- a/providers/deepinfra/models/ByteDance/Seed-2.0-pro.toml
+++ b/providers/deepinfra/models/ByteDance/Seed-2.0-pro.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-pro"
-structured_output = true
 reasoning_options = []
 
 [cost]

From 64b0a7f5e84b71cad27ef25a2bb61b6dacb73d4b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:32 +0000
Subject: [PATCH 0151/2012] chore(sync): update DigitalOcean model catalog
 (#5750)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/deepseek-3.2.toml      |  6 +++---
 providers/digitalocean/models/deepseek-4-flash.toml  |  6 +++---
 .../digitalocean/models/deepseek-v4-flash-0731.toml  |  6 +++---
 providers/digitalocean/models/deepseek-v4-pro.toml   |  6 +++---
 providers/digitalocean/models/glm-5.2.toml           |  6 +++---
 providers/digitalocean/models/kimi-k3.toml           |  6 +++---
 providers/digitalocean/models/llama-4-maverick.toml  |  4 ++--
 providers/digitalocean/models/mimo-v2.5-pro.toml     |  6 +++---
 .../digitalocean/models/openai-gpt-5.6-sol.toml      | 12 ++++++------
 .../digitalocean/models/openai-gpt-oss-120b.toml     |  4 ++--
 10 files changed, 31 insertions(+), 31 deletions(-)

diff --git a/providers/digitalocean/models/deepseek-3.2.toml b/providers/digitalocean/models/deepseek-3.2.toml
index d8d92fa42fa..86a0601e001 100644
--- a/providers/digitalocean/models/deepseek-3.2.toml
+++ b/providers/digitalocean/models/deepseek-3.2.toml
@@ -18,9 +18,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.5
-output = 1.6
-cache_read = 0.15
+input = 0.25
+output = 0.8
+cache_read = 0.075
 
 [limit]
 context = 163_840
diff --git a/providers/digitalocean/models/deepseek-4-flash.toml b/providers/digitalocean/models/deepseek-4-flash.toml
index 143346a359b..2e033308f91 100644
--- a/providers/digitalocean/models/deepseek-4-flash.toml
+++ b/providers/digitalocean/models/deepseek-4-flash.toml
@@ -10,9 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.0679
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-flash-0731.toml b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
index 2e9e64c5632..74ce3b0ae91 100644
--- a/providers/digitalocean/models/deepseek-v4-flash-0731.toml
+++ b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.08
+output = 0.252
+cache_read = 0.0252
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-pro.toml b/providers/digitalocean/models/deepseek-v4-pro.toml
index b4ebe0207c1..8a4edb98729 100644
--- a/providers/digitalocean/models/deepseek-v4-pro.toml
+++ b/providers/digitalocean/models/deepseek-v4-pro.toml
@@ -19,9 +19,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.74
-output = 3.48
-cache_read = 0.348
+input = 0.87
+output = 1.74
+cache_read = 0.174
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/glm-5.2.toml b/providers/digitalocean/models/glm-5.2.toml
index eb3b3d1b78d..3f01602c179 100644
--- a/providers/digitalocean/models/glm-5.2.toml
+++ b/providers/digitalocean/models/glm-5.2.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["medium", "high", "xhigh"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.21
+input = 0.7
+output = 2.2
+cache_read = 0.105
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/kimi-k3.toml b/providers/digitalocean/models/kimi-k3.toml
index 62d0f7b7404..b71a3e29e2e 100644
--- a/providers/digitalocean/models/kimi-k3.toml
+++ b/providers/digitalocean/models/kimi-k3.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.85
+output = 14.25
+cache_read = 0.285
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/digitalocean/models/llama-4-maverick.toml b/providers/digitalocean/models/llama-4-maverick.toml
index 095b2da6674..a72af75a264 100644
--- a/providers/digitalocean/models/llama-4-maverick.toml
+++ b/providers/digitalocean/models/llama-4-maverick.toml
@@ -12,8 +12,8 @@ knowledge = "2024-08"
 open_weights = true
 
 [cost]
-input = 0.25
-output = 0.87
+input = 0.2
+output = 0.696
 
 [limit]
 context = 128_000
diff --git a/providers/digitalocean/models/mimo-v2.5-pro.toml b/providers/digitalocean/models/mimo-v2.5-pro.toml
index 365235fb39f..20f6f6db3ea 100644
--- a/providers/digitalocean/models/mimo-v2.5-pro.toml
+++ b/providers/digitalocean/models/mimo-v2.5-pro.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.8
-output = 3
-cache_read = 0.16
+input = 0.4
+output = 1.5
+cache_read = 0.08
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/openai-gpt-5.6-sol.toml b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
index 3c9af7df850..e14ff09a4ee 100644
--- a/providers/digitalocean/models/openai-gpt-5.6-sol.toml
+++ b/providers/digitalocean/models/openai-gpt-5.6-sol.toml
@@ -6,15 +6,15 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
+input = 4
+output = 20
+cache_read = 0.4
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 10
-output = 45
-cache_read = 1
+input = 8
+output = 30
+cache_read = 0.8
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/digitalocean/models/openai-gpt-oss-120b.toml b/providers/digitalocean/models/openai-gpt-oss-120b.toml
index 4af4fef37ad..fc45d3aac8b 100644
--- a/providers/digitalocean/models/openai-gpt-oss-120b.toml
+++ b/providers/digitalocean/models/openai-gpt-oss-120b.toml
@@ -19,8 +19,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.1
-output = 0.7
+input = 0.055
+output = 0.385
 cache_read = 0.02
 
 [limit]

From 89ebe17f4705fb7bcf164a7bc4dc8837809c5006 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:34 +0000
Subject: [PATCH 0152/2012] chore(sync): update Venice model catalog (#5756)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/seed-2-1-turbo.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/venice/models/seed-2-1-turbo.toml b/providers/venice/models/seed-2-1-turbo.toml
index 5920f325f9b..fe1ac7fc657 100644
--- a/providers/venice/models/seed-2-1-turbo.toml
+++ b/providers/venice/models/seed-2-1-turbo.toml
@@ -2,7 +2,6 @@ base_model = "bytedance-seed/seed-2.1-turbo"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 release_date = "2026-06-28"
 last_updated = "2026-07-24"
-structured_output = true
 reasoning_options = []
 
 [cost]

From bc83ab017ba6275d31ee53d07219e1c64c535df1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:36 +0000
Subject: [PATCH 0153/2012] chore(sync): update NanoGPT model catalog (#5757)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/bytedance-seed/seed-2.0-lite.toml  |  1 -
 providers/nano-gpt/models/claw-high.toml      |  6 ++---
 providers/nano-gpt/models/claw-low.toml       |  6 ++---
 providers/nano-gpt/models/claw-medium.toml    |  6 ++---
 providers/nano-gpt/models/hermes-high.toml    |  6 ++---
 providers/nano-gpt/models/hermes-low.toml     |  6 ++---
 providers/nano-gpt/models/hermes-medium.toml  |  6 ++---
 .../models/inflection/inflection-3-pi.toml    | 24 -------------------
 .../models/qwen/qwen3.8-27b-fable.toml        |  2 +-
 .../nano-gpt/models/zai-org/glm-5.3.toml      |  6 ++---
 .../models/zai-org/glm-5.3:thinking.toml      |  6 ++---
 11 files changed, 25 insertions(+), 50 deletions(-)
 delete mode 100644 providers/nano-gpt/models/inflection/inflection-3-pi.toml

diff --git a/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml b/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
index ff8f59ead35..3b07182e39b 100644
--- a/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
+++ b/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
@@ -4,7 +4,6 @@ name = "ByteDance Seed 2.0 Lite"
 attachment = false
 reasoning = false
 tool_call = false
-structured_output = true
 
 [cost]
 input = 0.25
diff --git a/providers/nano-gpt/models/claw-high.toml b/providers/nano-gpt/models/claw-high.toml
index 7de86921a6d..eb151da1c35 100644
--- a/providers/nano-gpt/models/claw-high.toml
+++ b/providers/nano-gpt/models/claw-high.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/claw-low.toml b/providers/nano-gpt/models/claw-low.toml
index 1da6864013a..2f2bd7a41a0 100644
--- a/providers/nano-gpt/models/claw-low.toml
+++ b/providers/nano-gpt/models/claw-low.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 cache_write = 0.08333
 
 [limit]
diff --git a/providers/nano-gpt/models/claw-medium.toml b/providers/nano-gpt/models/claw-medium.toml
index da9a2c4bb73..c54c7e146a7 100644
--- a/providers/nano-gpt/models/claw-medium.toml
+++ b/providers/nano-gpt/models/claw-medium.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/hermes-high.toml b/providers/nano-gpt/models/hermes-high.toml
index 0a1235c971b..9d6bf229e10 100644
--- a/providers/nano-gpt/models/hermes-high.toml
+++ b/providers/nano-gpt/models/hermes-high.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/hermes-low.toml b/providers/nano-gpt/models/hermes-low.toml
index 51d8dc85b5c..aec46394a9a 100644
--- a/providers/nano-gpt/models/hermes-low.toml
+++ b/providers/nano-gpt/models/hermes-low.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 cache_write = 0.08333
 
 [limit]
diff --git a/providers/nano-gpt/models/hermes-medium.toml b/providers/nano-gpt/models/hermes-medium.toml
index 687214f1c58..223fb950e99 100644
--- a/providers/nano-gpt/models/hermes-medium.toml
+++ b/providers/nano-gpt/models/hermes-medium.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/inflection/inflection-3-pi.toml b/providers/nano-gpt/models/inflection/inflection-3-pi.toml
deleted file mode 100644
index a528d3f5cb6..00000000000
--- a/providers/nano-gpt/models/inflection/inflection-3-pi.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Inflection 3 Pi"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2025-11-21"
-last_updated = "2024-10-11"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.499
-output = 9.996
-cache_read = 1.2495
-
-[limit]
-context = 8_000
-input = 8_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
index 8a877ac220d..e0ad1ce2427 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
@@ -5,7 +5,7 @@ release_date = "2026-08-28"
 last_updated = "2026-08-28"
 attachment = true
 reasoning = true
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = true
 reasoning_options = []
diff --git a/providers/nano-gpt/models/zai-org/glm-5.3.toml b/providers/nano-gpt/models/zai-org/glm-5.3.toml
index e5177fe2893..15216a41c8d 100644
--- a/providers/nano-gpt/models/zai-org/glm-5.3.toml
+++ b/providers/nano-gpt/models/zai-org/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml b/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml
index 723d238b9bc..bb1c9b20d63 100644
--- a/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml
+++ b/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576

From e6ebea3f0e623e0ab61b0d442ddc3103f6699ea2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:40 +0000
Subject: [PATCH 0154/2012] chore(sync): update Kilo model catalog (#5744)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/bytedance-seed/seed-2.0-lite.toml      |  1 -
 .../kilo/models/bytedance-seed/seed-2.0-mini.toml      |  1 -
 providers/kilo/models/meta/muse-glimmer-30b.toml       |  2 +-
 .../kilo/models/mistralai/voxtral-small-24b-2507.toml  |  4 ++--
 .../kilo/models/nvidia/nemotron-3.5-lightning.toml     |  2 +-
 .../kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml    |  4 ++--
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml      |  4 ----
 providers/kilo/models/~google/gemini-flash-latest.toml | 10 +++++-----
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 9 files changed, 16 insertions(+), 22 deletions(-)

diff --git a/providers/kilo/models/bytedance-seed/seed-2.0-lite.toml b/providers/kilo/models/bytedance-seed/seed-2.0-lite.toml
index b4f478fc036..cfdf2d37937 100644
--- a/providers/kilo/models/bytedance-seed/seed-2.0-lite.toml
+++ b/providers/kilo/models/bytedance-seed/seed-2.0-lite.toml
@@ -1,6 +1,5 @@
 base_model = "bytedance-seed/seed-2.0-lite"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-structured_output = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/bytedance-seed/seed-2.0-mini.toml b/providers/kilo/models/bytedance-seed/seed-2.0-mini.toml
index 22dd8120f8b..9d67dc91391 100644
--- a/providers/kilo/models/bytedance-seed/seed-2.0-mini.toml
+++ b/providers/kilo/models/bytedance-seed/seed-2.0-mini.toml
@@ -1,6 +1,5 @@
 base_model = "bytedance-seed/seed-2.0-mini"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-structured_output = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/meta/muse-glimmer-30b.toml b/providers/kilo/models/meta/muse-glimmer-30b.toml
index 2c352c0d845..aa01201020a 100644
--- a/providers/kilo/models/meta/muse-glimmer-30b.toml
+++ b/providers/kilo/models/meta/muse-glimmer-30b.toml
@@ -11,7 +11,7 @@ output = 1.1
 cache_read = 0.04
 
 [limit]
-output = 117_964
+output = 16_384
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml b/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
index 8f7727e8b7b..e73d3ea91b9 100644
--- a/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
+++ b/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
@@ -16,8 +16,8 @@ output = 0.3
 cache_read = 0.01
 
 [limit]
-context = 32_000
-output = 25_600
+context = 32_768
+output = 26_214
 
 [modalities]
 input = ["text", "audio", "pdf"]
diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index 28181cd4e48..f6b9541a819 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -11,4 +11,4 @@ output = 0.2
 cache_read = 0.04
 
 [limit]
-output = 235_929
+output = 131_072
diff --git a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index dee6c4b128f..83d4e34b47c 100644
--- a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 131_072
-output = 32_768
+context = 262_144
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index c79bdfce589..ad993c8f794 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,7 +13,3 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
-
-[limit]
-context = 1_000_000
-output = 262_144
diff --git a/providers/kilo/models/~google/gemini-flash-latest.toml b/providers/kilo/models/~google/gemini-flash-latest.toml
index 944d89b8d27..9e99c8689bf 100644
--- a/providers/kilo/models/~google/gemini-flash-latest.toml
+++ b/providers/kilo/models/~google/gemini-flash-latest.toml
@@ -16,11 +16,11 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-reasoning = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 8f563849520..8b99c94330d 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.2
-output = 1.2
-cache_read = 0.6
+input = 1.25
+output = 4.4
+cache_read = 0.26
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 131_072
 
 [modalities]
 input = ["text"]

From 3487a0d7e5b3a6b557246cbdd99e742f74af5a66 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:42 +0000
Subject: [PATCH 0155/2012] chore(sync): update Charm Hyper model catalog
 (#5743)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml              | 6 +++---
 providers/hyper/models/glm-5.1.toml                         | 6 +++---
 .../models/llama-4-maverick-17b-128e-instruct-fp8.toml      | 6 +++---
 providers/hyper/models/minimax-m2.7.toml                    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 85218385ac6..443e26fe9ac 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.12
-output = 0.42
-cache_write = 0.06
+input = 0.11
+output = 0.408
+cache_write = 0.055
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 4c86ffebd03..78d701cc4e0 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.29
-output = 4.22
-cache_write = 0.645
+input = 1.314
+output = 4.268
+cache_write = 0.657
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
index 9b588de13d6..b36adf9ee14 100644
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.274
-output = 0.8992
-cache_write = 0.137
+input = 0.284
+output = 0.934
+cache_write = 0.142
 
 [limit]
 context = 430_000
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 8dc18f04455..de3e51335c2 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.408
-output = 1.512
-cache_write = 0.204
+input = 0.404
+output = 1.496
+cache_write = 0.202
 
 [limit]
 context = 262_100

From 304613911835ca89d36c69c4cadbeff70b499009 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:43 +0000
Subject: [PATCH 0156/2012] chore(sync): update OpenRouter model catalog
 (#5749)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/bytedance-seed/seed-2.0-lite.toml           |  3 ++-
 .../models/bytedance-seed/seed-2.0-mini.toml           |  3 ++-
 .../models/deepseek/deepseek-v4-flash-0731.toml        |  6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml    |  6 +++---
 .../openrouter/models/google/gemini-3.7-flash.toml     | 10 +++++-----
 providers/openrouter/models/meta/muse-glimmer-30b.toml |  6 +++---
 .../models/mistralai/voxtral-small-24b-2507.toml       |  4 ++--
 providers/openrouter/models/moonshotai/kimi-k3.toml    |  8 +++++---
 .../models/nvidia/nemotron-3.5-lightning.toml          |  8 ++++----
 .../models/qwen/qwen3-vl-30b-a3b-instruct.toml         |  6 +++---
 .../openrouter/models/qwen/qwen3.8-2.4t-a95b.toml      |  3 +--
 providers/openrouter/models/z-ai/glm-5.3.toml          |  2 +-
 .../openrouter/models/~google/gemini-flash-latest.toml | 10 +++++-----
 providers/openrouter/models/~z-ai/glm-latest.toml      |  6 +++---
 14 files changed, 42 insertions(+), 39 deletions(-)

diff --git a/providers/openrouter/models/bytedance-seed/seed-2.0-lite.toml b/providers/openrouter/models/bytedance-seed/seed-2.0-lite.toml
index b86b4bc5916..33cfe7cf3ab 100644
--- a/providers/openrouter/models/bytedance-seed/seed-2.0-lite.toml
+++ b/providers/openrouter/models/bytedance-seed/seed-2.0-lite.toml
@@ -1,6 +1,7 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "bytedance-seed/seed-2.0-lite"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/openrouter/models/bytedance-seed/seed-2.0-mini.toml b/providers/openrouter/models/bytedance-seed/seed-2.0-mini.toml
index e07bb957bbd..abe379d9357 100644
--- a/providers/openrouter/models/bytedance-seed/seed-2.0-mini.toml
+++ b/providers/openrouter/models/bytedance-seed/seed-2.0-mini.toml
@@ -1,6 +1,7 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "bytedance-seed/seed-2.0-mini"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 06d23ff5e01..1273df92943 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.14
-cache_read = 0.014
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 32c1ed84218..ec122acaa20 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.772038
-output = 1.544076
-cache_read = 0.064337
+input = 0.741588
+output = 1.483176
+cache_read = 0.061799
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemini-3.7-flash.toml b/providers/openrouter/models/google/gemini-3.7-flash.toml
index aca24855c7b..db204acd0ef 100644
--- a/providers/openrouter/models/google/gemini-3.7-flash.toml
+++ b/providers/openrouter/models/google/gemini-3.7-flash.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-reasoning = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index 3b46ee2987a..d4077f81891 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.35
-output = 1.5
+input = 0.3
+output = 1.2
 cache_read = 0.04
 
 [limit]
-output = 117_964
+output = 16_384
diff --git a/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml b/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
index 1fdb2e715d2..a6827d692ac 100644
--- a/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
+++ b/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
@@ -16,8 +16,8 @@ output = 0.3
 cache_read = 0.01
 
 [limit]
-context = 32_000
-output = 25_600
+context = 32_768
+output = 26_214
 
 [modalities]
 input = ["text", "audio", "pdf"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 50bd15db035..a3d2ffa5980 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "moonshotai/kimi-k3"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 temperature = true
@@ -10,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.55
+output = 12.75
+cache_read = 0.256
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
index 757af4e5d09..3b251679339 100644
--- a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
@@ -7,9 +7,9 @@ description = "Nemotron model for efficient reasoning, coding, and specialized A
 type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.25
-cache_read = 0.05
+input = 0.08
+output = 0.2
+cache_read = 0.04
 
 [limit]
-output = 235_929
+output = 131_072
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index d857fa53dab..3f78c99ff08 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,12 +12,12 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.13
-output = 0.52
+input = 0.15
+output = 0.6
 
 [limit]
 context = 262_144
-output = 32_768
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index abacdd68c88..51225871db8 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,8 +14,7 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.25
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
-output = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 66d809ab3b9..55248129361 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -11,4 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_310_720
diff --git a/providers/openrouter/models/~google/gemini-flash-latest.toml b/providers/openrouter/models/~google/gemini-flash-latest.toml
index 8dfb0cceba0..b664a0fc44a 100644
--- a/providers/openrouter/models/~google/gemini-flash-latest.toml
+++ b/providers/openrouter/models/~google/gemini-flash-latest.toml
@@ -18,11 +18,11 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-reasoning = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index ccfe85e96f1..2cd92a0f255 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
+input = 1.25
 output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_310_720
+output = 131_072
 
 [modalities]
 input = ["text"]

From 6bb9cecce1b24982bc7048df7a394fbc72adb29b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:46 +0000
Subject: [PATCH 0157/2012] chore(sync): update Eden AI model catalog (#5748)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
index 5b3a9d1765c..9d7aa010e62 100644
--- a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
+++ b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-mini"
-structured_output = true
 
 [[reasoning_options]]
 type = "effort"

From 89a20e7c4e05cc59da9469922f9d2c11c3c01934 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:49 +0000
Subject: [PATCH 0158/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#5758)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3-flash.toml |  6 +++---
 providers/llmgateway/models/qwen3.8-27b.toml   | 10 ++++++++++
 2 files changed, 13 insertions(+), 3 deletions(-)
 create mode 100644 providers/llmgateway/models/qwen3.8-27b.toml

diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index ef942f3d875..e11959adcb4 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.13
+output = 0.4
+cache_read = 0.024
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway/models/qwen3.8-27b.toml b/providers/llmgateway/models/qwen3.8-27b.toml
new file mode 100644
index 00000000000..79646fa55fc
--- /dev/null
+++ b/providers/llmgateway/models/qwen3.8-27b.toml
@@ -0,0 +1,10 @@
+base_model = "alibaba/qwen3.8-27b"
+reasoning_options = []
+
+[cost]
+input = 0.42
+output = 3
+cache_read = 0.085
+
+[limit]
+context = 1_000_000

From 9946ad4c041f3a85af552d13aee9ce90b29c7567 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:00:57 +0000
Subject: [PATCH 0159/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5745)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-2.4t-a95b.toml       |  5 ++++-
 .../models/alibaba/wan-v3.0-video-prime.toml    | 17 +++++++++++++++++
 providers/vercel/models/minimax/minimax-m3.toml |  3 +--
 .../models/moonshotai/kimi-k2.7-code.toml       |  2 +-
 .../models/nvidia/nemotron-3.5-lightning.toml   |  4 ++--
 providers/vercel/models/tencent/hy3.toml        |  9 +++++----
 providers/vercel/models/zai/glm-5.3.toml        |  9 ++++++---
 7 files changed, 36 insertions(+), 13 deletions(-)
 create mode 100644 providers/vercel/models/alibaba/wan-v3.0-video-prime.toml

diff --git a/providers/vercel/models/alibaba/qwen3.8-2.4t-a95b.toml b/providers/vercel/models/alibaba/qwen3.8-2.4t-a95b.toml
index 970bb5a1653..2e1295a93db 100644
--- a/providers/vercel/models/alibaba/qwen3.8-2.4t-a95b.toml
+++ b/providers/vercel/models/alibaba/qwen3.8-2.4t-a95b.toml
@@ -12,7 +12,10 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.2
+cache_read = 0.25
+
+[limit]
+output = 128_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml b/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml
new file mode 100644
index 00000000000..c61b39ecb77
--- /dev/null
+++ b/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml
@@ -0,0 +1,17 @@
+name = "Wan v3.0 Video Prime"
+description = "Video model for prompt-guided generation, editing, and motion workflows"
+family = "o"
+release_date = "2026-08-28"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["video"]
diff --git a/providers/vercel/models/minimax/minimax-m3.toml b/providers/vercel/models/minimax/minimax-m3.toml
index e63dc74aa5e..c17b68ca01e 100644
--- a/providers/vercel/models/minimax/minimax-m3.toml
+++ b/providers/vercel/models/minimax/minimax-m3.toml
@@ -10,8 +10,7 @@ output = 1.2
 cache_read = 0.06
 
 [limit]
-context = 1_000_000
-output = 1_000_000
+context = 512_000
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/moonshotai/kimi-k2.7-code.toml b/providers/vercel/models/moonshotai/kimi-k2.7-code.toml
index ad12aa3fa47..c0e7590b4d2 100644
--- a/providers/vercel/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/vercel/models/moonshotai/kimi-k2.7-code.toml
@@ -5,7 +5,7 @@ reasoning_options = []
 [cost]
 input = 0.95
 output = 4
-cache_read = 0.19
+cache_read = 0.16
 
 [limit]
 context = 256_000
diff --git a/providers/vercel/models/nvidia/nemotron-3.5-lightning.toml b/providers/vercel/models/nvidia/nemotron-3.5-lightning.toml
index a6055f6e02d..cd71f8120c5 100644
--- a/providers/vercel/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/vercel/models/nvidia/nemotron-3.5-lightning.toml
@@ -17,8 +17,8 @@ max = 32_768
 
 [cost]
 input = 0.05
-output = 0.15
-cache_read = 0.05
+output = 0.2
+cache_read = 0.01
 
 [limit]
 output = 131_072
diff --git a/providers/vercel/models/tencent/hy3.toml b/providers/vercel/models/tencent/hy3.toml
index 92236761e0a..066f2f8f4ec 100644
--- a/providers/vercel/models/tencent/hy3.toml
+++ b/providers/vercel/models/tencent/hy3.toml
@@ -12,9 +12,10 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
-output = 128_000
+context = 262_144
+output = 262_144
diff --git a/providers/vercel/models/zai/glm-5.3.toml b/providers/vercel/models/zai/glm-5.3.toml
index a1ee58233c9..05cdccc7976 100644
--- a/providers/vercel/models/zai/glm-5.3.toml
+++ b/providers/vercel/models/zai/glm-5.3.toml
@@ -1,11 +1,14 @@
 base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
 input = 1.4
 output = 4.4
-cache_read = 0.26
+cache_read = 0.14
 
 [limit]
-output = 12_800
+output = 1_000_000

From 6a9fb0d24e32dda0525ec8b048cb2d10d1602f48 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:01:03 +0000
Subject: [PATCH 0160/2012] chore(sync): update LLM Gateway model catalog
 (#5747)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/novita/glm-5.3.toml                | 15 +++++++++++++++
 .../models/novita/qwen3.8-27b.toml            | 17 +++++++++++++++++
 .../models/novita/qwen3.8-flash.toml          | 15 +++++++++++++++
 .../models/scx-ai-gp/glm-5.3-flash.toml       | 19 +++++++++++++++++++
 4 files changed, 66 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/novita/glm-5.3.toml
 create mode 100644 providers/llmgateway-providers/models/novita/qwen3.8-27b.toml
 create mode 100644 providers/llmgateway-providers/models/novita/qwen3.8-flash.toml
 create mode 100644 providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml

diff --git a/providers/llmgateway-providers/models/novita/glm-5.3.toml b/providers/llmgateway-providers/models/novita/glm-5.3.toml
new file mode 100644
index 00000000000..3b82595c37a
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (NovitaAI)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
diff --git a/providers/llmgateway-providers/models/novita/qwen3.8-27b.toml b/providers/llmgateway-providers/models/novita/qwen3.8-27b.toml
new file mode 100644
index 00000000000..32648c39e02
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/qwen3.8-27b.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8 27B (NovitaAI)"
+attachment = false
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 0.42
+output = 3
+cache_read = 0.085
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text"]
diff --git a/providers/llmgateway-providers/models/novita/qwen3.8-flash.toml b/providers/llmgateway-providers/models/novita/qwen3.8-flash.toml
new file mode 100644
index 00000000000..336a549d36a
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/qwen3.8-flash.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-flash"
+name = "Qwen3.8 Flash (NovitaAI)"
+attachment = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+
+[modalities]
+input = ["text"]
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
new file mode 100644
index 00000000000..49ae721b432
--- /dev/null
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
@@ -0,0 +1,19 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (SCX.ai)"
+attachment = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.13
+output = 0.4
+cache_read = 0.024
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text"]

From 67f2df4ef270a430a5de3483aa5544cc3f3cdc48 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 21:01:11 +0000
Subject: [PATCH 0161/2012] chore(sync): update EmpirioLabs AI model catalog
 (#5751)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/seed-2-0-lite.toml  | 1 -
 providers/empiriolabs/models/seed-2-0-mini.toml  | 1 -
 providers/empiriolabs/models/seed-2-0-pro.toml   | 1 -
 providers/empiriolabs/models/seed-2-1-turbo.toml | 1 -
 4 files changed, 4 deletions(-)

diff --git a/providers/empiriolabs/models/seed-2-0-lite.toml b/providers/empiriolabs/models/seed-2-0-lite.toml
index 299714a3910..b2ba0932611 100644
--- a/providers/empiriolabs/models/seed-2-0-lite.toml
+++ b/providers/empiriolabs/models/seed-2-0-lite.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-lite"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/empiriolabs/models/seed-2-0-mini.toml b/providers/empiriolabs/models/seed-2-0-mini.toml
index da2918942a9..6ea20e5b116 100644
--- a/providers/empiriolabs/models/seed-2-0-mini.toml
+++ b/providers/empiriolabs/models/seed-2-0-mini.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-mini"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/empiriolabs/models/seed-2-0-pro.toml b/providers/empiriolabs/models/seed-2-0-pro.toml
index cfec380b517..907c80ad87c 100644
--- a/providers/empiriolabs/models/seed-2-0-pro.toml
+++ b/providers/empiriolabs/models/seed-2-0-pro.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.0-pro"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/empiriolabs/models/seed-2-1-turbo.toml b/providers/empiriolabs/models/seed-2-1-turbo.toml
index 7d3367af0f2..d0749a11c06 100644
--- a/providers/empiriolabs/models/seed-2-1-turbo.toml
+++ b/providers/empiriolabs/models/seed-2-1-turbo.toml
@@ -1,5 +1,4 @@
 base_model = "bytedance-seed/seed-2.1-turbo"
-structured_output = true
 
 [[reasoning_options]]
 type = "toggle"

From 2aa691511115fdc1c790aab8430c836db690d4f7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 22:12:06 +0000
Subject: [PATCH 0162/2012] chore(sync): update Eden AI model catalog (#5761)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../accounts/fireworks/models/gpt-oss-20b.toml     | 10 ----------
 .../nvidia/nemotron-3-ultra-550b-a55b.toml         | 14 --------------
 2 files changed, 24 deletions(-)
 delete mode 100644 providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-20b.toml
 delete mode 100644 providers/edenai/models/together_ai/nvidia/nemotron-3-ultra-550b-a55b.toml

diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-20b.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-20b.toml
deleted file mode 100644
index 52e5e9d4321..00000000000
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-20b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "openai/gpt-oss-20b"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.07
-output = 0.3
-cache_read = 0.035
diff --git a/providers/edenai/models/together_ai/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/edenai/models/together_ai/nvidia/nemotron-3-ultra-550b-a55b.toml
deleted file mode 100644
index e4eec863471..00000000000
--- a/providers/edenai/models/together_ai/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "nvidia/nemotron-3-ultra-550b-a55b"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "medium", "high"]
-
-[cost]
-input = 0.6
-output = 3.6
-cache_read = 0.2
-
-[limit]
-context = 512_288

From f88ee04a427326f6d82259f34af80e992a531f1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 22:12:14 +0000
Subject: [PATCH 0163/2012] chore(sync): update OpenRouter model catalog
 (#5762)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k3.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml | 3 ++-
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index a3d2ffa5980..c11eb6c3dae 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 3
+output = 15
+cache_read = 0.3
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index 51225871db8..abacdd68c88 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,7 +14,8 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.2
+cache_read = 0.25
 
 [limit]
 context = 1_048_576
+output = 262_144

From 7940d7570150ff5550ee4fbee56f45837c5d4ff2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 22:13:08 +0000
Subject: [PATCH 0164/2012] chore(sync): update Kilo model catalog (#5760)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-4-maverick.toml | 2 +-
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml      | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 5443201b6e8..3db8291511b 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [cost]
 input = 0.2
-output = 0.8
+output = 0.696
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index ad993c8f794..c79bdfce589 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,3 +13,7 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
+
+[limit]
+context = 1_000_000
+output = 262_144

From ca762deb0389b0f7750a8d1bde3e61e35a2bf32a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 18:02:57 -0500
Subject: [PATCH 0165/2012] chore(sync): update Weights & Biases model catalog
 (#5752)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-ai/DeepSeek-V4-Pro-0813.toml       | 17 +++++++++++++++++
 providers/wandb/models/moonshotai/Kimi-K3.toml  | 17 -----------------
 .../wandb/models/zai-org/GLM-5.3-Flash.toml     | 17 -----------------
 3 files changed, 17 insertions(+), 34 deletions(-)
 create mode 100644 providers/wandb/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
 delete mode 100644 providers/wandb/models/moonshotai/Kimi-K3.toml
 delete mode 100644 providers/wandb/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/wandb/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/wandb/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
new file mode 100644
index 00000000000..4e032260521
--- /dev/null
+++ b/providers/wandb/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -0,0 +1,17 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+description = "DeepSeek V4-Pro-0813 is a 1.6T-parameter MoE model excelling at advanced reasoning, coding, and complex agentic workloads."
+family = "deepseek"
+release_date = "2026-08-13"
+last_updated = "2026-08-13"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 1.31
+output = 3.96
+cache_read = 0.044
+
+[limit]
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/wandb/models/moonshotai/Kimi-K3.toml b/providers/wandb/models/moonshotai/Kimi-K3.toml
deleted file mode 100644
index 68fa428979c..00000000000
--- a/providers/wandb/models/moonshotai/Kimi-K3.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "moonshotai/kimi-k3"
-description = "Kimi K3 is a 2.8T-parameter multimodal MoE model with 104B active parameters built for long-horizon coding and agentic workflows."
-temperature = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-output = 1_048_576
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
deleted file mode 100644
index 0edd2acb8c5..00000000000
--- a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "zhipuai/glm-5.3-flash"
-name = "GLM 5.3 Flash"
-description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.5
-cache_read = 0.05
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text", "image"]

From cbdddb0350b5ddb9e3bd2306aec797625c62e2f3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:07:33 +0000
Subject: [PATCH 0166/2012] chore(sync): update Hugging Face model catalog
 (#5754)

* chore(sync): update Hugging Face model catalog

* fix(huggingface): correct GLM-5.3 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/huggingface/models/zai-org/GLM-5.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/huggingface/models/zai-org/GLM-5.3.toml

diff --git a/providers/huggingface/models/zai-org/GLM-5.3.toml b/providers/huggingface/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..63a6b009d06
--- /dev/null
+++ b/providers/huggingface/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = low|high|max; GLM-5.3 always reasons.
+# HF exposes model/provider-specific effort; use the lab's supported levels.
+# https://docs.z.ai/guides/llm/glm-5.3
+# https://huggingface.co/docs/inference-providers/en/tasks/chat-completion
+base_model = "zhipuai/glm-5.3"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+open_weights = true
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 1.4
+output = 4.4
+
+[limit]
+context = 1_048_576

From 941740ce067507f17d81bf420c57bafce10779a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:07:36 +0000
Subject: [PATCH 0167/2012] chore(sync): update Requesty model catalog (#5755)

* chore(sync): update Requesty model catalog

* fix(requesty): correct GLM-5.3 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/requesty/models/glm-5.3@eu.toml | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/requesty/models/glm-5.3@eu.toml

diff --git a/providers/requesty/models/glm-5.3@eu.toml b/providers/requesty/models/glm-5.3@eu.toml
new file mode 100644
index 00000000000..a5a072b72d9
--- /dev/null
+++ b/providers/requesty/models/glm-5.3@eu.toml
@@ -0,0 +1,20 @@
+# Effort: reasoning_effort = low|high|max, matching GLM-5.3's native controls.
+# Thinking is always enabled. Requesty's generic budget conversions document
+# other model families, not a GLM-5.3 reasoning-token budget.
+# https://docs.z.ai/guides/llm/glm-5.3
+# https://docs.requesty.ai/features/reasoning
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.75
+output = 4.5
+cache_read = 0.44
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 813dc773ff2e6f689fa9b6a5a12e188401459980 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:07:39 +0000
Subject: [PATCH 0168/2012] chore(sync): update Cortecs model catalog (#5746)

* chore(sync): update Cortecs model catalog

* fix(cortecs): correct GLM-5.3 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/cortecs/models/glm-5.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/cortecs/models/glm-5.3.toml

diff --git a/providers/cortecs/models/glm-5.3.toml b/providers/cortecs/models/glm-5.3.toml
new file mode 100644
index 00000000000..39cd4d9b019
--- /dev/null
+++ b/providers/cortecs/models/glm-5.3.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = low|high|max, matching the lab and relay peers.
+# Cortecs maps reasoning_effort to backend controls; GLM-5.3 always reasons.
+# https://docs.z.ai/guides/llm/glm-5.3
+# https://docs.cortecs.ai/examples/reasoning
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 1.75
+output = 4.499
+cache_read = 0.438
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 28762977e34a9bb028b5cd6fabcd6889a02a647b Mon Sep 17 00:00:00 2001
From: Ahmad Shahzad <107808273+shzdehmd@users.noreply.github.com>
Date: Sat, 29 Aug 2026 04:13:46 +0500
Subject: [PATCH 0169/2012] feat(fireworks-ai): add GLM 5.3, remove deprecated
 DeepSeek V4 Flash (#5763)

Add GLM 5.3 (accounts/fireworks/models/glm-5p3) to the Fireworks AI
provider. Released 2026-08-28 on the Serverless API.
Reference: https://fireworks.ai/models/fireworks/glm-5p3
Pricing: https://docs.fireworks.ai/serverless/pricing

- Thinking-only model: reasoning cannot be disabled (reasoning_effort
  "none" and thinking.type "disabled" both return HTTP 400)
- Two effective effort tiers (high, max) matching the GLM 5.2 pattern;
  low/medium are silently collapsed to high, max/xhigh select max
- reasoning_content returned via interleaved field
- Cost: $1.40/$0.26/$4.40 per 1M tokens (input/cache_read/output)

Remove DeepSeek V4 Flash (accounts/fireworks/models/deepseek-v4-flash),
deprecated from the Serverless API on 2026-08-14 and no longer available.
The dated variant deepseek-v4-flash-0731 remains.
---
 .../fireworks/models/deepseek-v4-flash.toml   | 19 -----------------
 .../accounts/fireworks/models/glm-5p3.toml    | 21 +++++++++++++++++++
 2 files changed, 21 insertions(+), 19 deletions(-)
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash.toml
deleted file mode 100644
index 21fab17a91c..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash"
-# Fireworks' published schema does not document this model's toggle or "max"
-# effort mapping. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
-last_updated = "2026-06-16"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
new file mode 100644
index 00000000000..8fb8037dd97
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
@@ -0,0 +1,21 @@
+# Pricing: https://docs.fireworks.ai/serverless/pricing (accessed 2026-08-29)
+# Released on Fireworks Serverless API 2026-08-28.
+# Thinking-only model: reasoning cannot be disabled (reasoning_effort="none"
+# and thinking.type="disabled" both return 400). Two effective tiers like
+# GLM 5.2: low/medium collapse to high, max/xhigh select max.
+# https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-08-29)
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3"
+last_updated = "2026-08-28"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.26

From 04c8586c96fd8cf8cc8be430230ae317fd5cae4f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:20:06 +0000
Subject: [PATCH 0170/2012] chore(sync): update Kilo model catalog (#5765)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/allenai/olmo-3-32b-think.toml | 27 -------------------
 .../kilo/models/arcee-ai/virtuoso-large.toml  | 22 ---------------
 2 files changed, 49 deletions(-)
 delete mode 100644 providers/kilo/models/allenai/olmo-3-32b-think.toml
 delete mode 100644 providers/kilo/models/arcee-ai/virtuoso-large.toml

diff --git a/providers/kilo/models/allenai/olmo-3-32b-think.toml b/providers/kilo/models/allenai/olmo-3-32b-think.toml
deleted file mode 100644
index d51cca6fc33..00000000000
--- a/providers/kilo/models/allenai/olmo-3-32b-think.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "AllenAI: Olmo 3 32B Think"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "allenai"
-release_date = "2025-11-21"
-last_updated = "2025-11-21"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = false
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["high"]
-
-[cost]
-input = 0.15
-output = 0.5
-
-[limit]
-context = 65_536
-output = 58_982
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/arcee-ai/virtuoso-large.toml b/providers/kilo/models/arcee-ai/virtuoso-large.toml
deleted file mode 100644
index f7d03cc0ca4..00000000000
--- a/providers/kilo/models/arcee-ai/virtuoso-large.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Arcee AI: Virtuoso Large"
-description = "Flagship model for demanding analysis, coding, and production agent workflows"
-release_date = "2025-05-05"
-last_updated = "2025-05-05"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.75
-output = 1.2
-
-[limit]
-context = 131_072
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 986d8033d1e1dcd93aeb5b4ade0409ae7397e710 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:20:12 +0000
Subject: [PATCH 0171/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5769)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/alibaba/qwen3.8-27b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/vercel/models/alibaba/qwen3.8-27b.toml b/providers/vercel/models/alibaba/qwen3.8-27b.toml
index fd3361678fa..84643c210eb 100644
--- a/providers/vercel/models/alibaba/qwen3.8-27b.toml
+++ b/providers/vercel/models/alibaba/qwen3.8-27b.toml
@@ -8,9 +8,10 @@ type = "effort"
 values = ["none", "low", "medium", "xhigh"]
 
 [cost]
-input = 0.55
-output = 3.3
-cache_read = 0.11
+input = 0.5
+output = 3
+cache_read = 0.1
+cache_write = 0.625
 
 [limit]
 context = 1_000_000

From 1c0937c8702dc67a116dbd98b30d19d1ef5194c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:20:14 +0000
Subject: [PATCH 0172/2012] chore(sync): update LLM Gateway model catalog
 (#5768)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/fireworks/deepseek-v4-flash.toml   |  6 +++---
 .../models/permafrost/kimi-k3.toml            | 15 ---------------
 .../models/tundra/kimi-k2.6.toml              | 19 -------------------
 3 files changed, 3 insertions(+), 37 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/permafrost/kimi-k3.toml
 delete mode 100644 providers/llmgateway-providers/models/tundra/kimi-k2.6.toml

diff --git a/providers/llmgateway-providers/models/fireworks/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/fireworks/deepseek-v4-flash.toml
index bdd63f19f70..dd2699004e1 100644
--- a/providers/llmgateway-providers/models/fireworks/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/fireworks/deepseek-v4-flash.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway-providers/models/permafrost/kimi-k3.toml b/providers/llmgateway-providers/models/permafrost/kimi-k3.toml
deleted file mode 100644
index 2ef88145ac8..00000000000
--- a/providers/llmgateway-providers/models/permafrost/kimi-k3.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "moonshotai/kimi-k3"
-name = "Kimi K3 (Permafrost)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.6
-
-[limit]
-output = 1_048_576
diff --git a/providers/llmgateway-providers/models/tundra/kimi-k2.6.toml b/providers/llmgateway-providers/models/tundra/kimi-k2.6.toml
deleted file mode 100644
index d4f3bbcf767..00000000000
--- a/providers/llmgateway-providers/models/tundra/kimi-k2.6.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "moonshotai/kimi-k2.6"
-name = "Kimi K2.6 (Tundra)"
-attachment = false
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.4
-output = 2.2
-cache_read = 0.08
-
-[modalities]
-input = ["text"]

From 678cc87d34798d6c10f1ed472fce1af489f95843 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:20:23 +0000
Subject: [PATCH 0173/2012] chore(sync): update OpenRouter model catalog
 (#5766)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/allenai/olmo-3-32b-think.toml      | 24 -------------------
 .../models/arcee-ai/virtuoso-large.toml       | 23 ------------------
 .../deepseek/deepseek-v4-flash-0731.toml      |  6 ++---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++---
 5 files changed, 9 insertions(+), 56 deletions(-)
 delete mode 100644 providers/openrouter/models/allenai/olmo-3-32b-think.toml
 delete mode 100644 providers/openrouter/models/arcee-ai/virtuoso-large.toml

diff --git a/providers/openrouter/models/allenai/olmo-3-32b-think.toml b/providers/openrouter/models/allenai/olmo-3-32b-think.toml
deleted file mode 100644
index e9191571fa0..00000000000
--- a/providers/openrouter/models/allenai/olmo-3-32b-think.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Olmo 3 32B Think"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "allenai"
-release_date = "2025-11-21"
-last_updated = "2025-11-21"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = false
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.5
-
-[limit]
-context = 65_536
-output = 58_982
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/arcee-ai/virtuoso-large.toml b/providers/openrouter/models/arcee-ai/virtuoso-large.toml
deleted file mode 100644
index 87f8e66b8db..00000000000
--- a/providers/openrouter/models/arcee-ai/virtuoso-large.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Virtuoso Large"
-description = "Flagship model for demanding analysis, coding, and production agent workflows"
-release_date = "2025-05-05"
-last_updated = "2025-05-05"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-knowledge = "2025-03-31"
-open_weights = false
-
-[cost]
-input = 0.75
-output = 1.2
-
-[limit]
-context = 131_072
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 1273df92943..159a52f39fa 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.01
+input = 0.045
+output = 0.09
+cache_read = 0.009
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e9b629fb28e..713b581e8d8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0868
-output = 0.1736
-cache_read = 0.01736
+input = 0.08652
+output = 0.17304
+cache_read = 0.017304
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ec122acaa20..09de32358ac 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.741588
-output = 1.483176
-cache_read = 0.061799
+input = 0.731496
+output = 1.462992
+cache_read = 0.060958
 
 [limit]
 context = 1_048_576

From 62a3fd677f642bfe7b6fb069e4e8f78e8e545b60 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 28 Aug 2026 23:20:40 +0000
Subject: [PATCH 0174/2012] chore(sync): update Venice model catalog (#5771)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/z-ai-glm-5-3.toml | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/providers/venice/models/z-ai-glm-5-3.toml b/providers/venice/models/z-ai-glm-5-3.toml
index 5c3a60289ec..4d41a6a336e 100644
--- a/providers/venice/models/z-ai-glm-5-3.toml
+++ b/providers/venice/models/z-ai-glm-5-3.toml
@@ -3,10 +3,7 @@ name = "GLM 5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 release_date = "2026-08-18"
 last_updated = "2026-08-18"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
+reasoning_options = []
 
 [cost]
 input = 1.75

From 5ef63a3f319ede5400069a6d10ee5d169ecdb434 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 00:52:30 +0000
Subject: [PATCH 0175/2012] chore(sync): update Kilo model catalog (#5775)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v3.2.toml | 2 +-
 providers/kilo/models/tencent/hy3.toml            | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v3.2.toml b/providers/kilo/models/deepseek/deepseek-v3.2.toml
index a29137a7cc0..1af0a0ca64c 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2.toml
@@ -12,4 +12,4 @@ cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 65_536
+output = 147_456
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144

From a2bec47cd96e07e2c58c8d970d2acb0dece46203 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 00:52:38 +0000
Subject: [PATCH 0176/2012] chore(sync): update OpenRouter model catalog
 (#5774)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v3.2.toml   | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v3.2.toml b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
index 80cd9bc29e1..cf2b742ab8d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v3.2.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
@@ -7,10 +7,10 @@ description = "DeepSeek chat model for instruction following, coding, and analys
 type = "toggle"
 
 [cost]
-input = 0.269
-output = 0.4
-cache_read = 0.1345
+input = 0.26
+output = 0.38
+cache_read = 0.13
 
 [limit]
 context = 163_840
-output = 65_536
+output = 147_456
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 713b581e8d8..0a16c6a0710 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08652
-output = 0.17304
-cache_read = 0.017304
+input = 0.08624
+output = 0.17248
+cache_read = 0.017248
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 09de32358ac..557fc7cd752 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.731496
-output = 1.462992
-cache_read = 0.060958
+input = 0.721404
+output = 1.442808
+cache_read = 0.060117
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 467387f9d666aa4e73d41a6c7b7b792081ee0c55 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 01:24:46 +0000
Subject: [PATCH 0177/2012] chore(sync): update NanoGPT model catalog (#5778)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 5213c551274c42e32b61815291736abd746a2600 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 01:24:50 +0000
Subject: [PATCH 0178/2012] chore(sync): update OpenRouter model catalog
 (#5779)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0a16c6a0710..02b6932bc45 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08624
-output = 0.17248
-cache_read = 0.017248
+input = 0.08596
+output = 0.17192
+cache_read = 0.017192
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 557fc7cd752..238455a7870 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.721404
-output = 1.442808
-cache_read = 0.060117
+input = 0.711312
+output = 1.422624
+cache_read = 0.059276
 
 [limit]
 context = 1_048_576

From a4d0d13ea815d52df984f897d1f61f38c4d6bbaf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 01:24:57 +0000
Subject: [PATCH 0179/2012] chore(sync): update Deep Infra model catalog
 (#5777)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/zai-org/GLM-5.3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/deepinfra/models/zai-org/GLM-5.3.toml b/providers/deepinfra/models/zai-org/GLM-5.3.toml
index 517df8be057..16174adddac 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.3.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.3.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.2
+output = 4
+cache_read = 0.24
 
 [limit]
 context = 1_048_576

From b24aed886f002289d743e41d2ceab7daf46e0fb1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 02:29:32 +0000
Subject: [PATCH 0180/2012] chore(sync): update Kilo model catalog (#5780)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/google/gemma-4-31b-it.toml |  4 ++--
 providers/kilo/models/~z-ai/glm-latest.toml      | 10 +++++-----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/google/gemma-4-31b-it.toml b/providers/kilo/models/google/gemma-4-31b-it.toml
index 14b6cfb6e84..8d55ceb50b5 100644
--- a/providers/kilo/models/google/gemma-4-31b-it.toml
+++ b/providers/kilo/models/google/gemma-4-31b-it.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.08
+input = 0.07
 output = 0.35
-cache_read = 0.01
+cache_read = 0.1
 
 [limit]
 output = 16_384
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 8b99c94330d..117dbfcefe9 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.25
-output = 4.4
-cache_read = 0.26
+input = 1.2
+output = 4
+cache_read = 0.24
 
 [limit]
-context = 262_144
-output = 131_072
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 214d6f846f639eb9004234bc41b26b48f1cb5328 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 02:29:41 +0000
Subject: [PATCH 0181/2012] chore(sync): update OpenRouter model catalog
 (#5781)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 2cd92a0f255..a237c5f83a3 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.25
-output = 4.4
-cache_read = 0.26
+input = 1.2
+output = 4
+cache_read = 0.24
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 3eeaa00199b646864ebde5ba774d89d1ebab0081 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 03:27:10 +0000
Subject: [PATCH 0182/2012] chore(sync): update Kilo model catalog (#5783)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 117dbfcefe9..e47c9b45de9 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.2
-output = 4
-cache_read = 0.24
+input = 0.8415
+output = 3.1365
+cache_read = 0.1683
 
 [limit]
 context = 1_048_576

From 773275fbc504f128843b4898088d8b336e5fd313 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 03:27:20 +0000
Subject: [PATCH 0183/2012] chore(sync): update OpenRouter model catalog
 (#5782)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 02b6932bc45..a1e3802eb9d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08596
-output = 0.17192
-cache_read = 0.017192
+input = 0.08568
+output = 0.17136
+cache_read = 0.017136
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 238455a7870..e45becad75b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.711312
-output = 1.422624
-cache_read = 0.059276
+input = 0.701046
+output = 1.402092
+cache_read = 0.058421
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index a237c5f83a3..ca263cd9a02 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.2
-output = 4
-cache_read = 0.24
+input = 0.8415
+output = 3.1365
+cache_read = 0.1683
 
 [limit]
 context = 1_310_720

From a12055a57041fe1931f96eb9f7be0a4765ebbe28 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 04:26:55 +0000
Subject: [PATCH 0184/2012] chore(sync): update OpenRouter model catalog
 (#5785)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a1e3802eb9d..4dca8a54662 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08568
-output = 0.17136
-cache_read = 0.017136
+input = 0.0854
+output = 0.1708
+cache_read = 0.01708
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index ca263cd9a02..a237c5f83a3 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8415
-output = 3.1365
-cache_read = 0.1683
+input = 1.2
+output = 4
+cache_read = 0.24
 
 [limit]
 context = 1_310_720

From 02c4f49b859c8ceea06b88b6c43eacfc4f279070 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 04:26:57 +0000
Subject: [PATCH 0185/2012] chore(sync): update NanoGPT model catalog (#5786)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From d9b639780bad444fff9185d287b22db6c4a98406 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 04:27:21 +0000
Subject: [PATCH 0186/2012] chore(sync): update Kilo model catalog (#5784)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index e47c9b45de9..117dbfcefe9 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8415
-output = 3.1365
-cache_read = 0.1683
+input = 1.2
+output = 4
+cache_read = 0.24
 
 [limit]
 context = 1_048_576

From ebca5e1873bf2757704cc0804dbe9a91d7ab6aad Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 05:24:12 +0000
Subject: [PATCH 0187/2012] chore(sync): update Eden AI model catalog (#5789)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/gpt-oss-20b.toml    |  4 ++--
 .../nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml |  2 +-
 .../models/scaleway/deepseek-v4-flash-0731.toml    |  1 -
 providers/edenai/models/scaleway/gpt-oss-120b.toml |  1 -
 .../models/scaleway/llama-3.3-70b-instruct.toml    |  1 -
 .../tensorx/deepseek/deepseek-v4-pro-0813.toml     | 14 ++++++++++++++
 6 files changed, 17 insertions(+), 6 deletions(-)
 create mode 100644 providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml

diff --git a/providers/edenai/models/flexai/gpt-oss-20b.toml b/providers/edenai/models/flexai/gpt-oss-20b.toml
index cd92dbb1355..64a0c7acc13 100644
--- a/providers/edenai/models/flexai/gpt-oss-20b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-20b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.03
-output = 0.13
+input = 0.02
+output = 0.1
diff --git a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index e7b6da98f38..957b9dd7a3d 100644
--- a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -10,4 +10,4 @@ output = 0.28
 cache_read = 0.14
 
 [limit]
-context = 1_048_576
+context = 1_024_000
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index ca937b79460..31f9aa58835 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -1,5 +1,4 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
-tool_call = false
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 75b0577f6a1..75b35e954bc 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -1,5 +1,4 @@
 base_model = "openai/gpt-oss-120b"
-tool_call = false
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index b10ebf79d5f..9304c96d6a1 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 attachment = false
-tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml b/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..74e9ef4da95
--- /dev/null
+++ b/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 2
+output = 4
+cache_read = 0.5
+
+[limit]
+context = 1_048_576

From 77d23a28b35ea96dee43dd63e0e31d579293ae1c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 05:24:16 +0000
Subject: [PATCH 0188/2012] chore(sync): update OpenRouter model catalog
 (#5788)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 4dca8a54662..ac4930e9112 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0854
-output = 0.1708
-cache_read = 0.01708
+input = 0.08512
+output = 0.17024
+cache_read = 0.017024
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e45becad75b..93a0189705f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.701046
-output = 1.402092
-cache_read = 0.058421
+input = 0.690954
+output = 1.381908
+cache_read = 0.05758
 
 [limit]
 context = 1_048_576

From 00123b841010ac6391663ca71ed27d4d340d0825 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 06:34:55 +0000
Subject: [PATCH 0189/2012] chore(sync): update OpenRouter model catalog
 (#5790)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ac4930e9112..a878d771be8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08512
-output = 0.17024
-cache_read = 0.017024
+input = 0.08484
+output = 0.16968
+cache_read = 0.016968
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 93a0189705f..a1720951ed4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.690954
-output = 1.381908
-cache_read = 0.05758
+input = 0.680862
+output = 1.361724
+cache_read = 0.056739
 
 [limit]
 context = 1_048_576

From 19162441a4b0d312f2223a27b57a4138c61707a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 06:35:05 +0000
Subject: [PATCH 0190/2012] chore(sync): update NanoGPT model catalog (#5791)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 6b3d2579326e02ef83a412b17f8c5c8ffa3d7e83 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 07:25:00 +0000
Subject: [PATCH 0191/2012] chore(sync): update OpenRouter model catalog
 (#5794)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v3.2.toml   | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v3.2.toml b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
index cf2b742ab8d..80cd9bc29e1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v3.2.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
@@ -7,10 +7,10 @@ description = "DeepSeek chat model for instruction following, coding, and analys
 type = "toggle"
 
 [cost]
-input = 0.26
-output = 0.38
-cache_read = 0.13
+input = 0.269
+output = 0.4
+cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a878d771be8..96acdd95652 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08484
-output = 0.16968
-cache_read = 0.016968
+input = 0.08456
+output = 0.16912
+cache_read = 0.016912
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a1720951ed4..92fee3d8004 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.680862
-output = 1.361724
-cache_read = 0.056739
+input = 0.670596
+output = 1.341192
+cache_read = 0.055883
 
 [limit]
 context = 1_048_576

From 403cf1ca7c194162e2628645c1e192cb3583b69f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 07:25:02 +0000
Subject: [PATCH 0192/2012] chore(sync): update Kilo model catalog (#5796)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v3.2.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v3.2.toml b/providers/kilo/models/deepseek/deepseek-v3.2.toml
index 1af0a0ca64c..a29137a7cc0 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2.toml
@@ -12,4 +12,4 @@ cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536

From cfdc17812e73f39f718d84da7b632ef82ee303df Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 07:25:16 +0000
Subject: [PATCH 0193/2012] chore(sync): update NanoGPT model catalog (#5795)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../google/gemma-4-31b-it-gemsicle.toml       | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml

diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml b/providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml
new file mode 100644
index 00000000000..44d7cd91e9d
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml
@@ -0,0 +1,24 @@
+name = "Gemsicle"
+description = "Gemsicle is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
+family = "gemma"
+release_date = "2026-08-29"
+last_updated = "2026-08-29"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 0722469f7f9a5ba4769db8dfb45de044b18a207f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 08:28:17 +0000
Subject: [PATCH 0194/2012] chore(sync): update OpenRouter model catalog
 (#5799)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 96acdd95652..fb4a3f8d17b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08456
-output = 0.16912
-cache_read = 0.016912
+input = 0.08428
+output = 0.16856
+cache_read = 0.016856
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92fee3d8004..342ec931efb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.670596
-output = 1.341192
-cache_read = 0.055883
+input = 0.660504
+output = 1.321008
+cache_read = 0.055042
 
 [limit]
 context = 1_048_576

From faa6ca2771c40c32f0ded42ccf146e1def4937b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 08:28:25 +0000
Subject: [PATCH 0195/2012] chore(sync): update NanoGPT model catalog (#5800)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/yi-lightning.toml | 24 ---------------------
 1 file changed, 24 deletions(-)
 delete mode 100644 providers/nano-gpt/models/yi-lightning.toml

diff --git a/providers/nano-gpt/models/yi-lightning.toml b/providers/nano-gpt/models/yi-lightning.toml
deleted file mode 100644
index 97c5e1929a4..00000000000
--- a/providers/nano-gpt/models/yi-lightning.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Yi Lightning"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "yi"
-release_date = "2025-06-15"
-last_updated = "2024-10-16"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.2006
-output = 0.2006
-cache_read = 0.1003
-
-[limit]
-context = 12_000
-input = 12_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]

From e1d4ef08c2aaba06e5149b71f2e4b90e10f765ce Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 09:23:42 +0000
Subject: [PATCH 0196/2012] chore(sync): update OpenRouter model catalog
 (#5801)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index fb4a3f8d17b..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08428
-output = 0.16856
-cache_read = 0.016856
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 342ec931efb..923a5702f14 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.660504
-output = 1.321008
-cache_read = 0.055042
+input = 0.650412
+output = 1.300824
+cache_read = 0.054201
 
 [limit]
 context = 1_048_576

From 9dfb8774173d077355f6adc25dd9f9e9f8559945 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 10:22:59 +0000
Subject: [PATCH 0197/2012] chore(sync): update OpenRouter model catalog
 (#5803)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..7890afacc42 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08372
+output = 0.16744
+cache_read = 0.016744
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 923a5702f14..8ce1e2e483c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.650412
-output = 1.300824
-cache_read = 0.054201
+input = 0.64032
+output = 1.28064
+cache_read = 0.05336
 
 [limit]
 context = 1_048_576

From efdbe97a8bdb15da5d4dbbbf31cb897160cb4e01 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 10:23:10 +0000
Subject: [PATCH 0198/2012] chore(sync): update NanoGPT model catalog (#5804)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From ef75469d66e46854f8b86ea13d19fb0ffe0f73f0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 11:21:10 +0000
Subject: [PATCH 0199/2012] chore(sync): update OpenRouter model catalog
 (#5807)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 8ce1e2e483c..7661c5c8794 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.64032
-output = 1.28064
-cache_read = 0.05336
+input = 0.630054
+output = 1.260108
+cache_read = 0.052505
 
 [limit]
 context = 1_048_576

From df6579b54252fb45cee4e8f690b4685284d54c3e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 12:31:12 +0000
Subject: [PATCH 0200/2012] chore(sync): update OpenRouter model catalog
 (#5810)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 7890afacc42..110d9a84fa6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08372
-output = 0.16744
-cache_read = 0.016744
+input = 0.08344
+output = 0.16688
+cache_read = 0.016688
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 7661c5c8794..4c63f685375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.630054
-output = 1.260108
-cache_read = 0.052505
+input = 0.619962
+output = 1.239924
+cache_read = 0.051664
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index a237c5f83a3..0e324e4270a 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.2
-output = 4
-cache_read = 0.24
+input = 1.1875
+output = 4.18
+cache_read = 0.247
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From ab56db345224c34f9db8c46005a1239076bd7d11 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 12:31:24 +0000
Subject: [PATCH 0201/2012] chore(sync): update Kilo model catalog (#5811)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 117dbfcefe9..4fbd2317736 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.2
-output = 4
-cache_read = 0.24
+input = 1.1875
+output = 4.18
+cache_read = 0.247
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 131_072
 
 [modalities]
 input = ["text"]

From d174759350b2ff139d71894591903f9f70e8bc57 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 13:21:24 +0000
Subject: [PATCH 0202/2012] chore(sync): update OpenRouter model catalog
 (#5815)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 110d9a84fa6..b9598043f12 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08344
-output = 0.16688
-cache_read = 0.016688
+input = 0.08316
+output = 0.16632
+cache_read = 0.016632
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4c63f685375..16d0dad8b3c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.619962
-output = 1.239924
-cache_read = 0.051664
+input = 0.60987
+output = 1.21974
+cache_read = 0.050823
 
 [limit]
 context = 1_048_576

From d51a4b6dead678a50c352696098fb3b1015431bc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 14:22:26 +0000
Subject: [PATCH 0203/2012] chore(sync): update OpenRouter model catalog
 (#5816)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b9598043f12..110d9a84fa6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08316
-output = 0.16632
-cache_read = 0.016632
+input = 0.08344
+output = 0.16688
+cache_read = 0.016688
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 16d0dad8b3c..7f35618a98a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.60987
-output = 1.21974
-cache_read = 0.050823
+input = 0.599778
+output = 1.199556
+cache_read = 0.049982
 
 [limit]
 context = 1_048_576

From ca7157fe5acca57b0c6b2bd24409041834ec787d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 15:21:51 +0000
Subject: [PATCH 0204/2012] chore(sync): update OpenRouter model catalog
 (#5817)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 110d9a84fa6..b9598043f12 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08344
-output = 0.16688
-cache_read = 0.016688
+input = 0.08316
+output = 0.16632
+cache_read = 0.016632
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 7f35618a98a..318c1d0ec69 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.599778
-output = 1.199556
-cache_read = 0.049982
+input = 0.589512
+output = 1.179024
+cache_read = 0.049126
 
 [limit]
 context = 1_048_576

From f50e1759e07bfd28f02de9d50071ec6c065dbc4b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 16:23:59 +0000
Subject: [PATCH 0205/2012] chore(sync): update OpenRouter model catalog
 (#5820)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 318c1d0ec69..5561881c9fa 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.589512
-output = 1.179024
-cache_read = 0.049126
+input = 0.57942
+output = 1.15884
+cache_read = 0.048285
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From b3db1ac80f7297acd3e1ae131b94ac3ab6fc162c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 16:24:02 +0000
Subject: [PATCH 0206/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#5822)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.3.toml b/providers/llmgateway/models/glm-5.3.toml
index 35b4946bedd..9bc1eda393c 100644
--- a/providers/llmgateway/models/glm-5.3.toml
+++ b/providers/llmgateway/models/glm-5.3.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.3
+output = 4
+cache_read = 0.25
 
 [limit]
 context = 1_048_576

From dbf59f1b32edf55833daa4f44591cb83a26233ae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 16:24:13 +0000
Subject: [PATCH 0207/2012] chore(sync): update Kilo model catalog (#5819)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From e95bbd79d0e66ddf98fe53e421ac2830da6aa7c9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 16:24:20 +0000
Subject: [PATCH 0208/2012] chore(sync): update LLM Gateway model catalog
 (#5821)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/scx-ai-gp/glm-5.3.toml                 | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml

diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml
new file mode 100644
index 00000000000..9e9ce7cdee3
--- /dev/null
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (SCX.ai)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.3
+output = 4
+cache_read = 0.25
+
+[limit]
+output = 128_000

From 256da29a78274b3b4b0b365fbb7f5b6193b37711 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 17:21:19 +0000
Subject: [PATCH 0209/2012] chore(sync): update OpenRouter model catalog
 (#5823)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b9598043f12..716939011bf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08316
-output = 0.16632
-cache_read = 0.016632
+input = 0.08288
+output = 0.16576
+cache_read = 0.016576
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 5561881c9fa..828f6926def 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.57942
-output = 1.15884
-cache_read = 0.048285
+input = 0.569328
+output = 1.138656
+cache_read = 0.047444
 
 [limit]
 context = 1_048_576

From b6a58e4608b1c03ea3f0470292ee6eec86956f6f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 18:26:47 +0000
Subject: [PATCH 0210/2012] chore(sync): update OpenRouter model catalog
 (#5826)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 716939011bf..365c85db2f9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08288
-output = 0.16576
-cache_read = 0.016576
+input = 0.0826
+output = 0.1652
+cache_read = 0.01652
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index fe64f45e1df..08df6887228 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.007
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 262_144
 
 [modalities]
 input = ["text"]

From 6b50c5bf0c3d6a4390bc34b683a9c5d6571fad54 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 18:27:24 +0000
Subject: [PATCH 0211/2012] chore(sync): update Kilo model catalog (#5827)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 7aa572fb995..c3e871235b5 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.007
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 262_144
 
 [modalities]
 input = ["text"]

From 82a91a0595394f9fe4fed41f76adf492af94606b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 19:19:48 +0000
Subject: [PATCH 0212/2012] chore(sync): update OpenRouter model catalog
 (#5830)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 365c85db2f9..6c940659715 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0826
-output = 0.1652
-cache_read = 0.01652
+input = 0.08232
+output = 0.16464
+cache_read = 0.016464
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 828f6926def..0a9a2a70cdf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.569328
-output = 1.138656
-cache_read = 0.047444
+input = 0.559062
+output = 1.118124
+cache_read = 0.046589
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 08df6887228..64bd01fa1ab 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.007
 
 [limit]
 context = 1_310_720
-output = 262_144
+output = 393_216
 
 [modalities]
 input = ["text"]

From d630ca6b89d219c2549afac3b82669df10cf4304 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 19:19:57 +0000
Subject: [PATCH 0213/2012] chore(sync): update Kilo model catalog (#5831)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index c3e871235b5..92950df32ad 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.007
 
 [limit]
 context = 1_048_576
-output = 262_144
+output = 393_216
 
 [modalities]
 input = ["text"]

From 9e01e804bfe029d136c27339050d4258489fe069 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 20:23:34 +0000
Subject: [PATCH 0214/2012] chore(sync): update OpenRouter model catalog
 (#5832)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6c940659715..98144b8c68a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08232
-output = 0.16464
-cache_read = 0.016464
+input = 0.08204
+output = 0.16408
+cache_read = 0.016408
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 0a9a2a70cdf..cfffc22c9a8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.559062
-output = 1.118124
-cache_read = 0.046589
+input = 0.54897
+output = 1.09794
+cache_read = 0.045748
 
 [limit]
 context = 1_048_576

From db8cf18764f7911138085f587447f3bdc1c12e54 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 15:41:56 -0500
Subject: [PATCH 0215/2012] chore(sync): update DigitalOcean model catalog
 (#5767)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/glm-5.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/digitalocean/models/glm-5.3.toml

diff --git a/providers/digitalocean/models/glm-5.3.toml b/providers/digitalocean/models/glm-5.3.toml
new file mode 100644
index 00000000000..b736e699d4c
--- /dev/null
+++ b/providers/digitalocean/models/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 59fa8dc14157bf4376ae80de8fc9b9427f6d2814 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 15:48:01 -0500
Subject: [PATCH 0216/2012] Automate Cloudflare AI Gateway model sync (#5764)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
Co-authored-by: superhighfives <449385+superhighfives@users.noreply.github.com>
---
 .github/workflows/sync-models.yml             |   2 +
 .../script/generate-cloudflare-ai-gateway.ts  | 417 +--------------
 packages/core/src/sync/auto-merge.ts          |   8 +
 packages/core/src/sync/index.ts               |  70 ++-
 .../sync/providers/cloudflare-ai-gateway.ts   | 502 ++++++++++++++++++
 packages/core/test/auto-merge.test.ts         |  12 +
 .../core/test/cloudflare-ai-gateway.test.ts   | 433 +++++++++++++++
 providers/cloudflare-ai-gateway/README.md     |   7 +-
 sync.md                                       |  12 +-
 9 files changed, 1038 insertions(+), 425 deletions(-)
 create mode 100644 packages/core/src/sync/providers/cloudflare-ai-gateway.ts
 create mode 100644 packages/core/test/cloudflare-ai-gateway.test.ts

diff --git a/.github/workflows/sync-models.yml b/.github/workflows/sync-models.yml
index 7edd0324244..f741985ae65 100644
--- a/.github/workflows/sync-models.yml
+++ b/.github/workflows/sync-models.yml
@@ -88,6 +88,8 @@ jobs:
           GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
           GOOGLE_GENERATIVE_AI_API_KEY: ${{ secrets.GOOGLE_GENERATIVE_AI_API_KEY }}
           XAI_API_KEY: ${{ secrets.XAI_API_KEY }}
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }}
           CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID }}
           CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN }}
 
diff --git a/packages/core/script/generate-cloudflare-ai-gateway.ts b/packages/core/script/generate-cloudflare-ai-gateway.ts
index f44c36e38c9..b74d62a8c8f 100644
--- a/packages/core/script/generate-cloudflare-ai-gateway.ts
+++ b/packages/core/script/generate-cloudflare-ai-gateway.ts
@@ -1,417 +1,14 @@
 #!/usr/bin/env bun
-//
-// Regenerate the cloudflare-ai-gateway provider model TOMLs from Cloudflare's own sources,
-// with human curation reduced to providers/cloudflare-ai-gateway/curation.toml.
-//
-// Scope: proxied third-party models only (anthropic, openai, google, xai, alibaba, deepseek,
-// moonshotai, …). Cloudflare's own Workers AI (@cf/...) models are a different pathway — hosted
-// on Cloudflare, CF-token auth, model agreements — and live in their own provider,
-// providers/cloudflare-workers-ai, so they are deliberately not mirrored here.
-//
-// Source of truth (live):
-//   - GET /accounts/{id}/ai/catalog/models — canonical dotted model_id, name, description,
-//       context_length, max_output_tokens, and pricing (flat or context-tiered).
-//
-// curation.toml holds only what that source cannot express: structured_output (a quality
-// judgement — Cloudflare advertises response_format broadly but several models do not honour
-// it), reasoning_options (the catalog has no reasoning schema), limit divergences, and a skip
-// list for catalog ids with no lab file (or not reachable via unified billing).
-//
-// Env: CLOUDFLARE_API_TOKEN, CLOUDFLARE_ACCOUNT_ID
-//      CF_AIG_FIXTURE_DIR (optional) — read cached catalog* JSON instead of the network.
-//
-// Usage:
-//   CLOUDFLARE_API_TOKEN=… CLOUDFLARE_ACCOUNT_ID=… bun run cloudflare-ai-gateway:generate
-//   bun run cloudflare-ai-gateway:generate --check     # fail if the tree would change
 
-import path from "node:path";
-import { readdirSync, readFileSync, statSync, existsSync, rmSync } from "node:fs";
-import { z } from "zod";
-import { formatToml } from "../src/sync/index.ts";
+import { syncProviderByID } from "../src/sync/index.js";
 
-const PROVIDER_DIR = path.join(
-  import.meta.dirname, "..", "..", "..", "providers", "cloudflare-ai-gateway",
-);
-const MODELS_DIR = path.join(PROVIDER_DIR, "models");
-const MODELS_ROOT = path.join(import.meta.dirname, "..", "..", "..", "models");
-const CURATION_PATH = path.join(PROVIDER_DIR, "curation.toml");
+const check = process.argv.includes("--check");
+const result = await syncProviderByID("cloudflare-ai-gateway", { dryRun: check });
 
-const TEXT_GENERATION = "Text Generation";
-
-// Proxied providers that Cloudflare fronts with a *native* passthrough route rather than the
-// gateway's generic OpenAI-compatible transform: Anthropic keeps the Messages API, OpenAI keeps
-// the Responses API. Advertise each model's native SDK so consumers route to the endpoint that
-// serves it best instead of falling back to the provider default (ai-gateway-provider). Other
-// third-party providers Cloudflare only exposes over the compat route inherit that default.
-// https://developers.cloudflare.com/ai-gateway/usage/providers/anthropic/
-// https://developers.cloudflare.com/ai-gateway/usage/providers/openai/
-const NATIVE_NPM: Record<string, string> = {
-  anthropic: "@ai-sdk/anthropic",
-  openai: "@ai-sdk/openai",
-};
-function nativeNpm(id: string): string | undefined {
-  return NATIVE_NPM[id.split("/")[0]!];
-}
-
-// ---------------------------------------------------------------------------
-// curation.toml schema
-// ---------------------------------------------------------------------------
-const ReasoningOption = z.record(z.any());
-const CuratedModel = z
-  .object({
-    base_model: z.string().min(1).optional(),
-    structured_output: z.boolean().optional(),
-    reasoning_options: z.array(ReasoningOption).optional(),
-    limit: z.record(z.number()).optional(),
-    interleaved: z
-      .union([z.literal(true), z.object({ field: z.enum(["reasoning_content", "reasoning_details"]) }).strict()])
-      .optional(),
-    // Leading `#` comment lines, e.g. a toggle/effort wire-path note (AGENTS.md requires one
-    // for every `toggle` reasoning option) or a source citation. Rendered verbatim above the
-    // generated fields so hand-verified host behavior survives every regeneration.
-    note: z.array(z.string()).optional(),
-  })
-  .strict();
-
-const Curation = z
-  .object({
-    skip: z.array(z.string()).default([]),
-    models: z.record(CuratedModel).default({}),
-  })
-  .strict();
-
-// ---------------------------------------------------------------------------
-// Fetch (with fixture fallback)
-// ---------------------------------------------------------------------------
-async function fetchAllPages(url: string, token: string, perPage: number) {
-  const out: any[] = [];
-  for (let page = 1; page < 50; page++) {
-    const res = await fetch(`${url}?page=${page}&per_page=${perPage}`, {
-      headers: { Authorization: `Bearer ${token}` },
-    });
-    if (!res.ok) throw new Error(`Fetch failed ${res.status} ${res.statusText} for ${url}`);
-    const json: any = await res.json();
-    out.push(...(json.result ?? []));
-    const total = json.result_info?.total_count ?? out.length;
-    if (out.length >= total || (json.result ?? []).length === 0) break;
-  }
-  return out;
-}
-
-// Fetch with retry on 429/5xx. raw.githubusercontent.com rate-limits bursts, so honour
-// Retry-After when present and otherwise back off exponentially. Returns the Response;
-// callers decide how to treat a final !ok (throw vs. tolerate).
-async function fetchWithRetry(url: string, init?: RequestInit, tries = 5): Promise<Response> {
-  let delay = 500;
-  for (let attempt = 1; ; attempt++) {
-    const res = await fetch(url, init);
-    if (res.ok || (res.status !== 429 && res.status < 500) || attempt >= tries) return res;
-    const retryAfter = Number(res.headers.get("retry-after"));
-    const wait = Number.isFinite(retryAfter) && retryAfter > 0 ? retryAfter * 1000 : delay;
-    await new Promise((r) => setTimeout(r, wait));
-    delay = Math.min(delay * 2, 8000);
-  }
-}
-
-// Run async tasks with bounded concurrency to avoid tripping rate limits.
-async function mapLimit<T, R>(items: T[], limit: number, fn: (item: T) => Promise<R>): Promise<R[]> {
-  const results = new Array<R>(items.length);
-  let next = 0;
-  const workers = Array.from({ length: Math.min(limit, items.length) }, async () => {
-    while (true) {
-      const i = next++;
-      if (i >= items.length) break;
-      results[i] = await fn(items[i]!);
-    }
-  });
-  await Promise.all(workers);
-  return results;
-}
-
-// Load rows from every fixture file whose name starts with prefix, de-duplicated by their
-// catalog model_id. Dedup guards against overlapping snapshot files inflating the model set.
-function loadFixtureRows(dir: string, prefix: string): any[] {
-  const byId = new Map<string, any>();
-  for (const f of readdirSync(dir).filter((f) => f.startsWith(prefix) && f.endsWith(".json"))) {
-    for (const row of JSON.parse(readFileSync(path.join(dir, f), "utf8")).result ?? []) {
-      const key = row.model_id ?? row.name ?? JSON.stringify(row);
-      byId.set(key, row);
-    }
-  }
-  return [...byId.values()];
-}
-
-async function loadProxied() {
-  const fixtureDir = process.env.CF_AIG_FIXTURE_DIR;
-  if (fixtureDir) return loadFixtureRows(fixtureDir, "catalog");
-  const token = process.env.CLOUDFLARE_API_TOKEN;
-  const account = process.env.CLOUDFLARE_ACCOUNT_ID;
-  if (!token || !account) {
-    throw new Error(
-      "Set CLOUDFLARE_API_TOKEN + CLOUDFLARE_ACCOUNT_ID (or CF_AIG_FIXTURE_DIR for offline runs).",
-    );
-  }
-  const base = `https://api.cloudflare.com/client/v4/accounts/${account}/ai`;
-  return fetchAllPages(`${base}/catalog/models`, token, 50);
-}
-
-// Load the per-model catalog schema for a proxied model. The list endpoint omits `schema`;
-// the single-model schema endpoint returns schema.input, from which reasoning_options are
-// derivable for OpenAI-compatible providers (xai, alibaba, openai). Providers whose schema
-// is their native shape (google, anthropic, deepseek, moonshotai) return no reasoning
-// property — those fall back to curation.toml. Returns schema.input or undefined.
-async function loadCatalogSchemaInput(id: string): Promise<unknown> {
-  const fixtureDir = process.env.CF_AIG_FIXTURE_DIR;
-  if (fixtureDir) {
-    const p = path.join(fixtureDir, "schema", `${id.replace(/\//g, "_")}.json`);
-    if (!existsSync(p)) return undefined; // schema is optional per-model
-    return JSON.parse(readFileSync(p, "utf8")).result?.schema?.input;
-  }
-  const token = process.env.CLOUDFLARE_API_TOKEN!;
-  const account = process.env.CLOUDFLARE_ACCOUNT_ID!;
-  const res = await fetchWithRetry(
-    `https://api.cloudflare.com/client/v4/accounts/${account}/ai/catalog/models/${id}/schema`,
-    { headers: { Authorization: `Bearer ${token}` } },
-  );
-  if (!res.ok) return undefined; // treat missing schema as "not derivable"
-  return ((await res.json()) as any).result?.schema?.input;
-}
-
-// Read the `reasoning` flag from a base lab file (models/<base>.toml). A base model that
-// declares reasoning=true MUST carry reasoning_options in the provider file (schema validates
-// this), so we hard-fail when neither the catalog schema nor curation can supply them.
-function baseReasoning(base: string): boolean {
-  const p = path.join(MODELS_ROOT, `${base}.toml`);
-  if (!existsSync(p)) return false;
-  return (Bun.TOML.parse(readFileSync(p, "utf8")) as any).reasoning === true;
-}
-
-// ---------------------------------------------------------------------------
-// Pricing → cost
-// ---------------------------------------------------------------------------
-const FLAT_KEYS: Record<string, string> = {
-  "Input tokens (per 1M)": "input",
-  "Output tokens (per 1M)": "output",
-  "Cached input tokens (per 1M)": "cache_read",
-  "Cache creation tokens (per 1M)": "cache_write",
-};
-const TIER_RE = /^(Input|Output|Cached input)\s*(<=?|>=?)\s*(\d+)k\s*\(per 1M\)$/;
-const TIER_FIELD: Record<string, string> = {
-  Input: "input",
-  Output: "output",
-  "Cached input": "cache_read",
-};
-
-function proxiedCost(pricing: Record<string, number>, id: string, warnings: string[]) {
-  const base: Record<string, number> = {};
-  for (const [key, value] of Object.entries(pricing)) {
-    if (FLAT_KEYS[key]) {
-      base[FLAT_KEYS[key]] = value;
-      continue;
-    }
-    const m = key.match(TIER_RE);
-    if (m) {
-      const [, label, op] = m;
-      const field = TIER_FIELD[label!]!;
-      // Cloudflare AI Gateway entries don't carry tiered pricing (unsupported here). Fold the
-      // lower/default-context band (op "<") into the flat rate and drop the higher-context bands.
-      if (op!.startsWith("<")) base[field] = value;
-      continue;
-    }
-    warnings.push(`${id}: unmapped pricing key "${key}"`);
-  }
-  return { ...base };
-}
-
-// Derive reasoning_options from a docs schema.input by walking every named property.
-function deriveReasoningOptions(schemaInput: unknown): Array<Record<string, unknown>> {
-  let hasToggle = false;
-  let effortValues: string[] | undefined;
-
-  const visit = (node: unknown) => {
-    if (Array.isArray(node)) {
-      node.forEach(visit);
-      return;
-    }
-    if (!node || typeof node !== "object") return;
-    for (const [key, value] of Object.entries(node as Record<string, unknown>)) {
-      if (key === "properties" && value && typeof value === "object") {
-        for (const [propName, propSchema] of Object.entries(value as Record<string, any>)) {
-          if (propName === "enable_thinking" || propName === "thinking") hasToggle = true;
-          if (propName === "effort" || propName === "reasoning_effort") {
-            let enumVals: string[] | undefined = propSchema?.enum;
-            if (!enumVals) {
-              for (const branch of [...(propSchema?.anyOf ?? []), ...(propSchema?.oneOf ?? [])]) {
-                if (Array.isArray(branch?.enum)) enumVals = branch.enum;
-              }
-            }
-            if (enumVals) effortValues = enumVals;
-          }
-          visit(propSchema);
-        }
-      } else {
-        visit(value);
-      }
-    }
-  };
-  visit(schemaInput);
-
-  const opts: Array<Record<string, unknown>> = [];
-  if (hasToggle) opts.push({ type: "toggle" });
-  if (effortValues) opts.push({ type: "effort", values: effortValues });
-  return opts;
-}
-
-// Prepend curated leading `#` comment lines (a toggle/effort wire-path note, a source
-// citation, etc.) above the generated content. AGENTS.md requires one for every `toggle`
-// reasoning option since sync strips mid-file comments on every regeneration.
-function withNote(note: string[] | undefined, content: string): string {
-  if (!note || note.length === 0) return content;
-  return `${note.map((line) => `# ${line}`).join("\n")}\n\n${content}`;
-}
-
-// ---------------------------------------------------------------------------
-// base_model resolution
-// ---------------------------------------------------------------------------
-function labFileExists(id: string): boolean {
-  return existsSync(path.join(MODELS_ROOT, `${id}.toml`));
-}
-function autoResolveBase(catalogId: string): string | null {
-  if (labFileExists(catalogId)) return catalogId;
-  const dashed = catalogId.replace(/\./g, "-");
-  if (labFileExists(dashed)) return dashed;
-  return null;
-}
-
-// ---------------------------------------------------------------------------
-// Main
-// ---------------------------------------------------------------------------
-function walkToml(dir: string): string[] {
-  if (!existsSync(dir)) return [];
-  return readdirSync(dir).flatMap((e) => {
-    const p = path.join(dir, e);
-    return statSync(p).isDirectory() ? walkToml(p) : p.endsWith(".toml") ? [p] : [];
-  });
-}
-
-async function main() {
-  const check = process.argv.includes("--check");
-
-  const parsed = Curation.safeParse(Bun.TOML.parse(readFileSync(CURATION_PATH, "utf8")));
-  if (!parsed.success) {
-    console.error("Invalid curation.toml:", parsed.error.issues);
+if (check) {
+  if (result.files.length > 0) {
+    console.error(`--check: ${result.files.length} file(s) out of date`);
     process.exit(1);
   }
-  const curation = parsed.data;
-  const skip = new Set(curation.skip);
-  const errors: string[] = [];
-  const warnings: string[] = [];
-
-  const proxied = await loadProxied();
-  const proxiedTextGen = proxied.filter((m) => m.task === TEXT_GENERATION);
-
-  const wanted = new Map<string, string>(); // absolute path -> content
-
-  // Fetch per-model schemas up front for the proxied models we'll actually emit, with bounded
-  // concurrency so the catalog/docs endpoints don't rate-limit us.
-  const proxiedEmit = proxiedTextGen.filter((m) => !skip.has(m.model_id));
-  const schemaInputs = new Map<string, unknown>();
-  await mapLimit(proxiedEmit, 6, async (m) => {
-    schemaInputs.set(m.model_id, await loadCatalogSchemaInput(m.model_id));
-  });
-
-  // --- proxied models ---
-  for (const m of proxiedTextGen) {
-    const id: string = m.model_id;
-    if (skip.has(id)) continue;
-    const cur = curation.models[id] ?? {};
-    const base = cur.base_model ?? autoResolveBase(id);
-    if (!base) {
-      errors.push(`proxied ${id}: no lab file and no curation base_model (add to skip or map it)`);
-      continue;
-    }
-    // name/description are inherited from base_model (models.dev's canonical copy);
-    // the catalog only carries Cloudflare's own casing/marketing variants.
-    const model: Record<string, unknown> = { base_model: base };
-    if (cur.structured_output !== undefined) model.structured_output = cur.structured_output;
-    if (cur.interleaved !== undefined) model.interleaved = cur.interleaved;
-
-    // reasoning_options: only meaningful when the base actually reasons. The catalog schema
-    // advertises reasoning_effort for some non-reasoning models (gpt-4.1, gpt-4o) — schema
-    // acceptance is not capability, so gate on the base's reasoning flag. When the base does
-    // reason, prefer the per-model catalog schema, then curation, and fail loudly if neither
-    // supplies a shape (the schema requires reasoning_options whenever reasoning=true).
-    if (baseReasoning(base)) {
-      const derivedRo = deriveReasoningOptions(schemaInputs.get(id));
-      if (cur.reasoning_options !== undefined) model.reasoning_options = cur.reasoning_options;
-      else if (derivedRo.length > 0) model.reasoning_options = derivedRo;
-      else {
-        errors.push(
-          `proxied ${id}: base ${base} has reasoning=true but no reasoning_options ` +
-          `(catalog schema exposes none; add reasoning_options to curation.toml)`,
-        );
-        continue;
-      }
-    }
-
-    const cost = proxiedCost(m.pricing ?? {}, id, warnings);
-    if (Object.keys(cost).length === 0) errors.push(`proxied ${id}: catalog pricing empty`);
-    model.cost = cost;
-
-    // limit.output: the catalog's max_output_tokens is not a reliable ceiling — verified wrong
-    // against lab/first-party for gpt-5, gpt-5.5, and claude-haiku-4.5 (all understated by
-    // 4-8x). Only context_length has checked out, so that's all we auto-derive; output is
-    // either curated explicitly or left to inherit from base_model.
-    const limit: Record<string, number> = {};
-    if (m.context_length != null) limit.context = m.context_length;
-    if (cur.limit) Object.assign(limit, cur.limit); // curation overrides/adds (e.g. served output)
-    if (Object.keys(limit).length) model.limit = limit;
-
-    const npm = nativeNpm(id);
-    if (npm) model.provider = { npm };
-
-    wanted.set(path.join(MODELS_DIR, `${id}.toml`), withNote(cur.note, formatToml(model as any)));
-  }
-
-  // Guards: a curation model id that no longer appears in the live feed (warn only).
-  const liveIds = new Set<string>(proxiedTextGen.map((m) => m.model_id));
-  for (const id of Object.keys(curation.models)) {
-    if (!liveIds.has(id)) warnings.push(`curation id not in live feed: ${id}`);
-  }
-
-  if (errors.length > 0) {
-    console.error("Errors:\n" + errors.map((e) => `  - ${e}`).join("\n"));
-    process.exit(1);
-  }
-  for (const w of warnings) console.warn(`warning: ${w}`);
-
-  const existing = new Set(walkToml(MODELS_DIR));
-  const wantedPaths = new Set(wanted.keys());
-  const toRemove = [...existing].filter((p) => !wantedPaths.has(p));
-
-  if (check) {
-    let changed = 0;
-    for (const [p, content] of wanted) {
-      const cur = existing.has(p) ? readFileSync(p, "utf8") : undefined;
-      if (cur !== content) { console.error(`would change: ${path.relative(MODELS_DIR, p)}`); changed++; }
-    }
-    for (const p of toRemove) { console.error(`would remove: ${path.relative(MODELS_DIR, p)}`); changed++; }
-    if (changed > 0) { console.error(`--check: ${changed} file(s) out of date`); process.exit(1); }
-    console.log("--check: up to date");
-    return;
-  }
-
-  let changed = 0;
-  for (const [p, content] of wanted) {
-    const cur = existing.has(p) ? readFileSync(p, "utf8") : undefined;
-    if (cur !== content) { await Bun.write(p, content); changed++; }
-  }
-  for (const p of toRemove) { rmSync(p); changed++; }
-
-  console.log(
-    `cloudflare-ai-gateway: ${wanted.size} proxied model(s) ` +
-    `(${changed} written/removed, ${skip.size} skipped, ${warnings.length} warning(s)).`,
-  );
+  console.log("--check: up to date");
 }
-
-await main();
diff --git a/packages/core/src/sync/auto-merge.ts b/packages/core/src/sync/auto-merge.ts
index 5fb1934d6d3..244d2e277d3 100644
--- a/packages/core/src/sync/auto-merge.ts
+++ b/packages/core/src/sync/auto-merge.ts
@@ -55,6 +55,14 @@ export async function classifyAutoMerge(
   if (created + deleted > MAX_MODEL_CHURN) {
     reasons.push(`${created + deleted} models created or deleted (limit ${MAX_MODEL_CHURN})`);
   }
+  if (
+    models.some((change) =>
+      change.status === "deleted"
+      && change.path.startsWith("providers/cloudflare-ai-gateway/models/")
+    )
+  ) {
+    reasons.push("Cloudflare AI Gateway model deletions require manual review");
+  }
 
   const reasoningMetadata = async (path: string, loader: typeof load) => {
     const model = Bun.TOML.parse(await loader(path)) as Record<string, unknown>;
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 888773f3dd5..f08532afa14 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -9,6 +9,7 @@ import { ambient } from "./providers/ambient.js";
 import { anthropic } from "./providers/anthropic.js";
 import { baseten } from "./providers/baseten.js";
 import { chutes } from "./providers/chutes.js";
+import { cloudflareAiGateway } from "./providers/cloudflare-ai-gateway.js";
 import { cloudflareWorkersAi } from "./providers/cloudflare-workers-ai.js";
 import { cortecs } from "./providers/cortecs.js";
 import { crossmodel } from "./providers/crossmodel.js";
@@ -83,6 +84,8 @@ export interface SyncProvider<SourceModel> {
   preserveSymlinks?: boolean;
   preserveBaseModels?: boolean;
   preserveDescriptions?: boolean;
+  /** Replace existing leading comments with translateModel.header. */
+  authoritativeHeaders?: boolean;
   sameModel?(current: ExistingModel, desired: SyncedModel): boolean;
   missingNotice?(paths: string[]): string[];
   /**
@@ -104,9 +107,9 @@ export interface SyncProvider<SourceModel> {
     model: SyncedModel;
     metadata?: { id: string; model: SyncedMetadata };
     /**
-     * Leading comment block for the written file when it has none of its own
-     * (e.g. the wire-path header every toggle reasoning control requires). A
-     * header already present on the existing file always wins.
+     * Leading comment block for the written file (e.g. the wire-path header
+     * every toggle reasoning control requires). Existing headers win unless
+     * authoritativeHeaders is enabled.
      */
     header?: string;
   } | undefined;
@@ -129,6 +132,7 @@ export const providers: {
   anthropic: SyncProvider<any>;
   baseten: SyncProvider<any>;
   chutes: SyncProvider<any>;
+  "cloudflare-ai-gateway": SyncProvider<any>;
   "cloudflare-workers-ai": SyncProvider<any>;
   cortecs: SyncProvider<any>;
   crossmodel: SyncProvider<any>;
@@ -162,6 +166,7 @@ export const providers: {
   anthropic,
   baseten,
   chutes,
+  "cloudflare-ai-gateway": cloudflareAiGateway,
   "cloudflare-workers-ai": cloudflareWorkersAi,
   cortecs,
   crossmodel,
@@ -209,7 +214,7 @@ export const groups = {
     "openrouter",
     "vercel",
   ],
-  cloudflare: ["cloudflare-workers-ai"],
+  cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
   direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
@@ -235,7 +240,11 @@ export async function syncProvider<SourceModel>(
   const { models: existing, brokenSymlinks } = existingState;
   let { modelMetadata } = existingState;
   const sourceModels = provider.parseModels(await provider.fetchModels());
-  const desired = new Map<string, { model: z.infer<typeof SyncedAuthoredModel>; content: string }>();
+  const desired = new Map<string, {
+    model: z.infer<typeof SyncedAuthoredModel>;
+    content: string;
+    header: string;
+  }>();
   const caseNormalizedDesiredPaths = new Map<string, string>();
   const desiredMetadata = new Map<string, { model: z.infer<typeof ModelMetadata>; content: string }>();
   const skippedRemote: string[] = [];
@@ -325,9 +334,16 @@ export async function syncProvider<SourceModel>(
       throw parsed.error;
     }
 
+    const translatedHeader = translated.header === undefined
+      ? undefined
+      : leadingComments(translated.header);
+    const header = provider.authoritativeHeaders
+      ? translatedHeader ?? ""
+      : (existing.get(relativePath)?.header || translatedHeader) ?? "";
     desired.set(relativePath, {
       model: parsed.data,
-      content: ((existing.get(relativePath)?.header || translated.header) ?? "") + formatToml(parsed.data),
+      content: header + formatToml(parsed.data),
+      header,
     });
   }
 
@@ -336,7 +352,7 @@ export async function syncProvider<SourceModel>(
 
   const metadataDir = modelMetadataDir(provider.modelsDir);
   for (const [relativePath, file] of desiredMetadata) {
-    const filePath = path.join(metadataDir, relativePath);
+    const filePath = await safeWritePath(metadataDir, relativePath);
     const currentFile = Bun.file(filePath);
     const currentText = await currentFile.exists() ? await currentFile.text() : undefined;
     const current = currentText !== undefined
@@ -367,7 +383,7 @@ export async function syncProvider<SourceModel>(
         console.log(`Skipping metadata removal in new-only mode: ${relativePath}`);
         continue;
       }
-      const filePath = path.join(metadataDir, relativePath);
+      const filePath = await safeWritePath(metadataDir, relativePath);
       files.push({ status: "deleted", path: filePath });
       if (options.dryRun) {
         console.log(`Would remove metadata ${relativePath}`);
@@ -378,7 +394,7 @@ export async function syncProvider<SourceModel>(
   }
 
   for (const [relativePath, file] of desired) {
-    const filePath = path.join(provider.modelsDir, relativePath);
+    const filePath = await safeWritePath(provider.modelsDir, relativePath, true);
     const current = existing.get(relativePath);
 
     if (current === undefined) {
@@ -398,7 +414,12 @@ export async function syncProvider<SourceModel>(
       continue;
     }
 
-    if (!(provider.sameModel?.(current.authored, file.model) ?? sameModel(relativePath, current.authored, file.model))) {
+    const headerChanged = provider.authoritativeHeaders && current.header !== file.header;
+    if (
+      headerChanged
+      || !(provider.sameModel?.(current.authored, file.model)
+        ?? sameModel(relativePath, current.authored, file.model))
+    ) {
       if (options.newOnly) {
         unchanged++;
         continue;
@@ -431,7 +452,7 @@ export async function syncProvider<SourceModel>(
       continue;
     }
 
-    const filePath = path.join(provider.modelsDir, relativePath);
+    const filePath = await safeWritePath(provider.modelsDir, relativePath, true);
     files.push({ status: "deleted", path: filePath });
     if (options.dryRun) {
       console.log(`Would remove ${relativePath}`);
@@ -602,6 +623,31 @@ async function isSymlink(filePath: string) {
   }
 }
 
+async function safeWritePath(root: string, relativePath: string, allowLeafSymlink = false) {
+  const resolvedRoot = path.resolve(root);
+  const target = path.resolve(resolvedRoot, relativePath);
+  const relative = path.relative(resolvedRoot, target);
+  if (relative === "" || relative.startsWith(`..${path.sep}`) || path.isAbsolute(relative)) {
+    throw new Error(`Refusing to sync path outside ${root}: ${relativePath}`);
+  }
+  if (await isSymlink(resolvedRoot)) {
+    throw new Error(`Refusing to sync through symlink: ${resolvedRoot}`);
+  }
+
+  let current = resolvedRoot;
+  for (const segment of path.relative(resolvedRoot, path.dirname(target)).split(path.sep)) {
+    if (segment === "") continue;
+    current = path.join(current, segment);
+    if (await isSymlink(current)) {
+      throw new Error(`Refusing to sync through symlink: ${current}`);
+    }
+  }
+  if (!allowLeafSymlink && await isSymlink(target)) {
+    throw new Error(`Refusing to sync through symlink: ${target}`);
+  }
+  return target;
+}
+
 async function readModelMetadata(modelsDir: string) {
   const metadataDir = modelMetadataDir(modelsDir);
   const result: Record<string, Record<string, unknown>> = {};
@@ -829,7 +875,7 @@ async function writeReport(target: string, results: SyncResult[]) {
     }
   }
 
-  lines.push("", "This PR was created automatically by the daily model sync workflow.");
+  lines.push("", "This PR was created automatically by the model sync workflow.");
   await Bun.write(".sync/model-sync-report.md", `${lines.join("\n")}\n`);
 }
 
diff --git a/packages/core/src/sync/providers/cloudflare-ai-gateway.ts b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
new file mode 100644
index 00000000000..65c9499b7e3
--- /dev/null
+++ b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
@@ -0,0 +1,502 @@
+import { existsSync, readFileSync, readdirSync } from "node:fs";
+import path from "node:path";
+import { z } from "zod";
+
+import { ReasoningOption } from "../../schema.js";
+import type { ExistingModel, SyncProvider, SyncedBaseModel } from "../index.js";
+
+const API_BASE = "https://api.cloudflare.com/client/v4/accounts";
+const PROVIDER_DIR = path.join(
+  import.meta.dirname,
+  "..",
+  "..",
+  "..",
+  "..",
+  "..",
+  "providers",
+  "cloudflare-ai-gateway",
+);
+const MODELS_ROOT = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+const CURATION_PATH = path.join(PROVIDER_DIR, "curation.toml");
+const TEXT_GENERATION = "Text Generation";
+const REQUEST_TIMEOUT_MS = 30_000;
+const MAX_CATALOG_PAGES = 1_000;
+const MAX_BACKOFF_DELAY_MS = 8_000;
+const MAX_RETRY_DELAY_MS = 60_000;
+
+const NATIVE_NPM: Record<string, string> = {
+  anthropic: "@ai-sdk/anthropic",
+  openai: "@ai-sdk/openai",
+};
+
+const CatalogEntry = z.object({
+  model_id: z.string().refine(isSafeModelID, "model_id must be a safe relative provider/model path"),
+  task: z.string(),
+  context_length: z.number().int().positive().nullish(),
+  pricing: z.record(z.number().nonnegative()).nullish(),
+}).passthrough();
+const CatalogModel = CatalogEntry.extend({
+  task: z.literal(TEXT_GENERATION),
+  context_length: z.number().int().positive().nullish(),
+  pricing: z.record(z.number().nonnegative()),
+});
+
+const CloudflareResponse = z.object({
+  success: z.literal(true),
+  result: z.array(CatalogEntry),
+  result_info: z.object({
+    page: z.number().int().positive(),
+    per_page: z.number().int().positive(),
+    total_count: z.number().int().nonnegative(),
+    total_pages: z.number().int().positive().optional(),
+    count: z.number().int().nonnegative().optional(),
+  }).passthrough(),
+}).passthrough();
+
+const SourceModel = z.object({
+  catalog: CatalogModel,
+  schemaInput: z.unknown().optional(),
+});
+
+const CuratedModel = z.object({
+  base_model: z.string().refine(isSafeModelID, "base_model must be a safe relative provider/model path").optional(),
+  structured_output: z.boolean().optional(),
+  reasoning_options: z.array(ReasoningOption).optional(),
+  limit: z.object({
+    context: z.number().optional(),
+    input: z.number().optional(),
+    output: z.number().optional(),
+  }).strict().optional(),
+  interleaved: z.union([
+    z.literal(true),
+    z.object({ field: z.enum(["reasoning_content", "reasoning_details"]) }).strict(),
+  ]).optional(),
+  note: z.array(z.string().refine((value) => !/[\r\n]/.test(value))).optional(),
+}).strict();
+
+const Curation = z.object({
+  skip: z.array(z.string()).default([]),
+  models: z.record(CuratedModel).default({}),
+}).strict();
+
+type CatalogModel = z.infer<typeof CatalogModel>;
+type CatalogEntry = z.infer<typeof CatalogEntry>;
+type SourceModel = z.infer<typeof SourceModel>;
+type CuratedModel = z.infer<typeof CuratedModel>;
+
+const curation = Curation.parse(Bun.TOML.parse(readFileSync(CURATION_PATH, "utf8")));
+const skippedModels = new Set(curation.skip);
+
+export const cloudflareAiGateway = {
+  id: "cloudflare-ai-gateway",
+  name: "Cloudflare AI Gateway",
+  modelsDir: "providers/cloudflare-ai-gateway/models",
+  preserveDescriptions: false,
+  authoritativeHeaders: true,
+  async fetchModels() {
+    const catalog = CatalogEntry.array().parse(await loadCatalog());
+    const catalogIDs = new Set(catalog.map((model) => model.model_id));
+    if (catalogIDs.size !== catalog.length) {
+      throw new Error("Cloudflare AI Gateway catalog returned duplicate model IDs");
+    }
+    const textModels = catalog.filter((model) => model.task === TEXT_GENERATION);
+    if (textModels.length === 0) {
+      throw new Error("Cloudflare AI Gateway catalog returned no Text Generation models");
+    }
+
+    const emittedModels = textModels.filter(
+      (model) => !model.model_id.startsWith("@cf/") && !skippedModels.has(model.model_id),
+    ).map((model) => CatalogModel.parse(model));
+    if (emittedModels.length === 0) {
+      throw new Error("Cloudflare AI Gateway catalog returned no eligible proxied models");
+    }
+    const sources = await mapLimit(emittedModels, 6, async (model) => ({
+      catalog: model,
+      schemaInput: await loadCatalogSchemaInput(model.model_id),
+    }));
+
+    const liveIDs = new Set(textModels.map((model) => model.model_id));
+    for (const id of Object.keys(curation.models)) {
+      if (!liveIDs.has(id)) console.warn(`warning: curation id not in live feed: ${id}`);
+    }
+
+    return sources;
+  },
+  parseModels(raw) {
+    return SourceModel.array().parse(raw);
+  },
+  translateModel(source, context) {
+    const id = source.catalog.model_id;
+    const curated = curation.models[id] ?? {};
+    return {
+      id,
+      model: buildCloudflareAiGatewayModel(
+        source.catalog,
+        source.schemaInput,
+        curated,
+        context.authored(id),
+      ),
+      header: noteHeader(curated.note),
+    };
+  },
+} satisfies SyncProvider<SourceModel>;
+
+export function buildCloudflareAiGatewayModel(
+  catalog: CatalogModel,
+  schemaInput: unknown,
+  curated: CuratedModel = {},
+  existing?: ExistingModel,
+): SyncedBaseModel {
+  const id = catalog.model_id;
+  const baseModel = curated.base_model ?? resolveBaseModel(id);
+  if (baseModel === undefined) {
+    throw new Error(`${id}: no lab file and no curated base_model; add it to skip or map it`);
+  }
+
+  const model: SyncedBaseModel = { base_model: baseModel };
+  if (curated.structured_output !== undefined) {
+    model.structured_output = curated.structured_output;
+  }
+  if (curated.interleaved !== undefined) model.interleaved = curated.interleaved;
+
+  if (baseReasoning(baseModel)) {
+    const derived = deriveReasoningOptions(schemaInput);
+    const reasoningOptions = curated.reasoning_options ?? (derived.length > 0 ? derived : undefined);
+    if (reasoningOptions === undefined) {
+      throw new Error(
+        `${id}: base ${baseModel} reasons but the catalog schema and curation provide no reasoning_options`,
+      );
+    }
+    model.reasoning_options = reasoningOptions;
+  }
+
+  model.cost = proxiedCost(catalog.pricing, id);
+
+  const limit = {
+    ...(catalog.context_length == null && existing?.limit?.context === undefined
+      ? {}
+      : { context: catalog.context_length ?? existing?.limit?.context }),
+    ...curated.limit,
+  };
+  if (Object.keys(limit).length > 0) model.limit = limit;
+
+  const npm = NATIVE_NPM[id.split("/")[0]!];
+  if (npm !== undefined) model.provider = { npm };
+  return model;
+}
+
+export function deriveReasoningOptions(
+  schemaInput: unknown,
+): NonNullable<SyncedBaseModel["reasoning_options"]> {
+  let hasToggle = false;
+  let effortValues: Array<"none" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max" | "default">
+    | undefined;
+
+  const EffortValues = z.array(z.enum([
+    "none",
+    "minimal",
+    "low",
+    "medium",
+    "high",
+    "xhigh",
+    "max",
+    "default",
+  ]));
+
+  const visit = (node: unknown) => {
+    if (Array.isArray(node)) {
+      node.forEach(visit);
+      return;
+    }
+    if (node === null || typeof node !== "object") return;
+
+    for (const [key, value] of Object.entries(node)) {
+      if (key !== "properties" || value === null || typeof value !== "object") {
+        visit(value);
+        continue;
+      }
+
+      for (const [property, rawSchema] of Object.entries(value)) {
+        const propertySchema = rawSchema as Record<string, unknown>;
+        if (property === "enable_thinking" || property === "thinking") hasToggle = true;
+        if (property === "effort" || property === "reasoning_effort") {
+          const candidates = [propertySchema, ...arrayValue(propertySchema.anyOf), ...arrayValue(propertySchema.oneOf)];
+          for (const candidate of candidates) {
+            const parsed = EffortValues.safeParse(candidate.enum);
+            if (parsed.success) effortValues = parsed.data;
+          }
+        }
+        visit(rawSchema);
+      }
+    }
+  };
+  visit(schemaInput);
+
+  const options: NonNullable<SyncedBaseModel["reasoning_options"]> = [];
+  if (hasToggle) options.push({ type: "toggle" });
+  if (effortValues !== undefined) options.push({ type: "effort", values: effortValues });
+  return options;
+}
+
+function arrayValue(value: unknown): Array<Record<string, unknown>> {
+  return Array.isArray(value)
+    ? value.filter((item): item is Record<string, unknown> => item !== null && typeof item === "object")
+    : [];
+}
+
+async function loadCatalog() {
+  const fixtureDir = process.env.CF_AIG_FIXTURE_DIR;
+  if (fixtureDir !== undefined) return loadFixtureRows(fixtureDir, "catalog");
+
+  const { accountID, token } = credentials();
+  const pages: Array<z.infer<typeof CloudflareResponse>> = [];
+  for (let page = 1; page <= MAX_CATALOG_PAGES; page++) {
+    const url = new URL(`${API_BASE}/${accountID}/ai/catalog/models`);
+    url.searchParams.set("page", String(page));
+    url.searchParams.set("per_page", "50");
+    const { response, json } = await fetchJsonWithRetry(url, { headers: { Authorization: `Bearer ${token}` } });
+    if (!response.ok) {
+      throw new Error(`Cloudflare AI Gateway catalog request failed: ${response.status} ${response.statusText}`);
+    }
+    const body = CloudflareResponse.parse(json);
+    pages.push(body);
+    const expectedPages = catalogPageCount(pages[0]!);
+    if (expectedPages > MAX_CATALOG_PAGES) throw new Error(`Invalid Cloudflare AI Gateway catalog page count: ${expectedPages}`);
+    if (page === expectedPages) return validateCatalogPages(pages, "Cloudflare AI Gateway catalog");
+  }
+  throw new Error("Cloudflare AI Gateway catalog exceeded the pagination safety limit");
+}
+
+async function loadCatalogSchemaInput(id: string): Promise<unknown> {
+  const fixtureDir = process.env.CF_AIG_FIXTURE_DIR;
+  if (fixtureDir !== undefined) {
+    const file = path.join(fixtureDir, "schema", `${id.replaceAll("/", "_")}.json`);
+    if (!existsSync(file)) return undefined;
+    return z.object({
+      success: z.literal(true),
+      result: z.object({ schema: z.object({ input: z.unknown().optional() }).passthrough() }).passthrough(),
+    }).passthrough().parse(JSON.parse(readFileSync(file, "utf8"))).result.schema.input;
+  }
+
+  const { accountID, token } = credentials();
+  const { response, json } = await fetchJsonWithRetry(
+    `${API_BASE}/${accountID}/ai/catalog/models/${id.split("/").map(encodeURIComponent).join("/")}/schema`,
+    { headers: { Authorization: `Bearer ${token}` } },
+  );
+  if (response.status === 404) return undefined;
+  if (!response.ok) {
+    throw new Error(`Cloudflare AI Gateway schema request failed for ${id}: ${response.status} ${response.statusText}`);
+  }
+  return z.object({
+    success: z.literal(true),
+    result: z.object({ schema: z.object({ input: z.unknown().optional() }).passthrough() }).passthrough(),
+  }).passthrough().parse(json).result.schema.input;
+}
+
+function credentials() {
+  const token = process.env.CLOUDFLARE_API_TOKEN
+    ?? process.env.CLOUDFLARE_PRODUCTION_API_TOKEN;
+  const accountID = process.env.CLOUDFLARE_ACCOUNT_ID
+    ?? process.env.CLOUDFLARE_PRODUCTION_ACCOUNT_ID_AI_GATEWAY_SANDBOX;
+  if (!token || !accountID) {
+    throw new Error(
+      "Cloudflare AI Gateway sync requires Cloudflare API token and account ID credentials",
+    );
+  }
+  return { accountID, token };
+}
+
+function loadFixtureRows(directory: string, prefix: string): unknown[] {
+  const pages = readdirSync(directory)
+    .filter((name) => name.startsWith(prefix) && name.endsWith(".json"))
+    .sort()
+    .map((file) => CloudflareResponse.parse(JSON.parse(readFileSync(path.join(directory, file), "utf8"))))
+    .sort((a, b) => a.result_info.page - b.result_info.page);
+  return validateCatalogPages(pages, `Cloudflare AI Gateway fixtures in ${directory}`);
+}
+
+function catalogPageCount(page: z.infer<typeof CloudflareResponse>) {
+  const calculated = Math.max(1, Math.ceil(page.result_info.total_count / page.result_info.per_page));
+  if (page.result_info.total_pages !== undefined && page.result_info.total_pages !== calculated) {
+    throw new Error("Invalid Cloudflare AI Gateway catalog pagination: total_pages does not match total_count");
+  }
+  return page.result_info.total_pages ?? calculated;
+}
+
+function validateCatalogPages(pages: Array<z.infer<typeof CloudflareResponse>>, source: string) {
+  const first = pages[0];
+  if (first === undefined) throw new Error(`${source} contained no pages`);
+  const expectedPages = catalogPageCount(first);
+  if (pages.length !== expectedPages) {
+    throw new Error(`${source} contains ${pages.length}/${expectedPages} pages`);
+  }
+
+  const models: CatalogEntry[] = [];
+  const ids = new Set<string>();
+  for (const [index, page] of pages.entries()) {
+    if (page.result_info.page !== index + 1) {
+      throw new Error(`${source} expected page ${index + 1}, got ${page.result_info.page}`);
+    }
+    if (
+      page.result_info.total_count !== first.result_info.total_count
+      || page.result_info.per_page !== first.result_info.per_page
+      || catalogPageCount(page) !== expectedPages
+    ) {
+      throw new Error(`${source} pagination changed while reading pages`);
+    }
+    if (page.result_info.count !== undefined && page.result_info.count !== page.result.length) {
+      throw new Error(`${source} result count mismatch on page ${page.result_info.page}`);
+    }
+    if (page.result.length > page.result_info.per_page) {
+      throw new Error(`${source} page ${page.result_info.page} exceeds per_page`);
+    }
+    for (const model of page.result) {
+      if (ids.has(model.model_id)) throw new Error(`${source} returned duplicate model ID ${model.model_id}`);
+      ids.add(model.model_id);
+      models.push(model);
+    }
+  }
+  if (models.length !== first.result_info.total_count) {
+    throw new Error(`${source} pagination ended at ${models.length}/${first.result_info.total_count}`);
+  }
+  return models;
+}
+
+async function fetchJsonWithRetry(
+  url: string | URL,
+  init: RequestInit,
+  attempts = 5,
+): Promise<{ response: Response; json?: unknown }> {
+  let delay = 500;
+  for (let attempt = 1; attempt <= attempts; attempt++) {
+    try {
+      const timeout = AbortSignal.timeout(REQUEST_TIMEOUT_MS);
+      const response = await fetch(url, {
+        ...init,
+        signal: init.signal ? AbortSignal.any([init.signal, timeout]) : timeout,
+      });
+      if (!response.ok) {
+        if ((response.status === 429 || response.status >= 500) && attempt < attempts) {
+          await response.body?.cancel();
+          await waitForRetry(retryDelay(response, delay), init.signal);
+          delay = Math.min(delay * 2, MAX_BACKOFF_DELAY_MS);
+          continue;
+        }
+        await response.body?.cancel();
+        return { response };
+      }
+      try {
+        return { response, json: await response.json() };
+      } catch (error) {
+        if (attempt === attempts) throw error;
+      }
+    } catch (error) {
+      if (init.signal?.aborted || attempt === attempts) throw error;
+    }
+    await waitForRetry(delay, init.signal);
+    delay = Math.min(delay * 2, MAX_BACKOFF_DELAY_MS);
+  }
+  throw new Error("Cloudflare AI Gateway request exhausted retries");
+}
+
+function retryDelay(response: Response, fallback: number) {
+  const value = response.headers.get("retry-after");
+  if (value === null) return fallback;
+  const seconds = Number(value);
+  if (Number.isFinite(seconds) && seconds >= 0) return Math.min(seconds * 1_000, MAX_RETRY_DELAY_MS);
+  const timestamp = Date.parse(value);
+  return Number.isFinite(timestamp)
+    ? Math.min(Math.max(timestamp - Date.now(), 0), MAX_RETRY_DELAY_MS)
+    : fallback;
+}
+
+function waitForRetry(delay: number, signal: AbortSignal | null | undefined) {
+  return new Promise<void>((resolve, reject) => {
+    if (signal?.aborted) return reject(signal.reason);
+    const onAbort = () => {
+      clearTimeout(timer);
+      reject(signal.reason);
+    };
+    const timer = setTimeout(() => {
+      signal?.removeEventListener("abort", onAbort);
+      resolve();
+    }, delay);
+    signal?.addEventListener("abort", onAbort, { once: true });
+  });
+}
+
+async function mapLimit<T, R>(items: T[], limit: number, transform: (item: T) => Promise<R>) {
+  const results = new Array<R>(items.length);
+  let next = 0;
+  await Promise.all(Array.from({ length: Math.min(limit, items.length) }, async () => {
+    while (next < items.length) {
+      const index = next++;
+      results[index] = await transform(items[index]!);
+    }
+  }));
+  return results;
+}
+
+const FLAT_PRICING_KEYS: Record<string, "input" | "output" | "cache_read" | "cache_write"> = {
+  "Input tokens (per 1M)": "input",
+  "Output tokens (per 1M)": "output",
+  "Cached input tokens (per 1M)": "cache_read",
+  "Cache creation tokens (per 1M)": "cache_write",
+};
+const TIERED_PRICING_KEY = /^(Input|Output|Cached input)\s*(<=?|>=?)\s*(\d+)k\s*\(per 1M\)$/;
+const TIERED_PRICING_FIELDS = {
+  Input: "input",
+  Output: "output",
+  "Cached input": "cache_read",
+} as const;
+
+function proxiedCost(pricing: Record<string, number>, id: string): NonNullable<SyncedBaseModel["cost"]> {
+  const cost: NonNullable<SyncedBaseModel["cost"]> = {};
+  for (const [key, value] of Object.entries(pricing)) {
+    const flatField = FLAT_PRICING_KEYS[key];
+    if (flatField !== undefined) {
+      cost[flatField] = value;
+      continue;
+    }
+    const tier = TIERED_PRICING_KEY.exec(key);
+    if (tier !== null) {
+      const field = TIERED_PRICING_FIELDS[tier[1] as keyof typeof TIERED_PRICING_FIELDS];
+      if (tier[2]!.startsWith("<")) cost[field] = value;
+      continue;
+    }
+    throw new Error(`${id}: unmapped pricing key "${key}"`);
+  }
+  if (cost.input === undefined || cost.output === undefined) {
+    throw new Error(`${id}: catalog pricing must include input and output rates`);
+  }
+  return cost;
+}
+
+function isSafeModelID(id: string) {
+  if (path.isAbsolute(id) || id.includes("\\")) return false;
+  const segments = id.split("/");
+  return segments.length >= 2
+    && segments.every((segment) => /^[A-Za-z0-9@._-]+$/.test(segment) && segment !== "." && segment !== "..");
+}
+
+function resolveBaseModel(id: string) {
+  if (labFileExists(id)) return id;
+  const dashed = id.replaceAll(".", "-");
+  return labFileExists(dashed) ? dashed : undefined;
+}
+
+function labFileExists(id: string) {
+  return existsSync(path.join(MODELS_ROOT, `${id}.toml`));
+}
+
+function baseReasoning(id: string) {
+  const file = path.join(MODELS_ROOT, `${id}.toml`);
+  return existsSync(file) && z.object({ reasoning: z.boolean().optional() }).passthrough()
+    .parse(Bun.TOML.parse(readFileSync(file, "utf8"))).reasoning === true;
+}
+
+function noteHeader(note: string[] | undefined) {
+  return note === undefined || note.length === 0
+    ? undefined
+    : `${note.map((line) => `# ${line}`).join("\n")}\n`;
+}
diff --git a/packages/core/test/auto-merge.test.ts b/packages/core/test/auto-merge.test.ts
index aee07e799ff..1b4a44e3fd5 100644
--- a/packages/core/test/auto-merge.test.ts
+++ b/packages/core/test/auto-merge.test.ts
@@ -31,6 +31,18 @@ test("requires manual review for bulk additions", async () => {
   expect(decision.reasons).toContain("11 models created (limit 10)");
 });
 
+test("requires manual review for Cloudflare AI Gateway deletions", async () => {
+  const decision = await classifyAutoMerge([
+    {
+      status: "deleted",
+      path: "providers/cloudflare-ai-gateway/models/openai/gpt-4.1.toml",
+    },
+  ]);
+
+  expect(decision.safe).toBe(false);
+  expect(decision.reasons).toContain("Cloudflare AI Gateway model deletions require manual review");
+});
+
 test("requires manual review for added reasoning provider models", async () => {
   const withoutOptions = await classifyAutoMerge(
     [{ status: "created", path: "providers/test/models/reasoner.toml" }],
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
new file mode 100644
index 00000000000..3cf59f404c3
--- /dev/null
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -0,0 +1,433 @@
+import { expect, test } from "bun:test";
+import { mkdir, mkdtemp, readFile, rm, symlink, writeFile } from "node:fs/promises";
+import path from "node:path";
+
+import { syncProvider } from "../src/sync/index.js";
+import {
+  buildCloudflareAiGatewayModel,
+  cloudflareAiGateway,
+  deriveReasoningOptions,
+} from "../src/sync/providers/cloudflare-ai-gateway.js";
+
+test("builds Cloudflare AI Gateway overrides from catalog metadata", () => {
+  const model = buildCloudflareAiGatewayModel(
+    {
+      model_id: "openai/gpt-5.4",
+      task: "Text Generation",
+      context_length: 1_050_000,
+      pricing: {
+        "Input <= 200k (per 1M)": 2.5,
+        "Input > 200k (per 1M)": 5,
+        "Output tokens (per 1M)": 15,
+        "Cached input tokens (per 1M)": 0.25,
+      },
+    },
+    undefined,
+    {
+      reasoning_options: [{ type: "effort", values: ["none", "low", "medium", "high", "xhigh"] }],
+    },
+  );
+
+  expect(model).toEqual({
+    base_model: "openai/gpt-5.4",
+    reasoning_options: [{ type: "effort", values: ["none", "low", "medium", "high", "xhigh"] }],
+    cost: { input: 2.5, output: 15, cache_read: 0.25 },
+    limit: { context: 1_050_000 },
+    provider: { npm: "@ai-sdk/openai" },
+  });
+});
+
+test("derives nested Cloudflare reasoning controls", () => {
+  expect(deriveReasoningOptions({
+    properties: {
+      thinking: { type: "boolean" },
+      reasoning: {
+        properties: {
+          effort: {
+            anyOf: [{ enum: ["low", "medium", "high"] }],
+          },
+        },
+      },
+    },
+  })).toEqual([
+    { type: "toggle" },
+    { type: "effort", values: ["low", "medium", "high"] },
+  ]);
+});
+
+test("ignores advertised reasoning controls for non-reasoning base models", () => {
+  const model = buildCloudflareAiGatewayModel(
+    {
+      model_id: "openai/gpt-4.1",
+      task: "Text Generation",
+      context_length: 1_047_576,
+      pricing: {
+        "Input tokens (per 1M)": 2,
+        "Output tokens (per 1M)": 8,
+      },
+    },
+    {
+      properties: {
+        reasoning_effort: { enum: ["low", "medium", "high"] },
+      },
+    },
+  );
+
+  expect(model.reasoning_options).toBeUndefined();
+});
+
+test("fails closed on unknown pricing fields", () => {
+  expect(() => buildCloudflareAiGatewayModel({
+    model_id: "openai/gpt-4.1",
+    task: "Text Generation",
+    context_length: 1_047_576,
+    pricing: {
+      "Input tokens (per 1M)": 2,
+      "Output tokens (per 1M)": 8,
+      "New billing unit": 1,
+    },
+  }, undefined)).toThrow('unmapped pricing key "New billing unit"');
+});
+
+test("fails closed when Cloudflare pagination is incomplete", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  let page = 0;
+  globalThis.fetch = async () => new Response(JSON.stringify(catalogPage(
+    page++ === 0
+      ? [{
+        model_id: "openai/gpt-4.1",
+        task: "Text Generation",
+        context_length: 1_047_576,
+        pricing: {
+          "Input tokens (per 1M)": 2,
+          "Output tokens (per 1M)": 8,
+        },
+      }]
+      : [],
+    { page, total_count: 2 },
+  )));
+
+  try {
+    await expect(cloudflareAiGateway.fetchModels()).rejects.toThrow("pagination ended at 1/2");
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
+test("rejects unsafe catalog model paths", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  globalThis.fetch = async () => new Response(JSON.stringify(catalogPage([{
+      model_id: "../providers/openai/models/gpt-4.1",
+      task: "Text Generation",
+      context_length: 1_047_576,
+      pricing: {
+        "Input tokens (per 1M)": 2,
+        "Output tokens (per 1M)": 8,
+      },
+    }])));
+
+  try {
+    await expect(cloudflareAiGateway.fetchModels()).rejects.toThrow("safe relative provider/model path");
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
+test("rejects a catalog with no eligible proxied models", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  globalThis.fetch = async () => new Response(JSON.stringify(catalogPage([{
+    model_id: "@cf/meta/llama-3.1-8b-instruct",
+    task: "Text Generation",
+  }])));
+
+  try {
+    await expect(cloudflareAiGateway.fetchModels()).rejects.toThrow("no eligible proxied models");
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
+test("validates Cloudflare page metadata", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  globalThis.fetch = async () => new Response(JSON.stringify(catalogPage([], { page: 2, total_count: 0 })));
+
+  try {
+    await expect(cloudflareAiGateway.fetchModels()).rejects.toThrow("expected page 1, got 2");
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
+test("retries transient Cloudflare responses", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  let catalogRequests = 0;
+  globalThis.fetch = async (input) => {
+    if (String(input).endsWith("/schema")) return new Response(null, { status: 404 });
+    catalogRequests++;
+    if (catalogRequests === 1) return new Response(null, { status: 503, headers: { "retry-after": "0" } });
+    return new Response(JSON.stringify(catalogPage([{
+      model_id: "openai/gpt-4.1",
+      task: "Text Generation",
+      context_length: 1_047_576,
+      pricing: {
+        "Input tokens (per 1M)": 2,
+        "Output tokens (per 1M)": 8,
+      },
+    }])));
+  };
+
+  try {
+    expect(await cloudflareAiGateway.fetchModels()).toHaveLength(1);
+    expect(catalogRequests).toBe(2);
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
+test("replaces a stale generated base-model mapping", async () => {
+  const providersDir = path.join(import.meta.dirname, "..", "..", "..", "providers");
+  const providerDir = await mkdtemp(path.join(providersDir, ".base-model-sync-"));
+  const modelsDir = path.join(providerDir, "models");
+  await mkdir(modelsDir);
+  const file = path.join(modelsDir, "model.toml");
+  await writeFile(file, 'base_model = "anthropic/claude-opus-4-6"\n');
+
+  try {
+    const provider = {
+      id: "base-model-test",
+      name: "Base-model test",
+      modelsDir,
+      async fetchModels() {
+        return [{ id: "model" }];
+      },
+      parseModels(raw: unknown) {
+        return raw as Array<{ id: string }>;
+      },
+      translateModel(model: { id: string }) {
+        return { id: model.id, model: { base_model: "openai/gpt-4.1" } };
+      },
+    };
+    await syncProvider(provider);
+    expect(await readFile(file, "utf8")).toContain('base_model = "openai/gpt-4.1"');
+  } finally {
+    await rm(providerDir, { recursive: true, force: true });
+  }
+});
+
+test("reconciles authoritative generated headers", async () => {
+  const providersDir = path.join(import.meta.dirname, "..", "..", "..", "providers");
+  const providerDir = await mkdtemp(path.join(providersDir, ".cloudflare-ai-gateway-sync-"));
+  const modelsDir = path.join(providerDir, "models");
+  await mkdir(modelsDir);
+  const file = path.join(modelsDir, "gpt-4.1.toml");
+  await writeFile(file, "# Old note\n\nbase_model = \"openai/gpt-4.1\"\n");
+
+  try {
+    const provider = {
+      id: "cloudflare-ai-gateway-test",
+      name: "Cloudflare AI Gateway test",
+      modelsDir,
+      authoritativeHeaders: true,
+      async fetchModels() {
+        return [{ id: "gpt-4.1" }];
+      },
+      parseModels() {
+        return [{ id: "gpt-4.1" }];
+      },
+      translateModel(model) {
+        return {
+          id: model.id,
+          model: { base_model: "openai/gpt-4.1" },
+          header: "# New note\n\n",
+        };
+      },
+    };
+    const result = await syncProvider(provider);
+
+    expect(result.updated).toBe(1);
+    expect(await readFile(file, "utf8")).toStartWith("# New note\nbase_model");
+    expect((await syncProvider(provider)).updated).toBe(0);
+  } finally {
+    await rm(providerDir, { recursive: true, force: true });
+  }
+});
+
+test("refuses to write through a symlinked model directory", async () => {
+  const providersDir = path.join(import.meta.dirname, "..", "..", "..", "providers");
+  const providerDir = await mkdtemp(path.join(providersDir, ".sync-symlink-"));
+  const outsideDir = await mkdtemp(path.join(providersDir, ".sync-outside-"));
+  const modelsDir = path.join(providerDir, "models");
+  await mkdir(modelsDir);
+  await symlink(outsideDir, path.join(modelsDir, "linked"));
+
+  try {
+    const provider = {
+      id: "symlink-test",
+      name: "Symlink test",
+      modelsDir,
+      async fetchModels() {
+        return [{ id: "linked/model" }];
+      },
+      parseModels(raw: unknown) {
+        return raw as Array<{ id: string }>;
+      },
+      translateModel(model: { id: string }) {
+        return { id: model.id, model: { base_model: "openai/gpt-4.1" } };
+      },
+    };
+    await expect(syncProvider(provider)).rejects.toThrow("Refusing to sync through symlink");
+    expect(await Bun.file(path.join(outsideDir, "model.toml")).exists()).toBe(false);
+  } finally {
+    await rm(providerDir, { recursive: true, force: true });
+    await rm(outsideDir, { recursive: true, force: true });
+  }
+});
+
+test("refuses a symlinked models root", async () => {
+  const providersDir = path.join(import.meta.dirname, "..", "..", "..", "providers");
+  const providerDir = await mkdtemp(path.join(providersDir, ".sync-root-"));
+  const outsideDir = await mkdtemp(path.join(providersDir, ".sync-root-outside-"));
+  const modelsDir = path.join(providerDir, "models");
+  await symlink(outsideDir, modelsDir);
+
+  try {
+    const provider = testSyncProvider(modelsDir, "model");
+    await expect(syncProvider(provider)).rejects.toThrow("Refusing to sync through symlink");
+    expect(await Bun.file(path.join(outsideDir, "model.toml")).exists()).toBe(false);
+  } finally {
+    await rm(providerDir, { recursive: true, force: true });
+    await rm(outsideDir, { recursive: true, force: true });
+  }
+});
+
+test("refuses a symlinked metadata file", async () => {
+  const providersDir = path.join(import.meta.dirname, "..", "..", "..", "providers");
+  const providerDir = await mkdtemp(path.join(providersDir, ".sync-metadata-"));
+  const outsideDir = await mkdtemp(path.join(providersDir, ".sync-metadata-outside-"));
+  const modelsDir = path.join(providerDir, "models");
+  const namespace = `sync-symlink-${path.basename(providerDir).replaceAll(/[^a-z0-9-]/g, "")}`;
+  const metadataDir = path.join(providersDir, "..", "models", namespace);
+  const outsideFile = path.join(outsideDir, "model.toml");
+  await mkdir(modelsDir);
+  await mkdir(metadataDir);
+  await writeFile(outsideFile, "sentinel\n");
+  await symlink(outsideFile, path.join(metadataDir, "model.toml"));
+
+  try {
+    const provider = {
+      ...testSyncProvider(modelsDir, "provider-model"),
+      metadataNamespace: namespace,
+      translateModel(model: { id: string }) {
+        return {
+          id: model.id,
+          model: {
+            name: "Provider symlink test",
+            description: "Provider model used to test safe sync paths",
+            release_date: "2026-01-01",
+            last_updated: "2026-01-01",
+            attachment: false,
+            reasoning: false,
+            tool_call: false,
+            open_weights: false,
+            modalities: { input: ["text"], output: ["text"] },
+            limit: { context: 1_000, output: 100 },
+            cost: { input: 1, output: 2 },
+          },
+          metadata: {
+            id: `${namespace}/model`,
+            model: {
+              name: "Symlink test",
+              description: "Metadata used to test safe sync paths",
+              release_date: "2026-01-01",
+              last_updated: "2026-01-01",
+              attachment: false,
+              reasoning: false,
+              tool_call: false,
+              open_weights: false,
+              modalities: { input: ["text"], output: ["text"] },
+              limit: { context: 1_000, output: 100 },
+            },
+          },
+        };
+      },
+    };
+    await expect(syncProvider(provider)).rejects.toThrow("Refusing to sync through symlink");
+    expect(await readFile(outsideFile, "utf8")).toBe("sentinel\n");
+  } finally {
+    await rm(providerDir, { recursive: true, force: true });
+    await rm(outsideDir, { recursive: true, force: true });
+    await rm(metadataDir, { recursive: true, force: true });
+  }
+});
+
+function testSyncProvider(modelsDir: string, id: string) {
+  return {
+    id: "symlink-test",
+    name: "Symlink test",
+    modelsDir,
+    async fetchModels() {
+      return [{ id }];
+    },
+    parseModels(raw: unknown) {
+      return raw as Array<{ id: string }>;
+    },
+    translateModel(model: { id: string }) {
+      return { id: model.id, model: { base_model: "openai/gpt-4.1" } };
+    },
+  };
+}
+
+function catalogPage(
+  result: Array<Record<string, unknown>>,
+  resultInfo: Partial<{ page: number; per_page: number; total_count: number; total_pages: number }> = {},
+) {
+  const page = resultInfo.page ?? 1;
+  const perPage = resultInfo.per_page ?? 50;
+  const totalCount = resultInfo.total_count ?? result.length;
+  return {
+    success: true,
+    result,
+    result_info: {
+      page,
+      per_page: perPage,
+      count: result.length,
+      total_count: totalCount,
+      total_pages: resultInfo.total_pages ?? Math.max(1, Math.ceil(totalCount / perPage)),
+    },
+  };
+}
+
+function restoreEnv(name: string, value: string | undefined) {
+  if (value === undefined) delete process.env[name];
+  else process.env[name] = value;
+}
diff --git a/providers/cloudflare-ai-gateway/README.md b/providers/cloudflare-ai-gateway/README.md
index c12b859de6a..9abd759714c 100644
--- a/providers/cloudflare-ai-gateway/README.md
+++ b/providers/cloudflare-ai-gateway/README.md
@@ -18,10 +18,13 @@ One command regenerates every model TOML:
 CLOUDFLARE_API_TOKEN=xxx CLOUDFLARE_ACCOUNT_ID=xxx bun run cloudflare-ai-gateway:generate
 ```
 
+The sync also accepts `CLOUDFLARE_PRODUCTION_API_TOKEN` and
+`CLOUDFLARE_PRODUCTION_ACCOUNT_ID_AI_GATEWAY_SANDBOX` directly.
+
 It reads two live Cloudflare sources plus one local curation file:
 
 - **Proxied catalog** — `GET /accounts/{id}/ai/catalog/models`. The source of truth: one
-  canonical (dotted) `model_id` per model, description, context/output limits, and pricing.
+  canonical (dotted) `model_id` per model, description, context limits, and pricing.
   This is why ids look like `anthropic/claude-haiku-4.5`, not `claude-haiku-4-5`.
 - **Per-model catalog schema** — `GET /accounts/{id}/ai/catalog/models/{id}/schema`. Used to
   derive `reasoning_options` for providers whose schema is OpenAI-compatible (xAI, Alibaba,
@@ -44,7 +47,7 @@ bun run cloudflare-ai-gateway:generate --check
 ```
 
 Exits non-zero if the committed TOMLs are out of date with the live catalog + curation.
-Used in CI.
+The same shared provider runs hourly; use `--check` for local drift checks.
 
 ### Offline / fixtures
 
diff --git a/sync.md b/sync.md
index a9c59774309..2aa0ac34651 100644
--- a/sync.md
+++ b/sync.md
@@ -11,6 +11,7 @@ The grouped sync targets are available for local convenience, but CI syncs each
 - `bun models:sync aggregators` syncs every provider in the `aggregators` group.
 - `bun models:sync openrouter` syncs only OpenRouter.
 - `bun models:sync cloudflare-workers-ai` syncs only Cloudflare Workers AI.
+- `bun models:sync cloudflare-ai-gateway` syncs only Cloudflare AI Gateway's proxied catalog.
 - `bun models:sync cloudflare` syncs the Cloudflare sync group.
 - `bun models:sync direct` syncs every provider in the `direct` group.
 - `bun models:sync google` syncs only Google.
@@ -202,7 +203,16 @@ Cloudflare Workers AI is implemented in `packages/core/src/sync/providers/cloudf
 - Use a dedicated token scoped to Workers AI read access so sync automation does not share deploy credentials.
 - The endpoint is parsed as Cloudflare's OpenRouter-like Workers AI metadata.
 - Model IDs map directly to TOML paths under `providers/cloudflare-workers-ai/models`.
-- This sync target does not manage `providers/cloudflare-ai-gateway`, because the AI Gateway `/compat/models` endpoint does not support `format=openrouter` and does not provide enough model metadata for authoritative catalog sync.
+- This target only manages Workers AI; the separate Cloudflare AI Gateway target handles proxied third-party models.
+
+## Cloudflare AI Gateway Notes
+
+- Cloudflare AI Gateway is implemented in `packages/core/src/sync/providers/cloudflare-ai-gateway.ts`.
+- Source endpoints: `GET /accounts/{id}/ai/catalog/models` for model availability, context limits, and pricing, plus `GET /accounts/{id}/ai/catalog/models/{model}/schema` for reasoning controls exposed by compatible schemas.
+- Required auth: `CLOUDFLARE_ACCOUNT_ID` and `CLOUDFLARE_API_TOKEN`, or the production-token aliases documented in the provider README for local runs. The hourly workflow uses the canonical secret names.
+- The sync manages proxied third-party text-generation models only. Workers AI `@cf/...` models remain under `providers/cloudflare-workers-ai`.
+- `providers/cloudflare-ai-gateway/curation.toml` supplies base-model mappings, live-tested reasoning controls, structured-output support, limit overrides, and intentional skips that the catalog cannot express authoritatively.
+- New catalog entries without canonical lab metadata or required reasoning controls fail closed instead of generating incomplete TOMLs.
 
 ## Google Notes
 

From cc1c9478119d76ab6961bfb0edd2ec9d7c4d93c8 Mon Sep 17 00:00:00 2001
From: Zain Hasan <zain_has@hotmail.com>
Date: Sat, 29 Aug 2026 16:49:01 -0400
Subject: [PATCH 0217/2012] [Together AI] add glm5.3 (#5773)

---
 providers/togetherai/models/zai-org/GLM-5.3.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/togetherai/models/zai-org/GLM-5.3.toml

diff --git a/providers/togetherai/models/zai-org/GLM-5.3.toml b/providers/togetherai/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..3ce03e88dba
--- /dev/null
+++ b/providers/togetherai/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 262_144

From 2c2542b86db70f01acf232c29fa59803e6ec57ac Mon Sep 17 00:00:00 2001
From: Tejush <tejush1998@gmail.com>
Date: Sun, 30 Aug 2026 02:19:12 +0530
Subject: [PATCH 0218/2012] chore(sync): update CrofAI model catalog (#5829)

* sync crof providers

* Add GLM 5.3 for Crof provider

---------

Co-authored-by: tejush <mac@MacBook-Air.local>
---
 providers/crof/models/glm-5.3.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/crof/models/glm-5.3.toml

diff --git a/providers/crof/models/glm-5.3.toml b/providers/crof/models/glm-5.3.toml
new file mode 100644
index 00000000000..776c386dbf1
--- /dev/null
+++ b/providers/crof/models/glm-5.3.toml
@@ -0,0 +1,16 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.40
+output = 1.40
+cache_read = 0.06
+
+[limit]
+output = 131_072
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
\ No newline at end of file

From 3448fbcbe1a50b3f8bf79fedf9dd7c14506ea1d8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 21:21:47 +0000
Subject: [PATCH 0219/2012] chore(sync): update OpenRouter model catalog
 (#5833)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 159a52f39fa..d06dd2cd8d1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.045
-output = 0.09
-cache_read = 0.009
+input = 0.065
+output = 0.18
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cfffc22c9a8..b883d1f622b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.54897
-output = 1.09794
-cache_read = 0.045748
+input = 0.538878
+output = 1.077756
+cache_read = 0.044907
 
 [limit]
 context = 1_048_576

From 165e30577c65f235b357beda1561b09abcc5fe6e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 22:22:25 +0000
Subject: [PATCH 0220/2012] chore(sync): update OpenRouter model catalog
 (#5835)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 98144b8c68a..a0f20ed7621 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08204
-output = 0.16408
-cache_read = 0.016408
+input = 0.08176
+output = 0.16352
+cache_read = 0.016352
 
 [limit]
 context = 1_048_576

From 6c11fb9b719ec46850f5163677cb3a89e4351bc3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 22:22:34 +0000
Subject: [PATCH 0221/2012] chore(sync): update Kilo model catalog (#5834)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/google/gemma-4-31b-it.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/google/gemma-4-31b-it.toml b/providers/kilo/models/google/gemma-4-31b-it.toml
index 8d55ceb50b5..ccdbcb026b4 100644
--- a/providers/kilo/models/google/gemma-4-31b-it.toml
+++ b/providers/kilo/models/google/gemma-4-31b-it.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.07
-output = 0.35
-cache_read = 0.1
+input = 0.09
+output = 0.34
+cache_read = 0.05
 
 [limit]
 output = 16_384

From b5211ad8c1a95851843535684a07723a6e00eca6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 23:21:07 +0000
Subject: [PATCH 0222/2012] chore(sync): update OpenRouter model catalog
 (#5837)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/thinkingmachines/inkling.toml   | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 4 ++--
 4 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a0f20ed7621..c249bf42fef 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08176
-output = 0.16352
-cache_read = 0.016352
+input = 0.08148
+output = 0.16296
+cache_read = 0.016296
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b883d1f622b..1539a689311 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.538878
-output = 1.077756
-cache_read = 0.044907
+input = 0.528786
+output = 1.057572
+cache_read = 0.044066
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/thinkingmachines/inkling.toml b/providers/openrouter/models/thinkingmachines/inkling.toml
index 513922adcef..1a64ecd26f3 100644
--- a/providers/openrouter/models/thinkingmachines/inkling.toml
+++ b/providers/openrouter/models/thinkingmachines/inkling.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "max"]
 
 [cost]
-input = 0.95
+input = 1
 output = 4.05
-cache_read = 0.16
+cache_read = 0.17
 
 [limit]
-output = 262_144
+output = 471_859
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 64bd01fa1ab..364a0a34436 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,8 +21,8 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.1
-cache_read = 0.007
+output = 0.14
+cache_read = 0.01
 
 [limit]
 context = 1_310_720

From 8ce109f0393a8b84c89a5fc1130aff5e53ec26ce Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 23:21:20 +0000
Subject: [PATCH 0223/2012] chore(sync): update Kilo model catalog (#5838)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/thinkingmachines/inkling.toml           | 2 +-
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/thinkingmachines/inkling.toml b/providers/kilo/models/thinkingmachines/inkling.toml
index c46d7a4ad64..a367cecf27a 100644
--- a/providers/kilo/models/thinkingmachines/inkling.toml
+++ b/providers/kilo/models/thinkingmachines/inkling.toml
@@ -13,4 +13,4 @@ cache_read = 0.16
 
 [limit]
 context = 524_288
-output = 262_144
+output = 471_859
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 92950df32ad..25dc6e6dd90 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,8 +16,8 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.1
-cache_read = 0.007
+output = 0.14
+cache_read = 0.01
 
 [limit]
 context = 1_048_576

From 77ea61235c7864b3c7fc20ec21c7121a765fd419 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 01:03:34 +0000
Subject: [PATCH 0224/2012] chore(sync): update Kilo model catalog (#5840)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v3.2.toml           | 2 +-
 providers/kilo/models/meta-llama/llama-4-maverick.toml      | 4 ++--
 providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml | 2 +-
 providers/kilo/models/tencent/hy3.toml                      | 6 +++---
 providers/kilo/models/z-ai/glm-5.1.toml                     | 3 +--
 5 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v3.2.toml b/providers/kilo/models/deepseek/deepseek-v3.2.toml
index a29137a7cc0..1af0a0ca64c 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2.toml
@@ -12,4 +12,4 @@ cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 65_536
+output = 147_456
diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 3db8291511b..5c1df37c375 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -15,8 +15,8 @@ input = 0.2
 output = 0.696
 
 [limit]
-context = 1_048_576
-output = 16_384
+context = 128_000
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 8247cd4ca01..8071bb6d3c0 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.78
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.1.toml b/providers/kilo/models/z-ai/glm-5.1.toml
index 75a48589c32..882eaea90db 100644
--- a/providers/kilo/models/z-ai/glm-5.1.toml
+++ b/providers/kilo/models/z-ai/glm-5.1.toml
@@ -11,5 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 202_752
-output = 182_476
+output = 128_000

From 3e9d44222f44189b764667496d1c7e7fe631c021 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 01:03:41 +0000
Subject: [PATCH 0225/2012] chore(sync): update OpenRouter model catalog
 (#5839)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v3.2.toml   | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/meta-llama/llama-4-maverick.toml    | 4 ++--
 .../models/qwen/qwen3-next-80b-a3b-instruct.toml          | 5 ++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.1.toml             | 8 ++++----
 7 files changed, 21 insertions(+), 22 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v3.2.toml b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
index 80cd9bc29e1..cf2b742ab8d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v3.2.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
@@ -7,10 +7,10 @@ description = "DeepSeek chat model for instruction following, coding, and analys
 type = "toggle"
 
 [cost]
-input = 0.269
-output = 0.4
-cache_read = 0.1345
+input = 0.26
+output = 0.38
+cache_read = 0.13
 
 [limit]
 context = 163_840
-output = 65_536
+output = 147_456
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c249bf42fef..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08148
-output = 0.16296
-cache_read = 0.016296
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1539a689311..bee7b01d544 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.528786
-output = 1.057572
-cache_read = 0.044066
+input = 0.508428
+output = 1.016856
+cache_read = 0.042369
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index de6e5240ff9..f4b36c39886 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -13,11 +13,11 @@ open_weights = true
 
 [cost]
 input = 0.2
-output = 0.8
+output = 0.696
 
 [limit]
 context = 1_048_576
-output = 16_384
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 31bb83cc8bb..64deac24f44 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -2,10 +2,9 @@ base_model = "alibaba/qwen3-next-80b-a3b-instruct"
 structured_output = true
 
 [cost]
-input = 0.1
+input = 0.09
 output = 1.1
-cache_read = 0.07
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.1.toml b/providers/openrouter/models/z-ai/glm-5.1.toml
index 23d665940d6..5833f33e6e0 100644
--- a/providers/openrouter/models/z-ai/glm-5.1.toml
+++ b/providers/openrouter/models/z-ai/glm-5.1.toml
@@ -9,10 +9,10 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 1.26
-output = 3.96
-cache_read = 0.234
+input = 0.966
+output = 3.036
+cache_read = 0.1794
 
 [limit]
 context = 204_800
-output = 182_476
+output = 128_000

From 4ee7d4f92269141cdd2ecea2a7158e91aa475b1e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 02:33:24 +0000
Subject: [PATCH 0226/2012] chore(sync): update Kilo model catalog (#5841)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-4-maverick.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 5c1df37c375..3db8291511b 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -15,8 +15,8 @@ input = 0.2
 output = 0.696
 
 [limit]
-context = 128_000
-output = 115_200
+context = 1_048_576
+output = 16_384
 
 [modalities]
 input = ["text", "image"]

From 0843a537ff3e4d34e37c8ef0c8a44b672b371732 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 02:33:44 +0000
Subject: [PATCH 0227/2012] chore(sync): update OpenRouter model catalog
 (#5842)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/meta-llama/llama-4-maverick.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index f4b36c39886..de6e5240ff9 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -13,11 +13,11 @@ open_weights = true
 
 [cost]
 input = 0.2
-output = 0.696
+output = 0.8
 
 [limit]
 context = 1_048_576
-output = 115_200
+output = 16_384
 
 [modalities]
 input = ["text", "image"]

From c01c7dcbc08b3f9418c105a63d610b343ba35d61 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Sun, 30 Aug 2026 11:20:33 +0800
Subject: [PATCH 0228/2012] feat(go): end Hy3 promo

---
 providers/opencode-go/models/hy3.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/opencode-go/models/hy3.toml b/providers/opencode-go/models/hy3.toml
index 5b3b95a6c48..be97effd28f 100644
--- a/providers/opencode-go/models/hy3.toml
+++ b/providers/opencode-go/models/hy3.toml
@@ -1,11 +1,10 @@
 base_model = "tencent/hy3"
-name = "Hy3 (8x usage)"
 
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0175
-output = 0.0725
-cache_read = 0.004375
+input = 0.14
+output = 0.58
+cache_read = 0.035

From 04551ad7559ad91733411418a58e579bce39aebc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 03:29:19 +0000
Subject: [PATCH 0229/2012] chore(sync): update Kilo model catalog (#5843)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v3.2.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v3.2.toml b/providers/kilo/models/deepseek/deepseek-v3.2.toml
index 1af0a0ca64c..a29137a7cc0 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2.toml
@@ -12,4 +12,4 @@ cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536

From 682156da73d5c4d0ef77d895d44631fc5477ac1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 03:29:43 +0000
Subject: [PATCH 0230/2012] chore(sync): update OpenRouter model catalog
 (#5844)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v3.2.toml   | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v3.2.toml b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
index cf2b742ab8d..80cd9bc29e1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v3.2.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v3.2.toml
@@ -7,10 +7,10 @@ description = "DeepSeek chat model for instruction following, coding, and analys
 type = "toggle"
 
 [cost]
-input = 0.26
-output = 0.38
-cache_read = 0.13
+input = 0.269
+output = 0.4
+cache_read = 0.1345
 
 [limit]
 context = 163_840
-output = 147_456
+output = 65_536
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..3081f90ea05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.08064
+output = 0.16128
+cache_read = 0.016128
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index bee7b01d544..5bbf73f7cb8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.508428
-output = 1.016856
-cache_read = 0.042369
+input = 0.498336
+output = 0.996672
+cache_read = 0.041528
 
 [limit]
 context = 1_048_576

From 5340d31c23d2bc0a33051226a3122e5ab003e0c3 Mon Sep 17 00:00:00 2001
From: Ryan Scott <ryan@ryan-scott.me>
Date: Sun, 30 Aug 2026 13:31:02 +1000
Subject: [PATCH 0231/2012] feat(fireworks-ai): add GLM 5.3 Flash serverless
 model (#5836)

Expose Fireworks' multimodal GLM 5.3 Flash route with its reasoning controls and serverless pricing.
---
 .../fireworks/models/glm-5p3-flash.toml       | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
new file mode 100644
index 00000000000..9c92c6284f7
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
@@ -0,0 +1,19 @@
+# Pricing: https://fireworks.ai/models/fireworks/glm-5p3-flash (accessed 2026-08-29)
+# Released on Fireworks Serverless API 2026-08-26.
+# Thinking-only model: reasoning cannot be disabled. Use the same two effective
+# tiers as Fireworks GLM 5.3: low/medium collapse to high; max/xhigh select max.
+# https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-08-29)
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.50
+cache_read = 0.029

From be340379cf3e4af4e6336abd6df306189aeb723d Mon Sep 17 00:00:00 2001
From: Renaud Cerrato <renaud.cerrato@swissmarketplace.group>
Date: Sun, 30 Aug 2026 05:31:37 +0200
Subject: [PATCH 0232/2012] feat(ollama-cloud): add glm-5.3 model (#5798)

---
 providers/ollama-cloud/models/glm-5.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/ollama-cloud/models/glm-5.3.toml

diff --git a/providers/ollama-cloud/models/glm-5.3.toml b/providers/ollama-cloud/models/glm-5.3.toml
new file mode 100644
index 00000000000..d16b23a5f89
--- /dev/null
+++ b/providers/ollama-cloud/models/glm-5.3.toml
@@ -0,0 +1,15 @@
+# Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
+# or `reasoning.effort` with low|high|max (GLM-5.3 always reasons; thinking
+# cannot be disabled, so there is no toggle).
+# Sources: https://docs.ollama.com/capabilities/thinking, https://docs.ollama.com/openai
+# API-verified context: https://ollama.com/api/show (glm_dsa_moe.context_length = 1048576)
+base_model = "zhipuai/glm-5.3"
+open_weights = true
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+context = 1_048_576
\ No newline at end of file

From edaffcf7bdefab6f823f64c95fe9581a7f484367 Mon Sep 17 00:00:00 2001
From: Fenil Modi <fenilmodi4820@gmail.com>
Date: Sun, 30 Aug 2026 09:01:48 +0530
Subject: [PATCH 0233/2012] Add ai& GLM-5.3 and Qwen3.8 27B models (#5806)

* Add ai& GLM-5.3 and Qwen3.8 27B model catalog entries.

* Document ai& live reasoning_effort probes for GLM-5.3 and Qwen3.8 27B.

---------

Co-authored-by: fenilmodi00 <fenilmodi00@gmail.com>
---
 providers/aiand/models/qwen/qwen3.8-27b.toml | 22 ++++++++++++++++++++
 providers/aiand/models/zai-org/glm-5.3.toml  | 21 +++++++++++++++++++
 2 files changed, 43 insertions(+)
 create mode 100644 providers/aiand/models/qwen/qwen3.8-27b.toml
 create mode 100644 providers/aiand/models/zai-org/glm-5.3.toml

diff --git a/providers/aiand/models/qwen/qwen3.8-27b.toml b/providers/aiand/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..30b2de2637a
--- /dev/null
+++ b/providers/aiand/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,22 @@
+# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
+# 2026-08-29). GET /v1/models reports context_window = 262144, input = $0.40,
+# output = $3.00, cached_input = $0.20 per 1M tokens.
+# Modalities: text, image, video, and document (PDF) supported per catalog
+# capabilities (vision|video|document).
+# reasoning_effort verified by live probe on 2026-08-29: accepts none, low,
+# medium, xhigh (200); rejects minimal, high, max with 400 ("Supported
+# values: none, low, medium, xhigh"). effort=none returns no
+# reasoning_content (off).
+base_model = "alibaba/qwen3.8-27b"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[cost]
+input = 0.4
+output = 3
+cache_read = 0.2
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
diff --git a/providers/aiand/models/zai-org/glm-5.3.toml b/providers/aiand/models/zai-org/glm-5.3.toml
new file mode 100644
index 00000000000..9bdd6c737ee
--- /dev/null
+++ b/providers/aiand/models/zai-org/glm-5.3.toml
@@ -0,0 +1,21 @@
+# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
+# 2026-08-29). GET /v1/models reports context_window = 1048576 (shared base
+# model uses 1_000_000), so the exact figure is overridden here.
+# Lab / peer relays (zhipuai, zai, OpenRouter, Kilo) use always-on effort
+# low|high|max. ai&'s deployment differs: live reasoning_effort probe on
+# 2026-08-29 accepts none|low|xhigh|max (200); rejects minimal|medium|high
+# with 400 ("Supported values: none, low, xhigh, max"). effort=none returns
+# no reasoning_content (off); xhigh|max return reasoning_content.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "xhigh", "max"]
+
+[cost]
+input = 1
+output = 4
+cache_read = 0.3
+
+[limit]
+context = 1_048_576

From c66acb2fb46f3316499ed349833ec7e5f1c6ebd3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 22:32:05 -0500
Subject: [PATCH 0234/2012] chore(sync): update NanoGPT model catalog (#5808)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/Gemma-4-26B-A4B-MeroMero.toml      |  2 +-
 .../Gemma-4-26B-A4B-MeroMero:thinking.toml    |  2 +-
 .../models/Gemma-4-31B-MeroMero-v2.toml       |  2 +-
 .../Gemma-4-31B-MeroMero-v2:thinking.toml     |  2 +-
 .../abliterated-model-large.toml              |  4 +--
 .../abliterated-model.toml                    |  0
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 .../nano-gpt/models/dmind/dmind-1-mini.toml   | 24 ------------------
 .../gemma-4-26b-a4b-it-chimerax.toml          |  4 +--
 .../gemma-4-26b-a4b-it-darksoul.toml          |  4 +--
 .../gemma-4-26b-a4b-it-luminous.toml          |  4 +--
 .../gemma-4-26b-a4b-it-moonlight.toml         |  4 +--
 .../gemma-4-26b-a4b-it-musica.toml            |  4 +--
 .../gemma-4-26b-a4b-it-opusdistill.toml       |  4 +--
 .../gemma-4-26b-a4b-it-shadowsiren.toml       |  4 +--
 .../gemma-4-26b-a4b-uncensored.toml           |  6 ++---
 .../gemma-4-26b-a4b-uncensored:thinking.toml  |  4 +--
 .../{google => }/gemma-4-31b-it-darkidol.toml |  4 +--
 .../{google => }/gemma-4-31b-it-fabled.toml   |  4 +--
 .../{google => }/gemma-4-31b-it-garnet.toml   |  4 +--
 .../{google => }/gemma-4-31b-it-gembrain.toml |  4 +--
 .../{google => }/gemma-4-31b-it-gemsicle.toml |  7 +++---
 .../{google => }/gemma-4-31b-it-isometry.toml |  4 +--
 .../{google => }/gemma-4-31b-it-novelist.toml |  4 +--
 providers/nano-gpt/models/glm-4-air.toml      | 24 ------------------
 providers/nano-gpt/models/glm-4-airx.toml     | 24 ------------------
 providers/nano-gpt/models/glm-4-flash.toml    | 24 ------------------
 providers/nano-gpt/models/glm-4-plus.toml     | 24 ------------------
 providers/nano-gpt/models/glm-4.toml          | 24 ------------------
 .../nano-gpt/models/glm-zero-preview.toml     | 24 ------------------
 .../models/hunyuan-turbos-20250226.toml       | 24 ------------------
 .../inflection/inflection-3-productivity.toml | 24 ------------------
 .../nano-gpt/models/jamba-large-1.6.toml      | 24 ------------------
 .../nano-gpt/models/jamba-large-1.7.toml      | 24 ------------------
 providers/nano-gpt/models/jamba-large.toml    | 24 ------------------
 providers/nano-gpt/models/jamba-mini-1.6.toml | 24 ------------------
 providers/nano-gpt/models/jamba-mini-1.7.toml | 24 ------------------
 providers/nano-gpt/models/jamba-mini.toml     | 24 ------------------
 .../models/learnlm-1.5-pro-experimental.toml  | 24 ------------------
 .../nano-gpt/models/moonshotai/kimi-k3.toml   |  6 ++---
 .../models/moonshotai/kimi-latest.toml        |  6 ++---
 .../models/ornith-ai/ornith-1.5-35b-a3b.toml  |  2 +-
 .../ornith-1.5-35b-a3b:thinking.toml          |  2 +-
 .../models/ornith-ai/ornith-1.5-397b.toml     | 25 -------------------
 .../ornith-ai/ornith-1.5-397b:thinking.toml   | 25 -------------------
 .../models/ornith-ai/ornith-1.5-9b.toml       |  2 +-
 .../ornith-ai/ornith-1.5-9b:thinking.toml     |  2 +-
 .../qwen/qwen3.6-35b-a3b-uncensored.toml      |  2 +-
 .../qwen3.6-35b-a3b-uncensored:thinking.toml  |  2 +-
 .../models/qwen/qwen3.8-27b-fable.toml        |  2 +-
 .../models/qwen/qwen3.8-27b-obliterated.toml  |  2 +-
 .../qwen3.8-27b-obliterated:thinking.toml     |  2 +-
 .../models/qwen/qwen3.8-27b-uncensored.toml   |  2 +-
 .../qwen/qwen3.8-27b-uncensored:thinking.toml |  2 +-
 providers/nano-gpt/models/qwen3.8-27b.toml    |  4 +--
 .../nano-gpt/models/qwen3.8-27b:thinking.toml |  4 +--
 providers/nano-gpt/models/step-r1-v-mini.toml | 24 ------------------
 .../models/{zai-org => z-ai}/GLM-4.5-Air.toml |  1 -
 .../GLM-4.5-Air:thinking.toml                 |  1 -
 .../{zai-org => z-ai}/GLM-4.5:thinking.toml   |  1 -
 .../{zai-org => z-ai}/GLM-4.6-turbo.toml      |  3 +--
 .../GLM-4.6-turbo:thinking.toml               |  3 +--
 .../models/{zai-org => z-ai}/glm-4.5.toml     |  1 -
 .../{zai-org => z-ai}/glm-4.6-original.toml   |  7 ++----
 .../glm-4.6v-flash-original.toml              |  3 +--
 .../{zai-org => z-ai}/glm-4.6v-original.toml  |  3 +--
 .../models/{zai-org => z-ai}/glm-4.6v.toml    |  1 -
 .../glm-4.7-flash-original.toml               |  7 ++----
 .../glm-4.7-flash-original:thinking.toml      |  3 +--
 .../{zai-org => z-ai}/glm-4.7-flash.toml      |  4 +--
 .../glm-4.7-flash:thinking.toml               |  1 -
 .../{zai-org => z-ai}/glm-4.7-original.toml   |  7 ++----
 .../glm-4.7-original:thinking.toml            |  3 +--
 .../models/{zai-org => z-ai}/glm-4.7.toml     |  4 +--
 .../{zai-org => z-ai}/glm-4.7:thinking.toml   |  1 -
 .../{zai-org => z-ai}/glm-5-original.toml     |  7 ++----
 .../glm-5-original:thinking.toml              |  3 +--
 .../models/{zai-org => z-ai}/glm-5.1.toml     |  0
 .../{zai-org => z-ai}/glm-5.1:thinking.toml   |  0
 .../models/{zai-org => z-ai}/glm-5.2.toml     |  0
 .../{zai-org => z-ai}/glm-5.2:thinking.toml   |  0
 .../models/z-ai/glm-5.3-flash-uncensored.toml |  6 ++---
 .../models/{zai-org => z-ai}/glm-5.3.toml     |  2 +-
 .../{zai-org => z-ai}/glm-5.3:thinking.toml   |  2 +-
 .../models/{zai-org => z-ai}/glm-5.toml       |  4 +--
 .../{zai-org => z-ai}/glm-5:thinking.toml     |  0
 .../models/{zai-org => z-ai}/glm-latest.toml  |  8 ++----
 87 files changed, 90 insertions(+), 583 deletions(-)
 rename providers/nano-gpt/models/{ => abliteration-ai}/abliterated-model-large.toml (76%)
 rename providers/nano-gpt/models/{ => abliteration-ai}/abliterated-model.toml (100%)
 delete mode 100644 providers/nano-gpt/models/dmind/dmind-1-mini.toml
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-chimerax.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-darksoul.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-luminous.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-moonlight.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-musica.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-opusdistill.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-it-shadowsiren.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-uncensored.toml (58%)
 rename providers/nano-gpt/models/{google => }/gemma-4-26b-a4b-uncensored:thinking.toml (90%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-darkidol.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-fabled.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-garnet.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-gembrain.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-gemsicle.toml (81%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-isometry.toml (89%)
 rename providers/nano-gpt/models/{google => }/gemma-4-31b-it-novelist.toml (89%)
 delete mode 100644 providers/nano-gpt/models/glm-4-air.toml
 delete mode 100644 providers/nano-gpt/models/glm-4-airx.toml
 delete mode 100644 providers/nano-gpt/models/glm-4-flash.toml
 delete mode 100644 providers/nano-gpt/models/glm-4-plus.toml
 delete mode 100644 providers/nano-gpt/models/glm-4.toml
 delete mode 100644 providers/nano-gpt/models/glm-zero-preview.toml
 delete mode 100644 providers/nano-gpt/models/hunyuan-turbos-20250226.toml
 delete mode 100644 providers/nano-gpt/models/inflection/inflection-3-productivity.toml
 delete mode 100644 providers/nano-gpt/models/jamba-large-1.6.toml
 delete mode 100644 providers/nano-gpt/models/jamba-large-1.7.toml
 delete mode 100644 providers/nano-gpt/models/jamba-large.toml
 delete mode 100644 providers/nano-gpt/models/jamba-mini-1.6.toml
 delete mode 100644 providers/nano-gpt/models/jamba-mini-1.7.toml
 delete mode 100644 providers/nano-gpt/models/jamba-mini.toml
 delete mode 100644 providers/nano-gpt/models/learnlm-1.5-pro-experimental.toml
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-397b.toml
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-397b:thinking.toml
 delete mode 100644 providers/nano-gpt/models/step-r1-v-mini.toml
 rename providers/nano-gpt/models/{zai-org => z-ai}/GLM-4.5-Air.toml (87%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/GLM-4.5-Air:thinking.toml (88%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/GLM-4.5:thinking.toml (89%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/GLM-4.6-turbo.toml (73%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/GLM-4.6-turbo:thinking.toml (74%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.5.toml (89%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.6-original.toml (68%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.6v-flash-original.toml (60%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.6v-original.toml (52%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.6v.toml (90%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-flash-original.toml (65%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-flash-original:thinking.toml (72%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-flash.toml (82%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-flash:thinking.toml (88%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-original.toml (65%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7-original:thinking.toml (76%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7.toml (82%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-4.7:thinking.toml (88%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5-original.toml (68%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5-original:thinking.toml (76%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.1.toml (100%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.1:thinking.toml (100%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.2.toml (100%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.2:thinking.toml (100%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.3.toml (88%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.3:thinking.toml (84%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5.toml (82%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-5:thinking.toml (100%)
 rename providers/nano-gpt/models/{zai-org => z-ai}/glm-latest.toml (63%)

diff --git a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
index 1c6b8a983e9..1c6aa9516ee 100644
--- a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
+++ b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
@@ -1,7 +1,7 @@
 name = "Gemma 4 26B A4B MeroMero"
 description = "Gemma 4 26B A4B MeroMero is an NVFP4 multimodal mixture-of-experts fine-tune for emotive dialogue, relationship scenes, creative writing, and roleplay."
 family = "gemma"
-release_date = "2026-08-26"
+release_date = "2026-07-29"
 last_updated = "2026-08-26"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
index b02e770fbc4..1e636658f07 100644
--- a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
+++ b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
@@ -1,7 +1,7 @@
 name = "Gemma 4 26B A4B MeroMero Thinking"
 description = "Gemma 4 26B A4B MeroMero with thinking enabled for more deliberate emotive dialogue, relationship scenes, creative writing, and multimodal roleplay."
 family = "gemma"
-release_date = "2026-08-26"
+release_date = "2026-07-29"
 last_updated = "2026-08-26"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
index 9b5c3938098..fd2ec3d29f6 100644
--- a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
+++ b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
@@ -1,7 +1,7 @@
 name = "Gemma 4 31B MeroMero v2"
 description = "Gemma 4 31B MeroMero v2 is a LoRA finetune for emotive dialogue, relationship scenes, creative writing, and multimodal roleplay."
 family = "gemma"
-release_date = "2026-08-23"
+release_date = "2026-07-29"
 last_updated = "2026-08-23"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
index 378a06ef51c..d7df31bd6ff 100644
--- a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
+++ b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
@@ -1,7 +1,7 @@
 name = "Gemma 4 31B MeroMero v2 Thinking"
 description = "Gemma 4 31B MeroMero v2 with thinking enabled for more deliberate emotive dialogue, relationship scenes, creative writing, and multimodal roleplay."
 family = "gemma"
-release_date = "2026-08-24"
+release_date = "2026-07-29"
 last_updated = "2026-08-24"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/abliterated-model-large.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
similarity index 76%
rename from providers/nano-gpt/models/abliterated-model-large.toml
rename to providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
index 4e36a511c70..71a9a723881 100644
--- a/providers/nano-gpt/models/abliterated-model-large.toml
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
@@ -1,10 +1,10 @@
 name = "Abliterated Model Large"
-description = "Abliteration.ai's large text reasoning model is derived from GLM-5.2 and supports structured output, automatic prompt caching, and a one-million-token context window."
+description = "Abliteration.ai's large text reasoning model is derived from GLM-5.2 and supports native tool calling, structured output, automatic prompt caching, and a one-million-token context window."
 release_date = "2026-08-27"
 last_updated = "2026-08-27"
 attachment = false
 reasoning = true
-tool_call = false
+tool_call = true
 structured_output = true
 open_weights = false
 
diff --git a/providers/nano-gpt/models/abliterated-model.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model.toml
similarity index 100%
rename from providers/nano-gpt/models/abliterated-model.toml
rename to providers/nano-gpt/models/abliteration-ai/abliterated-model.toml
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/dmind/dmind-1-mini.toml b/providers/nano-gpt/models/dmind/dmind-1-mini.toml
deleted file mode 100644
index 3938a18079c..00000000000
--- a/providers/nano-gpt/models/dmind/dmind-1-mini.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "DMind-1-Mini"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-01"
-last_updated = "2025-06-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.2
-output = 0.4
-cache_read = 0.1
-
-[limit]
-context = 32_768
-input = 32_768
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
index f2c876ddf7b..0869768f349 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-chimerax.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
@@ -1,8 +1,8 @@
 name = "Chimera X"
 description = "Chimera X is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
index cf1dbd3b699..7b9406d64bb 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-darksoul.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
@@ -1,8 +1,8 @@
 name = "Dark Soul"
 description = "Dark Soul is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
index d1eeb0fe13d..3fe6fa6df43 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-luminous.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
@@ -1,8 +1,8 @@
 name = "Luminous Mirror"
 description = "Luminous Mirror is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = false
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
index 41e520abf2e..1128d4e0ca1 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-moonlight.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
@@ -1,8 +1,8 @@
 name = "Moonlight Dusk"
 description = "Moonlight Dusk is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
index 5bf3710aaa4..51a0a4560e2 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-musica.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
@@ -1,8 +1,8 @@
 name = "Musica"
 description = "Musica is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
index fc4e0eac717..d28838392ec 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-opusdistill.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
@@ -1,8 +1,8 @@
 name = "Opus Distill"
 description = "Opus Distill is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
index fde75187d86..df248b80165 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-shadowsiren.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
@@ -1,8 +1,8 @@
 name = "Shadow Siren"
 description = "Shadow Siren is a Gemma 4 26B A4B multimodal mixture-of-experts fine-tune for creative writing, expressive dialogue, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
similarity index 58%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
index f55e8645df6..4743aaa8aa7 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
@@ -1,8 +1,8 @@
 name = "Gemma 4 26B A4B Uncensored"
-description = "Gemma 4 26B A4B Uncensored is an NVFP4 open-weight multimodal mixture-of-experts model tuned for fewer refusals across chat, coding, tool use, and long-context work."
+description = "Gemma 4 26B A4B Uncensored is an FP8 open-weight multimodal mixture-of-experts model LoRA-tuned for fewer refusals across chat, coding, tool use, and long-context work."
 family = "gemma"
-release_date = "2026-08-22"
-last_updated = "2026-08-22"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored:thinking.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
similarity index 90%
rename from providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored:thinking.toml
rename to providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
index 36888fcce97..bda36ed39c3 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
@@ -1,8 +1,8 @@
 name = "Gemma 4 26B A4B Uncensored Thinking"
 description = "Gemma 4 26B A4B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and long-context problem solving."
 family = "gemma"
-release_date = "2026-08-24"
-last_updated = "2026-08-24"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml b/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
index 671290b4794..808caed4c24 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-darkidol.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
@@ -1,8 +1,8 @@
 name = "DarkIdol"
 description = "DarkIdol is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml b/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
index d2b178d01cd..747250c6fbe 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-fabled.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
@@ -1,8 +1,8 @@
 name = "Fabled"
 description = "Fabled is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml b/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
index cd2a2ade94d..aee93f31708 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-garnet.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
@@ -1,8 +1,8 @@
 name = "Garnet"
 description = "Garnet is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml b/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
index dbc4f8512ad..1f190959f1d 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-gembrain.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
@@ -1,8 +1,8 @@
 name = "Gembrain"
 description = "Gembrain is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml b/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
similarity index 81%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
index 44d7cd91e9d..aa9c62abf42 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-gemsicle.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
@@ -1,13 +1,14 @@
 name = "Gemsicle"
 description = "Gemsicle is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-29"
-last_updated = "2026-08-29"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
-reasoning = false
+reasoning = true
 tool_call = true
 structured_output = false
 open_weights = true
+reasoning_options = []
 
 [cost]
 input = 0.08
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml b/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
index 2c848317d24..8f9c876eb5d 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-isometry.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
@@ -1,8 +1,8 @@
 name = "Isometry"
 description = "Isometry is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml b/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
similarity index 89%
rename from providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml
rename to providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
index e2c70ac2f3b..5379ce74bf6 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it-novelist.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
@@ -1,8 +1,8 @@
 name = "Novelist"
 description = "Novelist is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay."
 family = "gemma"
-release_date = "2026-08-28"
-last_updated = "2026-08-28"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
 attachment = true
 reasoning = true
 tool_call = true
diff --git a/providers/nano-gpt/models/glm-4-air.toml b/providers/nano-gpt/models/glm-4-air.toml
deleted file mode 100644
index 860e39a9bbe..00000000000
--- a/providers/nano-gpt/models/glm-4-air.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM-4 Air"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-01-01"
-last_updated = "2024-06-05"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.2006
-output = 0.2006
-cache_read = 0.1003
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-4-airx.toml b/providers/nano-gpt/models/glm-4-airx.toml
deleted file mode 100644
index 300ccdee9ed..00000000000
--- a/providers/nano-gpt/models/glm-4-airx.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM-4 AirX"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-01-01"
-last_updated = "2024-06-05"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.006
-output = 2.006
-cache_read = 1.003
-
-[limit]
-context = 8_000
-input = 8_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-4-flash.toml b/providers/nano-gpt/models/glm-4-flash.toml
deleted file mode 100644
index bdddde5a4a4..00000000000
--- a/providers/nano-gpt/models/glm-4-flash.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM-4 Flash"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-01-01"
-last_updated = "2024-08-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1003
-output = 0.1003
-cache_read = 0.05015
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-4-plus.toml b/providers/nano-gpt/models/glm-4-plus.toml
deleted file mode 100644
index 7d24aa01459..00000000000
--- a/providers/nano-gpt/models/glm-4-plus.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM-4 Plus"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-09-20"
-last_updated = "2024-08-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 7.497
-output = 7.497
-cache_read = 3.7485
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-4.toml b/providers/nano-gpt/models/glm-4.toml
deleted file mode 100644
index 61ec5f9e9cf..00000000000
--- a/providers/nano-gpt/models/glm-4.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM-4"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-01-01"
-last_updated = "2024-01-16"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 14.994
-output = 14.994
-cache_read = 7.497
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-zero-preview.toml b/providers/nano-gpt/models/glm-zero-preview.toml
deleted file mode 100644
index 65f95b36a00..00000000000
--- a/providers/nano-gpt/models/glm-zero-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM Zero Preview"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2024-12-20"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 1.802
-output = 1.802
-cache_read = 0.901
-
-[limit]
-context = 8_000
-input = 8_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/hunyuan-turbos-20250226.toml b/providers/nano-gpt/models/hunyuan-turbos-20250226.toml
deleted file mode 100644
index 925e3de8f5e..00000000000
--- a/providers/nano-gpt/models/hunyuan-turbos-20250226.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Hunyuan Turbo S"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "hunyuan"
-release_date = "2025-02-27"
-last_updated = "2025-02-27"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.187
-output = 0.374
-cache_read = 0.0935
-
-[limit]
-context = 24_000
-input = 24_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/inflection/inflection-3-productivity.toml b/providers/nano-gpt/models/inflection/inflection-3-productivity.toml
deleted file mode 100644
index ebb75a204b8..00000000000
--- a/providers/nano-gpt/models/inflection/inflection-3-productivity.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Inflection 3 Productivity"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2025-11-21"
-last_updated = "2024-10-11"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.499
-output = 9.996
-cache_read = 1.2495
-
-[limit]
-context = 8_000
-input = 8_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-large-1.6.toml b/providers/nano-gpt/models/jamba-large-1.6.toml
deleted file mode 100644
index 44842744599..00000000000
--- a/providers/nano-gpt/models/jamba-large-1.6.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Large 1.6"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 1.989
-output = 7.99
-cache_read = 0.9945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-large-1.7.toml b/providers/nano-gpt/models/jamba-large-1.7.toml
deleted file mode 100644
index bb798ade9bd..00000000000
--- a/providers/nano-gpt/models/jamba-large-1.7.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Large 1.7"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2025-07-09"
-last_updated = "2025-07-09"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 1.989
-output = 7.99
-cache_read = 0.9945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-large.toml b/providers/nano-gpt/models/jamba-large.toml
deleted file mode 100644
index edfd4b655cc..00000000000
--- a/providers/nano-gpt/models/jamba-large.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Large"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2025-07-09"
-last_updated = "2025-07-09"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 1.989
-output = 7.99
-cache_read = 0.9945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-mini-1.6.toml b/providers/nano-gpt/models/jamba-mini-1.6.toml
deleted file mode 100644
index b694ca781d8..00000000000
--- a/providers/nano-gpt/models/jamba-mini-1.6.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Mini 1.6"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2024-01-01"
-last_updated = "2025-03-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1989
-output = 0.408
-cache_read = 0.09945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-mini-1.7.toml b/providers/nano-gpt/models/jamba-mini-1.7.toml
deleted file mode 100644
index 4db7b7dab3d..00000000000
--- a/providers/nano-gpt/models/jamba-mini-1.7.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Mini 1.7"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2025-07-09"
-last_updated = "2025-07-09"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1989
-output = 0.408
-cache_read = 0.09945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/jamba-mini.toml b/providers/nano-gpt/models/jamba-mini.toml
deleted file mode 100644
index 961f1a435ec..00000000000
--- a/providers/nano-gpt/models/jamba-mini.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Jamba Mini"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "jamba"
-release_date = "2025-07-09"
-last_updated = "2025-07-09"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1989
-output = 0.408
-cache_read = 0.09945
-
-[limit]
-context = 256_000
-input = 256_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/learnlm-1.5-pro-experimental.toml b/providers/nano-gpt/models/learnlm-1.5-pro-experimental.toml
deleted file mode 100644
index 1011ec3dfc1..00000000000
--- a/providers/nano-gpt/models/learnlm-1.5-pro-experimental.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Gemini LearnLM Experimental"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gemini"
-release_date = "2024-01-01"
-last_updated = "2024-05-14"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 3.502
-output = 10.506
-cache_read = 1.751
-
-[limit]
-context = 32_767
-input = 32_767
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/moonshotai/kimi-k3.toml b/providers/nano-gpt/models/moonshotai/kimi-k3.toml
index f626dad940c..fb933a243d2 100644
--- a/providers/nano-gpt/models/moonshotai/kimi-k3.toml
+++ b/providers/nano-gpt/models/moonshotai/kimi-k3.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 13.5
-cache_read = 0.25
+input = 2
+output = 10
+cache_read = 0.2
 
 [limit]
 input = 1_048_576
diff --git a/providers/nano-gpt/models/moonshotai/kimi-latest.toml b/providers/nano-gpt/models/moonshotai/kimi-latest.toml
index 0a406444695..d4fb34a816a 100644
--- a/providers/nano-gpt/models/moonshotai/kimi-latest.toml
+++ b/providers/nano-gpt/models/moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 13.5
-cache_read = 0.25
+input = 2
+output = 10
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b.toml
index 0b92da369ad..a2de9d1372a 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b.toml
@@ -1,7 +1,7 @@
 name = "Ornith 1.5 35B"
 description = "Ornith 1.5 35B A3B is an open-weight mixture-of-experts model for agentic coding, tool use, image understanding, and long-context work. This variant disables thinking for faster direct responses."
 family = "ornith"
-release_date = "2026-08-20"
+release_date = "2026-07-29"
 last_updated = "2026-08-20"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b:thinking.toml
index ead77c04359..5b27e673c4e 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b:thinking.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-35b-a3b:thinking.toml
@@ -1,7 +1,7 @@
 name = "Ornith 1.5 35B Thinking"
 description = "Ornith 1.5 35B A3B is an open-weight mixture-of-experts model for agentic coding, reasoning, tool use, image understanding, and long-context work. This variant enables thinking by default."
 family = "ornith"
-release_date = "2026-08-20"
+release_date = "2026-07-29"
 last_updated = "2026-08-20"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b.toml
deleted file mode 100644
index 6df2ea9f886..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 397B"
-description = "Ornith 1.5 397B is an open-weight mixture-of-experts model built for agentic coding, tool use, image understanding, and long-context workflows. This variant disables thinking for faster direct responses."
-family = "ornith"
-release_date = "2026-08-24"
-last_updated = "2026-08-24"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.9
-output = 3.6
-cache_read = 0.045
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b:thinking.toml
deleted file mode 100644
index d157ab44a6f..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-397b:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 397B Thinking"
-description = "Ornith 1.5 397B with thinking enabled for deeper agentic coding, reasoning, tool use, image understanding, and long-context workflows."
-family = "ornith"
-release_date = "2026-08-24"
-last_updated = "2026-08-24"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.9
-output = 3.6
-cache_read = 0.045
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
index bc5ab6144ba..1c1e75a810c 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
@@ -1,7 +1,7 @@
 name = "Ornith 1.5 9B"
 description = "Ornith 1.5 9B is an FP8 dense open-weight reasoning model built for agentic coding, tool use, visual understanding, and efficient long-context work."
 family = "ornith"
-release_date = "2026-08-23"
+release_date = "2026-07-29"
 last_updated = "2026-08-23"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
index 7c66d52c707..cccf569b7dd 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
@@ -1,7 +1,7 @@
 name = "Ornith 1.5 9B Thinking"
 description = "Ornith 1.5 9B with thinking enabled for more deliberate agentic coding, tool use, visual understanding, and long-context work."
 family = "ornith"
-release_date = "2026-08-24"
+release_date = "2026-07-29"
 last_updated = "2026-08-24"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
index f816868d2b5..7760a7632c0 100644
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.6 35B A3B Uncensored"
 description = "Qwen 3.6 35B A3B Uncensored is an FP8 open-weight mixture-of-experts model tuned for fewer refusals across chat, coding, tool use, and multimodal tasks."
 family = "qwen3.6"
-release_date = "2026-08-21"
+release_date = "2026-07-29"
 last_updated = "2026-08-21"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
index ee4b7ebe25e..0695f700ce0 100644
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.6 35B A3B Uncensored Thinking"
 description = "Qwen 3.6 35B A3B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and complex chat tasks."
 family = "qwen3.6"
-release_date = "2026-08-24"
+release_date = "2026-07-29"
 last_updated = "2026-08-24"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
index e0ad1ce2427..4796ed09894 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.8 27B Fable"
 description = "Qwen 3.8 27B Fable is an open-weight multimodal creative finetune for expressive dialogue, long-form storytelling, character work, and roleplay."
 family = "qwen"
-release_date = "2026-08-28"
+release_date = "2026-07-29"
 last_updated = "2026-08-28"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
index 279664ae347..80546d7654a 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.8 27B Obliterated"
 description = "Qwen 3.8 27B Obliterated is an open-weight multimodal model LoRA-tuned for fewer refusals across chat, coding, reasoning, tool use, and long-context work."
 family = "qwen"
-release_date = "2026-08-24"
+release_date = "2026-07-29"
 last_updated = "2026-08-24"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
index d1f6a593767..d3bdfe9f86d 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.8 27B Obliterated Thinking"
 description = "Qwen 3.8 27B Obliterated with thinking enabled for more deliberate creative work, coding, multimodal analysis, tool use, and long-context problem solving."
 family = "qwen"
-release_date = "2026-08-24"
+release_date = "2026-07-29"
 last_updated = "2026-08-24"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index 3f57c29d10e..efd6402590e 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.8 27B Uncensored"
 description = "Qwen 3.8 27B Uncensored is an NVFP4 open-weight multimodal model LoRA-tuned for fewer refusals across chat, coding, reasoning, tool use, and long-context work."
 family = "qwen"
-release_date = "2026-08-21"
+release_date = "2026-07-29"
 last_updated = "2026-08-21"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 47b5bd8483a..654490ff569 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -1,7 +1,7 @@
 name = "Qwen 3.8 27B Uncensored Thinking"
 description = "Qwen 3.8 27B Uncensored with thinking enabled for more deliberate creative work, coding, multimodal analysis, tool use, and long-context problem solving."
 family = "qwen"
-release_date = "2026-08-25"
+release_date = "2026-07-29"
 last_updated = "2026-08-25"
 attachment = true
 reasoning = true
diff --git a/providers/nano-gpt/models/qwen3.8-27b.toml b/providers/nano-gpt/models/qwen3.8-27b.toml
index ad923e81ad5..dc390a04804 100644
--- a/providers/nano-gpt/models/qwen3.8-27b.toml
+++ b/providers/nano-gpt/models/qwen3.8-27b.toml
@@ -2,8 +2,8 @@ base_model = "alibaba/qwen3.8-27b"
 reasoning_options = []
 
 [cost]
-input = 0.2
-output = 1.4
+input = 0.15
+output = 0.7
 cache_read = 0.04
 
 [limit]
diff --git a/providers/nano-gpt/models/qwen3.8-27b:thinking.toml b/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
index 33736f4ea82..fade9e57385 100644
--- a/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
+++ b/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
@@ -3,8 +3,8 @@ name = "Qwen3.8 27B Thinking"
 reasoning_options = []
 
 [cost]
-input = 0.2
-output = 1.4
+input = 0.15
+output = 0.7
 cache_read = 0.04
 
 [limit]
diff --git a/providers/nano-gpt/models/step-r1-v-mini.toml b/providers/nano-gpt/models/step-r1-v-mini.toml
deleted file mode 100644
index c5474728bce..00000000000
--- a/providers/nano-gpt/models/step-r1-v-mini.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Step R1 V Mini"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "step"
-release_date = "2024-01-01"
-last_updated = "2025-04-08"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.5
-output = 11
-cache_read = 1.25
-
-[limit]
-context = 128_000
-input = 128_000
-output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/zai-org/GLM-4.5-Air.toml b/providers/nano-gpt/models/z-ai/GLM-4.5-Air.toml
similarity index 87%
rename from providers/nano-gpt/models/zai-org/GLM-4.5-Air.toml
rename to providers/nano-gpt/models/z-ai/GLM-4.5-Air.toml
index 38e0ca5177d..8fd72c26d39 100644
--- a/providers/nano-gpt/models/zai-org/GLM-4.5-Air.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.5-Air.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.5-air"
 name = "GLM 4.5 Air"
 reasoning = false
diff --git a/providers/nano-gpt/models/zai-org/GLM-4.5-Air:thinking.toml b/providers/nano-gpt/models/z-ai/GLM-4.5-Air:thinking.toml
similarity index 88%
rename from providers/nano-gpt/models/zai-org/GLM-4.5-Air:thinking.toml
rename to providers/nano-gpt/models/z-ai/GLM-4.5-Air:thinking.toml
index 99872438ace..7b478335840 100644
--- a/providers/nano-gpt/models/zai-org/GLM-4.5-Air:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.5-Air:thinking.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.5-air"
 name = "GLM 4.5 Air (Thinking)"
 structured_output = true
diff --git a/providers/nano-gpt/models/zai-org/GLM-4.5:thinking.toml b/providers/nano-gpt/models/z-ai/GLM-4.5:thinking.toml
similarity index 89%
rename from providers/nano-gpt/models/zai-org/GLM-4.5:thinking.toml
rename to providers/nano-gpt/models/z-ai/GLM-4.5:thinking.toml
index a8b28164926..23cacc8f2ce 100644
--- a/providers/nano-gpt/models/zai-org/GLM-4.5:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.5:thinking.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.5"
 name = "GLM 4.5 (Thinking)"
 tool_call = false
diff --git a/providers/nano-gpt/models/zai-org/GLM-4.6-turbo.toml b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
similarity index 73%
rename from providers/nano-gpt/models/zai-org/GLM-4.6-turbo.toml
rename to providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
index 47baf06d381..66af8b5bcc7 100644
--- a/providers/nano-gpt/models/zai-org/GLM-4.6-turbo.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.6 Turbo"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
+description = "Fast variant of GLM 4.6 for general chat, coding, and analysis with improved latency and strong reasoning."
 family = "glm"
 release_date = "2025-10-02"
 last_updated = "2025-10-02"
diff --git a/providers/nano-gpt/models/zai-org/GLM-4.6-turbo:thinking.toml b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
similarity index 74%
rename from providers/nano-gpt/models/zai-org/GLM-4.6-turbo:thinking.toml
rename to providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
index 42470d6f6d9..8151551a0e2 100644
--- a/providers/nano-gpt/models/zai-org/GLM-4.6-turbo:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.6 Turbo (Thinking)"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
+description = "GLM 4.6 Turbo with thinking mode enabled for enhanced reasoning; shows internal reasoning and supports long context."
 family = "glm"
 release_date = "2025-10-02"
 last_updated = "2025-10-02"
diff --git a/providers/nano-gpt/models/zai-org/glm-4.5.toml b/providers/nano-gpt/models/z-ai/glm-4.5.toml
similarity index 89%
rename from providers/nano-gpt/models/zai-org/glm-4.5.toml
rename to providers/nano-gpt/models/z-ai/glm-4.5.toml
index 94a9691866a..ce0b8ee1f86 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.5.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.5.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.5"
 name = "GLM 4.5"
 reasoning = false
diff --git a/providers/nano-gpt/models/zai-org/glm-4.6-original.toml b/providers/nano-gpt/models/z-ai/glm-4.6-original.toml
similarity index 68%
rename from providers/nano-gpt/models/zai-org/glm-4.6-original.toml
rename to providers/nano-gpt/models/z-ai/glm-4.6-original.toml
index 61cfebbb3bc..19eca561652 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.6-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.6-original.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.6 Original"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "GLM-4.6, Zhipu's flagship text model with 256K context window and advanced reasoning capabilities. Direct via Z-AI (Zhipu)."
 family = "glm"
 release_date = "2025-12-11"
 last_updated = "2025-12-11"
@@ -9,9 +8,7 @@ reasoning = true
 tool_call = true
 structured_output = false
 open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.35
diff --git a/providers/nano-gpt/models/zai-org/glm-4.6v-flash-original.toml b/providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml
similarity index 60%
rename from providers/nano-gpt/models/zai-org/glm-4.6v-flash-original.toml
rename to providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml
index d3807ca58e0..cbe86397742 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.6v-flash-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.6V Flash"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+description = "GLM-4.6V-Flash (9B), a lightweight model optimized for local deployment and low-latency applications. Scales context window to 128k tokens and achieves SoTA performance in visual understanding among similar-scale models."
 family = "glm"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
diff --git a/providers/nano-gpt/models/zai-org/glm-4.6v-original.toml b/providers/nano-gpt/models/z-ai/glm-4.6v-original.toml
similarity index 52%
rename from providers/nano-gpt/models/zai-org/glm-4.6v-original.toml
rename to providers/nano-gpt/models/z-ai/glm-4.6v-original.toml
index 7d0d8e15729..337fa528938 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.6v-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.6v-original.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.6V Original"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+description = "GLM-4.6V scales its context window to 128k tokens in training, and achieves SoTA performance in visual understanding among models of similar parameter scales. Integrates native Function Calling capabilities, bridging 'visual perception' and 'executable action' for multimodal agents. Direct via Z-AI (Zhipu)."
 family = "glm"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
diff --git a/providers/nano-gpt/models/zai-org/glm-4.6v.toml b/providers/nano-gpt/models/z-ai/glm-4.6v.toml
similarity index 90%
rename from providers/nano-gpt/models/zai-org/glm-4.6v.toml
rename to providers/nano-gpt/models/z-ai/glm-4.6v.toml
index 52281c89590..24d900356a8 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.6v.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.6v.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.6v"
 name = "GLM 4.6V"
 reasoning = false
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-flash-original.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
similarity index 65%
rename from providers/nano-gpt/models/zai-org/glm-4.7-flash-original.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
index 4ddef908a94..4ff6c178da5 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-flash-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
@@ -1,6 +1,5 @@
-# Included in subscription
 name = "GLM 4.7 Flash Original"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
+description = "GLM-4.7-Flash is a lightweight 30B model optimized for coding and agentic tasks. Balances high performance with efficiency, perfect for local deployment."
 family = "glm"
 release_date = "2026-01-19"
 last_updated = "2026-01-19"
@@ -9,9 +8,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.07
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-flash-original:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
similarity index 72%
rename from providers/nano-gpt/models/zai-org/glm-4.7-flash-original:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
index 4ec696aa297..60fa9006fc9 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-flash-original:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
@@ -1,6 +1,5 @@
-# Included in subscription
 name = "GLM 4.7 Flash Original Thinking"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
+description = "GLM-4.7-Flash with extended thinking capabilities for complex reasoning. Lightweight 30B model optimized for coding and agentic tasks."
 family = "glm"
 release_date = "2026-01-19"
 last_updated = "2026-01-19"
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-flash.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash.toml
similarity index 82%
rename from providers/nano-gpt/models/zai-org/glm-4.7-flash.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-flash.toml
index a2b47088d3a..104b2a50bb1 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-flash.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash.toml
@@ -1,9 +1,7 @@
 base_model = "zhipuai/glm-4.7-flash"
 name = "GLM 4.7 Flash"
 structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.07
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-flash:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash:thinking.toml
similarity index 88%
rename from providers/nano-gpt/models/zai-org/glm-4.7-flash:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-flash:thinking.toml
index 2ed5d87a758..859ce6d9655 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-flash:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash:thinking.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.7-flash"
 name = "GLM 4.7 Flash Thinking"
 structured_output = false
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-original.toml b/providers/nano-gpt/models/z-ai/glm-4.7-original.toml
similarity index 65%
rename from providers/nano-gpt/models/zai-org/glm-4.7-original.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-original.toml
index 3c7eeac291e..a76ba1036d8 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-original.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.7 Original"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "GLM-4.7 is a next-gen GLM series text model with stronger reasoning, long-context chat, and reliable tool use. Routed directly via Z-AI (Zhipu)."
 family = "glm"
 release_date = "2025-12-22"
 last_updated = "2025-12-22"
@@ -9,9 +8,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.6
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7-original:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7-original:thinking.toml
similarity index 76%
rename from providers/nano-gpt/models/zai-org/glm-4.7-original:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7-original:thinking.toml
index ff799c8c168..c84b511b215 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7-original:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-original:thinking.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 4.7 Original Thinking"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "GLM-4.7 original with extended thinking capabilities for complex reasoning."
 family = "glm"
 release_date = "2025-12-22"
 last_updated = "2025-12-22"
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7.toml b/providers/nano-gpt/models/z-ai/glm-4.7.toml
similarity index 82%
rename from providers/nano-gpt/models/zai-org/glm-4.7.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7.toml
index 802f0c311fa..4a5b8b3579b 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7.toml
@@ -1,9 +1,7 @@
 base_model = "zhipuai/glm-4.7"
 name = "GLM 4.7"
 structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.2
diff --git a/providers/nano-gpt/models/zai-org/glm-4.7:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7:thinking.toml
similarity index 88%
rename from providers/nano-gpt/models/zai-org/glm-4.7:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-4.7:thinking.toml
index 4147631e9f2..e8de2402a9f 100644
--- a/providers/nano-gpt/models/zai-org/glm-4.7:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7:thinking.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "zhipuai/glm-4.7"
 name = "GLM 4.7 Thinking"
 structured_output = true
diff --git a/providers/nano-gpt/models/zai-org/glm-5-original.toml b/providers/nano-gpt/models/z-ai/glm-5-original.toml
similarity index 68%
rename from providers/nano-gpt/models/zai-org/glm-5-original.toml
rename to providers/nano-gpt/models/z-ai/glm-5-original.toml
index bba5e063d77..841b88ec93d 100644
--- a/providers/nano-gpt/models/zai-org/glm-5-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5-original.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 5 Original"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "GLM-5 is Zhipu's latest flagship model with advanced reasoning and instruction following. Routed directly via Z-AI (Zhipu)."
 family = "glm"
 release_date = "2026-02-11"
 last_updated = "2026-02-11"
@@ -9,9 +8,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 1
diff --git a/providers/nano-gpt/models/zai-org/glm-5-original:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5-original:thinking.toml
similarity index 76%
rename from providers/nano-gpt/models/zai-org/glm-5-original:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-5-original:thinking.toml
index dfd041f50f0..77ecbb12949 100644
--- a/providers/nano-gpt/models/zai-org/glm-5-original:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5-original:thinking.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "GLM 5 Original Thinking"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "GLM-5 original with extended thinking capabilities for complex reasoning."
 family = "glm"
 release_date = "2026-02-11"
 last_updated = "2026-02-11"
diff --git a/providers/nano-gpt/models/zai-org/glm-5.1.toml b/providers/nano-gpt/models/z-ai/glm-5.1.toml
similarity index 100%
rename from providers/nano-gpt/models/zai-org/glm-5.1.toml
rename to providers/nano-gpt/models/z-ai/glm-5.1.toml
diff --git a/providers/nano-gpt/models/zai-org/glm-5.1:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5.1:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/zai-org/glm-5.1:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-5.1:thinking.toml
diff --git a/providers/nano-gpt/models/zai-org/glm-5.2.toml b/providers/nano-gpt/models/z-ai/glm-5.2.toml
similarity index 100%
rename from providers/nano-gpt/models/zai-org/glm-5.2.toml
rename to providers/nano-gpt/models/z-ai/glm-5.2.toml
diff --git a/providers/nano-gpt/models/zai-org/glm-5.2:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5.2:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/zai-org/glm-5.2:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-5.2:thinking.toml
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index d6252728002..031db2e4bbf 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -1,9 +1,9 @@
 name = "GLM 5.3 Flash Uncensored"
 description = "GLM 5.3 Flash Uncensored is an uncensored fine-tune of the efficient 320B mixture-of-experts reasoning model, built for unrestricted chat, creative writing, coding, agentic work, tool use, and long-context tasks."
 family = "glm"
-release_date = "2026-08-27"
+release_date = "2026-07-29"
 last_updated = "2026-08-27"
-attachment = false
+attachment = true
 reasoning = true
 tool_call = true
 structured_output = true
@@ -24,5 +24,5 @@ input = 262_144
 output = 32_768
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/zai-org/glm-5.3.toml b/providers/nano-gpt/models/z-ai/glm-5.3.toml
similarity index 88%
rename from providers/nano-gpt/models/zai-org/glm-5.3.toml
rename to providers/nano-gpt/models/z-ai/glm-5.3.toml
index 15216a41c8d..a948ca548f4 100644
--- a/providers/nano-gpt/models/zai-org/glm-5.3.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3.toml
@@ -1,5 +1,5 @@
 base_model = "zhipuai/glm-5.3"
-name = "GLM 5.3 Preview"
+name = "GLM 5.3"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5.3:thinking.toml
similarity index 84%
rename from providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-5.3:thinking.toml
index bb1c9b20d63..ad1dfe31719 100644
--- a/providers/nano-gpt/models/zai-org/glm-5.3:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3:thinking.toml
@@ -1,5 +1,5 @@
 base_model = "zhipuai/glm-5.3"
-name = "GLM 5.3 Preview Thinking"
+name = "GLM 5.3 Thinking"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/nano-gpt/models/zai-org/glm-5.toml b/providers/nano-gpt/models/z-ai/glm-5.toml
similarity index 82%
rename from providers/nano-gpt/models/zai-org/glm-5.toml
rename to providers/nano-gpt/models/z-ai/glm-5.toml
index 1fe1ae7adfc..b8f99c25671 100644
--- a/providers/nano-gpt/models/zai-org/glm-5.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.toml
@@ -1,9 +1,7 @@
 base_model = "zhipuai/glm-5"
 name = "GLM 5"
 structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
+reasoning_options = []
 
 [cost]
 input = 0.5
diff --git a/providers/nano-gpt/models/zai-org/glm-5:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/zai-org/glm-5:thinking.toml
rename to providers/nano-gpt/models/z-ai/glm-5:thinking.toml
diff --git a/providers/nano-gpt/models/zai-org/glm-latest.toml b/providers/nano-gpt/models/z-ai/glm-latest.toml
similarity index 63%
rename from providers/nano-gpt/models/zai-org/glm-latest.toml
rename to providers/nano-gpt/models/z-ai/glm-latest.toml
index 0650bcb46cc..d1e853a3a90 100644
--- a/providers/nano-gpt/models/zai-org/glm-latest.toml
+++ b/providers/nano-gpt/models/z-ai/glm-latest.toml
@@ -1,6 +1,5 @@
-# Included in subscription (uses 2x input tokens)
 name = "GLM Latest"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+description = "Compatibility alias that routes to the newest thinking GLM model. Currently routes to GLM 5.2 Thinking."
 family = "glm"
 release_date = "2026-05-03"
 last_updated = "2026-05-03"
@@ -9,10 +8,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "xhigh"]
+reasoning_options = []
 
 [cost]
 input = 0.42

From c2caab1ab2eb77ceb5dc6a924a441a3b47d80748 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Sun, 30 Aug 2026 10:32:19 +0700
Subject: [PATCH 0235/2012] feat(orcarouter): add GLM-5.3-Flash (#5809)

* feat(orcarouter): add GLM-5.3-Flash model

* fix(orcarouter): add interleaved reasoning_content for glm-5.3-flash

---------

Co-authored-by: 4ebuRushka <metelev.vm@goodline.info>
---
 providers/orcarouter/models/z-ai/glm-5.3-flash.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/orcarouter/models/z-ai/glm-5.3-flash.toml

diff --git a/providers/orcarouter/models/z-ai/glm-5.3-flash.toml b/providers/orcarouter/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..7036c517990
--- /dev/null
+++ b/providers/orcarouter/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,13 @@
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.075
+output = 0.25
+
+[limit]
+context = 1_000_000
+output = 128_000

From 10bacdaeee0fd89777c714ad359c436f98145498 Mon Sep 17 00:00:00 2001
From: Jakob Thun <jakob.thun@gmail.com>
Date: Sun, 30 Aug 2026 05:32:41 +0200
Subject: [PATCH 0236/2012] feat: add klokintegration.se provider (#5812)

* feat: add klokintegration.se provider

Register Klok's first-party OpenAI-compatible models so OpenCode can list them from the catalog.

* docs: match Kloker catalog text to published AI-API page

* fix: set klokintegration catalog cost from published EUR list price

* docs: align klokintegration model text with published AI-API page

* fix: set klokintegration context window to 256k

* fix: set klokintegration limits to 200k context and 50k output

* fix: use Klok K mark for klokintegration logo

Replace the placeholder icon with the brand mark, currentColor and a square viewBox so models.dev can theme it.

---------

Co-authored-by: Jakob Thun <jakob.thun@integrera.com>
---
 providers/klokintegration/logo.svg            | 11 ++++++++
 .../models/Kloker-Integration-Architect.toml  | 27 +++++++++++++++++++
 .../models/Kloker-Integration-Developer.toml  | 27 +++++++++++++++++++
 providers/klokintegration/models/Kloker.toml  | 26 ++++++++++++++++++
 providers/klokintegration/provider.toml       |  5 ++++
 5 files changed, 96 insertions(+)
 create mode 100644 providers/klokintegration/logo.svg
 create mode 100644 providers/klokintegration/models/Kloker-Integration-Architect.toml
 create mode 100644 providers/klokintegration/models/Kloker-Integration-Developer.toml
 create mode 100644 providers/klokintegration/models/Kloker.toml
 create mode 100644 providers/klokintegration/provider.toml

diff --git a/providers/klokintegration/logo.svg b/providers/klokintegration/logo.svg
new file mode 100644
index 00000000000..c488f56b401
--- /dev/null
+++ b/providers/klokintegration/logo.svg
@@ -0,0 +1,11 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="-25.215 0 540.82 540.82" fill="currentColor">
+  <path d="M274.52,75.69c-2.96,3.58-5.64,7.43-8.91,10.69-30.51,30.49-61.12,60.89-91.71,91.31-12.51,12.45-24.97,24.96-37.57,37.32-6.17,6.04-12.57,11.85-18.97,17.65-2,1.82-2.74,3.75-2.74,6.51.03,60.44-.05,120.89-.09,181.33-.03,38.63-.04,77.26-.06,115.89,0,4.26-.01,4.33-4.49,4.34-11.33.03-22.66.01-33.99-.02-4.74-.02-4.77-.06-4.77-5.03,0-44.79.02-89.59,0-134.38-.02-130.55-.05-261.1-.08-391.65,0-2.97,0-5.95,0-9.4,5.24,0,10.03,0,14.82,0,7.66,0,15.33-.02,22.99,0,5.05,0,5.08.02,5.1,5.26.09,20.31.19,40.63.24,60.94.1,35.3.17,70.6.25,105.9,0,1.26,0,2.51,0,4.91,3.31-2.86,5.91-4.82,8.18-7.1,10.68-10.74,21.21-21.64,31.92-32.34,14.72-14.71,29.55-29.32,44.36-43.94,15.04-14.85,30.11-29.68,45.2-44.48.65-.64,1.69-.87,2.55-1.29,1.06,1.22,2.04,2.53,3.19,3.66,8.16,7.99,16.36,15.94,24.55,23.91Z"/>
+  <path d="M219.22,228.55c4.64,4.35,9.48,8.63,14.04,13.19,20.24,20.26,40.39,40.59,60.57,60.9,51.74,52.07,103.51,104.11,155.21,156.23,13.02,13.12,25.77,26.51,38.73,39.68,1.79,1.82,2.52,3.63,2.5,6.15-.06,10.49.1,20.98.11,31.47,0,4.37-.08,4.37-4.5,4.51-12.82.42-12.69.29-21.56-8.69-22.94-23.22-46.03-46.31-69.03-69.48-20.18-20.33-40.31-40.71-60.49-61.04-26.76-26.95-53.56-53.85-80.32-80.8-27.1-27.31-54.17-54.65-81.24-81.99-3.6-3.64-7.16-7.32-10.75-11.01,9.38-9.44,18.37-18.55,27.44-27.59,30.44-30.33,60.87-60.65,91.36-90.92,32.97-32.74,65.99-65.42,98.98-98.12Q391.42.01,407.17.01c12.16.01,24.33.05,36.49.08,1.45,0,2.89,0,4.6.5-24.38,26.4-50.35,51.03-75.55,76.41-25.35,25.53-50.81,50.96-76.41,76.24-25.57,25.24-51.38,50.23-77.08,75.32Z"/>
+  <path d="M41.43,540.6c-12.49.08-24.98.2-37.47.19-3.61,0-3.66-.21-3.67-4.24-.06-20.82-.15-41.63-.16-62.45C.08,335.38.05,196.66,0,57.95,0,40.29,0,22.64,0,4.99,0,.21,0,.17,4.99.16c12-.02,23.99,0,35.99,0,.28,2.13.8,4.26.8,6.39.05,49.95.04,99.89.04,149.84,0,124.37-.02,248.74,0,373.11,0,1.8.32,3.6.49,5.4,0,1.26,0,2.52,0,3.78-.09.08-.27.18-.26.25.03.23.14.45.21.68-.3.37-.6.75-.9,1.12,0,0,.06-.12.06-.12Z"/>
+  <path d="M431.44,540.11c-12.33.19-23.47.39-34.61.53-5.33.07-10.67-.08-16,.07-2.4.07-4.03-.86-5.65-2.52-12.54-12.89-25.19-25.68-37.74-38.56-14.98-15.38-29.77-30.95-44.89-46.2-25.91-26.14-52.07-52.02-78-78.13-28.51-28.72-56.86-57.59-85.29-86.38-3.74-3.79-7.58-7.49-11.31-11.3-2.43-2.48-2.58-3.4-.16-5.85,7.38-7.46,14.94-14.73,22.35-22.15,1.89-1.89,3.13-1.55,4.9.23,13.94,14.09,28.02,28.04,41.95,42.14,20.36,20.61,40.57,41.36,60.95,61.94,18.05,18.23,36.31,36.25,54.35,54.48,32.43,32.78,64.78,65.65,97.14,98.5,9.58,9.72,19.09,19.52,28.62,29.29.91.93,1.71,1.96,3.39,3.92Z"/>
+  <path d="M482.25,7.86c2.03-2.14,4.06-4.29,6.88-7.27.48,2.61,1.01,4.14,1.01,5.66.02,15.98-.02,31.96-.15,47.94,0,1.15-.74,2.39-1.41,3.4-1.01,1.51-2.14,2.98-3.45,4.23-10.58,10.09-21.32,20.03-31.8,30.23-4.98,4.85-9.31,10.37-14.26,15.26-10.04,9.9-20.42,19.46-30.44,29.38-9.54,9.44-18.68,19.27-28.2,28.74-14.2,14.13-28.65,28.01-42.85,42.15-15.45,15.38-30.7,30.94-46.1,46.38-3.4,3.41-7,6.6-10.81,10.18-2.62-2.16-5.32-4.1-7.68-6.39-6.34-6.14-12.52-12.44-18.78-18.66-1.36-1.35-2.53-2.37-.46-4.4,13.75-13.55,27.37-27.23,41.03-40.87.5-.49,1.01-.98,1.51-1.47.13-.2.3-.35.52-.43,1.4-1.36,2.81-2.71,4.21-4.07.4-.38.8-.76,1.2-1.14.35-.34.69-.69,1.04-1.04.27-.25.54-.51.81-.76,0,0,.12-.12.12-.12.34-.34.68-.67,1.02-1.01.7-.76,1.41-1.52,2.12-2.29.17-.2.38-.36.62-.47,0,0,.29-.07.29-.07l.05-.3c.15-.3.36-.55.62-.75,0,0,.26-.04.26-.04,0,0,.03-.26.03-.26.29-.26.58-.53.86-.79,1.12-1.12,2.24-2.24,3.36-3.36.29-.3.57-.59.86-.89,3.38-1.09,5.64-3.34,6.86-6.65.3-.34.61-.69.91-1.03.36-.29.72-.59,1.07-.88.35-.35.7-.69,1.05-1.04,0,0,.06-.06.06-.06.4-.36.79-.72,1.18-1.08.24-.22.48-.44.72-.66.49-.51.97-1.02,1.45-1.52,1.79-1.88,3.59-3.76,5.39-5.65.34-.39.67-.78,1.01-1.17,0,0,.15-.17.15-.17.34-.33.68-.66,1.02-.99,2-1.97,3.99-3.94,5.99-5.91.36-.34.72-.68,1.07-1.01,1.72-1.21,3.64-2.23,5.15-3.67,11.42-10.88,22.72-21.87,34.14-32.74,3.97-3.78,8.4-7.08,12.2-11,5.65-5.83,10.69-12.25,16.4-18.02,15.77-15.94,31.83-31.59,47.61-47.53,5.46-5.52,10.41-11.55,15.6-17.34.09-.19.22-.33.41-.41.48-.48.96-.96,1.44-1.44.18-.24.41-.44.68-.58,1.3-1.22,2.6-2.43,3.9-3.65.26-.27.53-.54.79-.81.18-.18.37-.37.55-.56.37-.35.73-.69,1.1-1.04,0,0,.04-.06.04-.06Z"/>
+  <path d="M143.59,345.23c1.58.92,2.37,1.19,2.92,1.72,11.34,11.02,22.62,22.09,33.99,33.08,1.98,1.91,1.76,4.16,1.76,6.47,0,32.63,0,65.26,0,97.89,0,16.98,0,33.96-.02,50.94,0,4.52-.05,4.67-4.59,4.75-9.66.17-19.32.22-28.98.22-5.08,0-5.12-.1-5.12-5.35,0-46.45.02-92.89.03-139.34,0-14.82,0-29.63,0-44.45,0-1.65,0-3.29,0-5.93Z"/>
+  <path d="M274.52,75.69c-8.19-7.96-16.38-15.92-24.55-23.91-1.15-1.13-2.13-2.44-3.19-3.66,8.65-8.52,17.3-17.03,25.94-25.55,6.63-6.54,13.27-13.07,19.84-19.68,1.86-1.87,3.8-2.91,6.56-2.88,16.31.16,32.62.17,48.93.24.62,0,1.23.09,1.96.15-2.53,5.19-19.84,21.88-37.43,37.85-13.17,11.95-25.4,24.93-38.05,37.45Z"/>
+  <path d="M174.69,4.64c10.81,3.98,18.08,11.61,19.87,22.83,3.25,20.43-11.84,34.8-29.57,34.41-14.88-.33-32.17-14.91-27.01-36.03,2.87-11.74,10.08-18.61,21.37-21.77,5.11.19,10.22.37,15.34.56Z"/>
+  <path d="M473.33,16.42c-5.18,5.8-10.14,11.82-15.6,17.34-15.77,15.94-31.84,31.59-47.61,47.53-5.71,5.77-10.75,12.19-16.4,18.02-3.8,3.92-8.24,7.22-12.2,11-11.42,10.87-22.72,21.87-34.14,32.74-1.51,1.44-3.43,2.46-5.15,3.67,8.41-8.49,16.77-17.03,25.23-25.46,14.01-13.97,28.1-27.85,42.11-41.81,20.36-20.29,40.68-40.61,61.04-60.91.81-.81,1.81-1.42,2.72-2.12Z"/>
+</svg>
diff --git a/providers/klokintegration/models/Kloker-Integration-Architect.toml b/providers/klokintegration/models/Kloker-Integration-Architect.toml
new file mode 100644
index 00000000000..a821123824e
--- /dev/null
+++ b/providers/klokintegration/models/Kloker-Integration-Architect.toml
@@ -0,0 +1,27 @@
+# Cost from https://klokintegration.se/docs/ai-api (email-token rate):
+# €0.20 input / €1.00 output per MTok. EUR/USD 1.1583 (2026-08-28) → 0.23 / 1.16 USD.
+# Context 200k and max output 50k from Klok.
+# Gateway appends Klok practice to the system prompt and runs lookup tools server-side.
+name = "Kloker Integration Architect"
+description = "Knows the customer integration environment and Klok best practices. Opinionated about structure. The gateway runs ecosystem lookup tools server-side and appends a system-prompt injection (data contracts, CloudEvents, event-driven flows). Client system prompts and OpenAI tool calls are preserved."
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+temperature = true
+release_date = "2026-08-29"
+last_updated = "2026-08-29"
+open_weights = false
+status = "beta"
+
+[cost]
+input = 0.23
+output = 1.16
+
+[limit]
+context = 200_000
+output = 50_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/klokintegration/models/Kloker-Integration-Developer.toml b/providers/klokintegration/models/Kloker-Integration-Developer.toml
new file mode 100644
index 00000000000..6e858630e60
--- /dev/null
+++ b/providers/klokintegration/models/Kloker-Integration-Developer.toml
@@ -0,0 +1,27 @@
+# Cost from https://klokintegration.se/docs/ai-api (email-token rate):
+# €0.20 input / €1.00 output per MTok. EUR/USD 1.1583 (2026-08-28) → 0.23 / 1.16 USD.
+# Context 200k and max output 50k from Klok.
+# Gateway appends Klok practice to the system prompt and runs lookup tools server-side.
+name = "Kloker Integration Developer"
+description = "Knows the customer integration environment and Klok best practices. Opinionated about implementation. The gateway runs ecosystem lookup tools server-side and appends a system-prompt injection. Client system prompts and OpenAI tool calls are preserved."
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+temperature = true
+release_date = "2026-08-29"
+last_updated = "2026-08-29"
+open_weights = false
+status = "beta"
+
+[cost]
+input = 0.23
+output = 1.16
+
+[limit]
+context = 200_000
+output = 50_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/klokintegration/models/Kloker.toml b/providers/klokintegration/models/Kloker.toml
new file mode 100644
index 00000000000..d8385b72d6d
--- /dev/null
+++ b/providers/klokintegration/models/Kloker.toml
@@ -0,0 +1,26 @@
+# Cost from https://klokintegration.se/docs/ai-api (email-token rate):
+# €0.20 input / €1.00 output per MTok. EUR/USD 1.1583 (2026-08-28) → 0.23 / 1.16 USD.
+# Context 200k and max output 50k from Klok.
+name = "Kloker"
+description = "Cheap general model with a clean context. Nothing from the customer environment is packed in. It tracks the current best open source model. The Klok team verifies it and upgrades it periodically."
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+temperature = true
+release_date = "2026-08-29"
+last_updated = "2026-08-29"
+open_weights = false
+status = "beta"
+
+[cost]
+input = 0.23
+output = 1.16
+
+[limit]
+context = 200_000
+output = 50_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/klokintegration/provider.toml b/providers/klokintegration/provider.toml
new file mode 100644
index 00000000000..39e11a5aed6
--- /dev/null
+++ b/providers/klokintegration/provider.toml
@@ -0,0 +1,5 @@
+name = "klokintegration.se"
+npm = "@ai-sdk/openai-compatible"
+env = ["KLOKINTEGRATION_API_KEY"]
+api = "https://api-gw.klok.ipaas.se/proxy/kloker-key/v1"
+doc = "https://klokintegration.se/docs/ai-api"

From 039a1e0296aa72906371f007b0cda33d84f20230 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Sun, 30 Aug 2026 10:33:51 +0700
Subject: [PATCH 0237/2012] feat: add bothub and tokenrouter providers (#5814)

* add providers/bothub/provider.toml

* add providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml

* add providers/bothub/models/gemma-4-31b-it:free.toml

* add providers/tokenrouter/provider.toml

* add providers/tokenrouter/models/z-ai/glm-5.3-free.toml

* fix providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml content

* fix providers/bothub/models/gemma-4-31b-it:free.toml content

* fix providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml per review

* fix providers/bothub/models/gemma-4-31b-it:free.toml per review

* fix providers/tokenrouter/models/z-ai/glm-5.3-free.toml per review

* add providers/bothub/logo.svg

* add providers/tokenrouter/logo.svg

* fix providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml interleaved and reasoning per review

* fix gemma interleaved and reasoning per review

* replace bothub logo with brand robot mark

* gemma: document toggle wire path

* bothub: align provider comment with per-model controls

* tokenrouter: narrow provider comment to verified effort set
---
 providers/bothub/logo.svg                     | 33 +++++++++++++++++++
 .../bothub/models/gemma-4-31b-it:free.toml    | 14 ++++++++
 .../nemotron-3-ultra-550b-a55b:free.toml      | 13 ++++++++
 providers/bothub/provider.toml                | 13 ++++++++
 providers/tokenrouter/logo.svg                |  4 +++
 .../tokenrouter/models/z-ai/glm-5.3-free.toml | 13 ++++++++
 providers/tokenrouter/provider.toml           | 10 ++++++
 7 files changed, 100 insertions(+)
 create mode 100644 providers/bothub/logo.svg
 create mode 100644 providers/bothub/models/gemma-4-31b-it:free.toml
 create mode 100644 providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml
 create mode 100644 providers/bothub/provider.toml
 create mode 100644 providers/tokenrouter/logo.svg
 create mode 100644 providers/tokenrouter/models/z-ai/glm-5.3-free.toml
 create mode 100644 providers/tokenrouter/provider.toml

diff --git a/providers/bothub/logo.svg b/providers/bothub/logo.svg
new file mode 100644
index 00000000000..01c3f46c1de
--- /dev/null
+++ b/providers/bothub/logo.svg
@@ -0,0 +1,33 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 500 500">
+  <path d="
+    M 238,170 
+    L 238,126 
+    A 20 20 0 1 1 262,126 
+    L 262,170 
+    L 335,170 
+    A 35 35 0 0 1 370,205 
+    L 370,225 
+    A 12 12 0 0 1 382,237 
+    L 382,263 
+    A 12 12 0 0 1 370,275 
+    L 370,315 
+    A 35 35 0 0 1 335,350 
+    L 165,350 
+    A 35 35 0 0 1 130,315 
+    L 130,275 
+    A 12 12 0 0 1 118,263 
+    L 118,237 
+    A 12 12 0 0 1 130,225 
+    L 130,205 
+    A 35 35 0 0 1 165,170 
+    Z" 
+    fill="none" 
+    stroke="currentColor" 
+    stroke-width="12" 
+    stroke-linecap="round" 
+    stroke-linejoin="round"
+  />
+  <circle cx="202" cy="242" r="21" fill="currentColor" />
+  <circle cx="298" cy="242" r="21" fill="currentColor" />
+  <rect x="205" y="288" width="90" height="26" rx="13" fill="currentColor" />
+</svg>
\ No newline at end of file
diff --git a/providers/bothub/models/gemma-4-31b-it:free.toml b/providers/bothub/models/gemma-4-31b-it:free.toml
new file mode 100644
index 00000000000..627a8338b29
--- /dev/null
+++ b/providers/bothub/models/gemma-4-31b-it:free.toml
@@ -0,0 +1,14 @@
+# Bothub wire: reasoning arrives in message.reasoning (mirrored in message.reasoning_details as reasoning.text).
+# Toggle: POST /v1/chat/completions request field reasoning_effort = "low" (on) | "none" (off);
+# live-verified 2026-08-29: any non-none value enables thinking, "none" disables it (no graded effect between on-values).
+base_model = "google/gemma-4-31b-it"
+name = "Gemma 4 31B IT (free)"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
\ No newline at end of file
diff --git a/providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml b/providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml
new file mode 100644
index 00000000000..63ca995b51c
--- /dev/null
+++ b/providers/bothub/models/nemotron-3-ultra-550b-a55b:free.toml
@@ -0,0 +1,13 @@
+# Bothub wire: reasoning in message.reasoning (also mirrored in message.reasoning_details as reasoning.text)
+base_model = "nvidia/nemotron-3-ultra-550b-a55b"
+name = "Nemotron 3 Ultra (free)"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
\ No newline at end of file
diff --git a/providers/bothub/provider.toml b/providers/bothub/provider.toml
new file mode 100644
index 00000000000..80c4ff82069
--- /dev/null
+++ b/providers/bothub/provider.toml
@@ -0,0 +1,13 @@
+name = "Bothub"
+env = ["BOTHUB_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+# Raw HTTP reasoning controls (live-verified 2026-08-29):
+# Chat POST `/v1/chat/completions` takes `reasoning_effort`; upstream
+# reasoning returns in the message side channel (`reasoning`, mirrored
+# in `reasoning_details`). Per-model request controls differ:
+# - nemotron-3-ultra-550b-a55b:free: effort none|medium|high
+# - gemma-4-31b-it:free: binary toggle (any non-none value enables
+#   thinking, `none` disables it)
+# https://bothub.ru
+api = "https://openai.bothub.ru/v1"
+doc = "https://bothub.ru/models"
\ No newline at end of file
diff --git a/providers/tokenrouter/logo.svg b/providers/tokenrouter/logo.svg
new file mode 100644
index 00000000000..56c8191b295
--- /dev/null
+++ b/providers/tokenrouter/logo.svg
@@ -0,0 +1,4 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none">
+  <rect x="3" y="7" width="18" height="10" rx="5" stroke="currentColor" stroke-width="1.7"/>
+  <circle cx="16" cy="12" r="3" fill="currentColor"/>
+</svg>
\ No newline at end of file
diff --git a/providers/tokenrouter/models/z-ai/glm-5.3-free.toml b/providers/tokenrouter/models/z-ai/glm-5.3-free.toml
new file mode 100644
index 00000000000..0e3d3ff1864
--- /dev/null
+++ b/providers/tokenrouter/models/z-ai/glm-5.3-free.toml
@@ -0,0 +1,13 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (free)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
\ No newline at end of file
diff --git a/providers/tokenrouter/provider.toml b/providers/tokenrouter/provider.toml
new file mode 100644
index 00000000000..a0fa188e055
--- /dev/null
+++ b/providers/tokenrouter/provider.toml
@@ -0,0 +1,10 @@
+name = "TokenRouter"
+env = ["TOKENROUTER_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+# Raw HTTP reasoning controls (live-verified 2026-08-29):
+# Chat POST `/v1/chat/completions` accepts `reasoning_effort =
+# low|high|max` (verified set for z-ai/glm-5.3-free); thinking is
+# always on and reasoning returns in `reasoning_content`.
+# https://www.tokenrouter.com
+api = "https://api.tokenrouter.com/v1"
+doc = "https://www.tokenrouter.com/docs/tokenrouter-feature-guide/"
\ No newline at end of file

From 448526f5292691b29bb4e6dcec8b21e00f490648 Mon Sep 17 00:00:00 2001
From: h3nnes <henrik+github@hnrk.io>
Date: Sun, 30 Aug 2026 05:34:00 +0200
Subject: [PATCH 0238/2012] feat(synthetic): add GLM-5.3-Flash (#5818)

Add GLM-5.3-Flash to the Synthetic provider. The model is served by
api.synthetic.new/openai/v1 (model id hf:zai-org/GLM-5.3-Flash) but was
missing from the catalog, so it never appeared in opencode's model picker
even though the live /models endpoint exposes it.

The provider file is override-only: base_model points at the existing
canonical lab entry models/zhipuai/glm-5.3-flash.toml. Provider-specific
values (cost, limits, modalities, reasoning efforts, interleaved reasoning
field) are sourced from the live model surface:
  GET https://api.synthetic.new/openai/v1/models

- context 524288, output 65536 (synthetic serving limits)
- input modalities text+image (canonical lists video+pdf too; synthetic
  API does not accept those)
- reasoning efforts low/high/max
- interleaved reasoning via reasoning_content
- pricing $0.15/$0.50/$0.04 per MTok (input/output/cache_read)
---
 .../models/hf:zai-org/GLM-5.3-Flash.toml      | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/synthetic/models/hf:zai-org/GLM-5.3-Flash.toml

diff --git a/providers/synthetic/models/hf:zai-org/GLM-5.3-Flash.toml b/providers/synthetic/models/hf:zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..f49d9a0acee
--- /dev/null
+++ b/providers/synthetic/models/hf:zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,19 @@
+# Pricing, limits, modalities, and reasoning efforts from the live model surface:
+# GET https://api.synthetic.new/openai/v1/models  (hf:zai-org/GLM-5.3-Flash)
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[modalities]
+input = ["text", "image"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.04
+
+[limit]
+context = 524_288
+output = 65_536

From 489a3f4ea08a56dd060f76eedba9ebfa18dc1641 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 22:34:09 -0500
Subject: [PATCH 0239/2012] fix: cloudflare-workers-ai: glm-5.3-flash limit
 1310720 exceeds gateway max 1048576 (same as #4422 for glm-5.2) (#5825)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3-flash.toml                  | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
index c242092c4a7..f8b5d01f800 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -1,3 +1,6 @@
+# Context/output capped to Cloudflare Workers AI context window (1,048,576).
+# Catalog/search previously reported 1,310,720, which exceeds AI Gateway max_completion_tokens.
+# https://developers.cloudflare.com/workers-ai/models/glm-5.3-flash/
 name = "Glm 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
 family = "glm"
@@ -17,8 +20,8 @@ output = 0.5
 cache_read = 0.03
 
 [limit]
-context = 1_310_720
-output = 1_310_720
+context = 1_048_576
+output = 1_048_576
 
 [modalities]
 input = ["text", "image"]

From 381cfb58bda4ba7466c7932eecd8154538eb04b6 Mon Sep 17 00:00:00 2001
From: realtimegb <support@realtimecomms.co.uk>
Date: Sun, 30 Aug 2026 04:35:44 +0100
Subject: [PATCH 0240/2012] Add gateway.fast provider (#5828)

* Add gateway.fast provider

OpenAI-compatible gateway serving DeepSeek V4, GLM 5.2, MiMo V2.5 and
Qwen 3.8 Max at upstream cost plus a flat 10%, with the full upstream
cache-hit discount passed through. DeepSeek models use time-of-day
pricing upstream; listed costs are the off-peak rate (peak is 2x,
weekdays 01:00-04:00 and 06:00-10:00 UTC), noted in provider.toml.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

* Address review: wire comments, Pro effort high|max, MiMo interleaved

- deepseek-v4-pro effort narrowed to [high, max] (Pro maps low->high)
- every toggle model now has a leading wire comment with the exact
  request fields (DeepSeek/MiMo thinking.type, Fireworks surface for
  GLM and Qwen)
- qwen3.8-max stays toggle-only, citing the Fireworks host surface
- MiMo entries gain [interleaved] reasoning_content, matching
  providers/xiaomi

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

* Address second review: verified wire syntax, price citations, host deltas

- DeepSeek comments now cite the current price sheet (effective
  2026-08-16: flash 0.22/0.007/0.66, pro 0.66/0.022/1.98 off-peak);
  authored costs are that sheet plus a flat 10%. The first-party
  catalog entries predate the 2026-08-16 change.
- GLM/Qwen toggle wire documented and verified end-to-end on this
  host: reasoning_effort="none" (or chat_template_kwargs
  thinking=false) disables thinking.
- qwen3.8-max gains host deltas: attachment=false, text+image
  modalities (image verified; video/pdf unsupported), context 1M,
  output 32768, interleaved reasoning_content.
- env renamed to GATEWAY_FAST_API_KEY, matching the provider docs.
- MiMo comments carry CNY rates, conversion rate + date, and markup.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

* Address third review: GLM effort enum with none, trim qwen restates

- glm-5.2 and glm-5.2-fast: single effort enum [none, high, max]
  replacing toggle + [high, max]; wire comments updated (off is
  reasoning_effort=none on the same field).
- qwen3.8-max: dropped restated limit.context and modalities.output;
  kept the real deltas (attachment, input modalities, output cap).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

* Rebrand provider: gateway-fast -> above (above.dev)

The service has rebranded from gateway.fast to above.dev. Provider id,
name, env (ABOVE_API_KEY), api base (https://api.above.dev/v1) and doc
links updated; models and costs unchanged. gateway.fast web pages 301
to above.dev and the old api.gateway.fast host remains alive for
existing clients.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

---------

Co-authored-by: realtimegb <pkreil3@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
---
 providers/above/logo.svg                      |  3 ++
 .../models/deepseek-v4-flash-vision-exp.toml  | 28 ++++++++++++++++++
 providers/above/models/deepseek-v4-flash.toml | 28 ++++++++++++++++++
 providers/above/models/deepseek-v4-pro.toml   | 29 +++++++++++++++++++
 providers/above/models/glm-5.2-fast.toml      | 18 ++++++++++++
 providers/above/models/glm-5.2.toml           | 18 ++++++++++++
 .../models/mimo-v2.5-pro-ultraspeed.toml      | 20 +++++++++++++
 providers/above/models/mimo-v2.5-pro.toml     | 19 ++++++++++++
 providers/above/models/qwen3.8-max.toml       | 27 +++++++++++++++++
 providers/above/provider.toml                 | 11 +++++++
 10 files changed, 201 insertions(+)
 create mode 100644 providers/above/logo.svg
 create mode 100644 providers/above/models/deepseek-v4-flash-vision-exp.toml
 create mode 100644 providers/above/models/deepseek-v4-flash.toml
 create mode 100644 providers/above/models/deepseek-v4-pro.toml
 create mode 100644 providers/above/models/glm-5.2-fast.toml
 create mode 100644 providers/above/models/glm-5.2.toml
 create mode 100644 providers/above/models/mimo-v2.5-pro-ultraspeed.toml
 create mode 100644 providers/above/models/mimo-v2.5-pro.toml
 create mode 100644 providers/above/models/qwen3.8-max.toml
 create mode 100644 providers/above/provider.toml

diff --git a/providers/above/logo.svg b/providers/above/logo.svg
new file mode 100644
index 00000000000..00d80ac6381
--- /dev/null
+++ b/providers/above/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M14.25 2 6 13h4.5L9.75 22 18 11h-4.5L14.25 2z"/>
+</svg>
diff --git a/providers/above/models/deepseek-v4-flash-vision-exp.toml b/providers/above/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..76b0d5df561
--- /dev/null
+++ b/providers/above/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,28 @@
+# Passthrough to the official DeepSeek API.
+# Toggle: `thinking.type = enabled|disabled`; effort: `reasoning_effort = low|high|max`
+# (Flash maps requested low→low). https://api-docs.deepseek.com/guides/thinking_mode/
+# Off-peak rate from the CURRENT DeepSeek price sheet effective 2026-08-16
+# (flash $0.22/$0.007/$0.66, pro $0.66/$0.022/$1.98 off-peak) plus a flat 10%;
+# peak (weekdays 01:00-04:00 and 06:00-10:00 UTC, Beijing-time weekends
+# off-peak) is 2x. https://api-docs.deepseek.com/quick_start/pricing/
+# (accessed 2026-08-29; note first-party catalog entries predate 2026-08-16).
+# Toggle verified end-to-end on this host 2026-08-29.
+# Reasoning tokens billed at the output rate.
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+name = "DeepSeek V4 Flash Vision (Exp)"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.242
+output = 0.726
+reasoning = 0.726
+cache_read = 0.0077
diff --git a/providers/above/models/deepseek-v4-flash.toml b/providers/above/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..d8ff403d9ea
--- /dev/null
+++ b/providers/above/models/deepseek-v4-flash.toml
@@ -0,0 +1,28 @@
+# Passthrough to the official DeepSeek API.
+# Toggle: `thinking.type = enabled|disabled`; effort: `reasoning_effort = low|high|max`
+# (Flash maps requested low→low). https://api-docs.deepseek.com/guides/thinking_mode/
+# Off-peak rate from the CURRENT DeepSeek price sheet effective 2026-08-16
+# (flash $0.22/$0.007/$0.66, pro $0.66/$0.022/$1.98 off-peak) plus a flat 10%;
+# peak (weekdays 01:00-04:00 and 06:00-10:00 UTC, Beijing-time weekends
+# off-peak) is 2x. https://api-docs.deepseek.com/quick_start/pricing/
+# (accessed 2026-08-29; note first-party catalog entries predate 2026-08-16).
+# Toggle verified end-to-end on this host 2026-08-29.
+# Reasoning tokens billed at the output rate.
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.242
+output = 0.726
+reasoning = 0.726
+cache_read = 0.0077
diff --git a/providers/above/models/deepseek-v4-pro.toml b/providers/above/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..f0d6b4759e9
--- /dev/null
+++ b/providers/above/models/deepseek-v4-pro.toml
@@ -0,0 +1,29 @@
+# Passthrough to the official DeepSeek API.
+# Toggle: `thinking.type = enabled|disabled`; effort: `reasoning_effort = high|max`
+# (Pro maps requested low→high, so low is not a real level).
+# https://api-docs.deepseek.com/guides/thinking_mode/
+# Off-peak rate from the CURRENT DeepSeek price sheet effective 2026-08-16
+# (flash $0.22/$0.007/$0.66, pro $0.66/$0.022/$1.98 off-peak) plus a flat 10%;
+# peak (weekdays 01:00-04:00 and 06:00-10:00 UTC, Beijing-time weekends
+# off-peak) is 2x. https://api-docs.deepseek.com/quick_start/pricing/
+# (accessed 2026-08-29; note first-party catalog entries predate 2026-08-16).
+# Toggle verified end-to-end on this host 2026-08-29.
+# Reasoning tokens billed at the output rate.
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.726
+output = 2.178
+reasoning = 2.178
+cache_read = 0.0242
diff --git a/providers/above/models/glm-5.2-fast.toml b/providers/above/models/glm-5.2-fast.toml
new file mode 100644
index 00000000000..aa8d211a43b
--- /dev/null
+++ b/providers/above/models/glm-5.2-fast.toml
@@ -0,0 +1,18 @@
+# Passthrough to the Fireworks fast router (~2x standard throughput).
+# Effort: `reasoning_effort = none|high|max`; "none" disables thinking
+# (`chat_template_kwargs = {"thinking": false}` also works as an alternate).
+# Verified end-to-end on this host 2026-08-29.
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2 Fast"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.31
+output = 7.26
+cache_read = 0.231
diff --git a/providers/above/models/glm-5.2.toml b/providers/above/models/glm-5.2.toml
new file mode 100644
index 00000000000..75a9b2317be
--- /dev/null
+++ b/providers/above/models/glm-5.2.toml
@@ -0,0 +1,18 @@
+# Passthrough to Fireworks serverless.
+# Effort: `reasoning_effort = none|high|max`; "none" disables thinking
+# (`chat_template_kwargs = {"thinking": false}` also works as an alternate).
+# Verified end-to-end on this host 2026-08-29.
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.54
+output = 4.84
+cache_read = 0.154
diff --git a/providers/above/models/mimo-v2.5-pro-ultraspeed.toml b/providers/above/models/mimo-v2.5-pro-ultraspeed.toml
new file mode 100644
index 00000000000..cebe2ae0c69
--- /dev/null
+++ b/providers/above/models/mimo-v2.5-pro-ultraspeed.toml
@@ -0,0 +1,20 @@
+# Passthrough to the official MiMo API (api.xiaomimimo.com); same model as
+# mimo-v2.5-pro served at up to ~1000 tok/s, 3x price for speed.
+# Toggle: `thinking.type = enabled|disabled`, as providers/xiaomi; verified
+# end-to-end on this host 2026-08-29.
+# RMB upstream (pro CNY 3/0.025/6, ultraspeed CNY 9/0.075/18 per MTok)
+# converted at 6.5 CNY/USD (operator rate set 2026-07, spot ~6.79; buffer
+# covers FX drift), then a flat 10% markup; rounded to 4dp. Checked 2026-08-29.
+base_model = "xiaomi/mimo-v2.5-pro-ultraspeed"
+name = "MiMo V2.5 Pro UltraSpeed"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.5231
+output = 3.0462
+cache_read = 0.0127
diff --git a/providers/above/models/mimo-v2.5-pro.toml b/providers/above/models/mimo-v2.5-pro.toml
new file mode 100644
index 00000000000..c1a7dec3005
--- /dev/null
+++ b/providers/above/models/mimo-v2.5-pro.toml
@@ -0,0 +1,19 @@
+# Passthrough to the official MiMo API (api.xiaomimimo.com).
+# Toggle: `thinking.type = enabled|disabled`, as providers/xiaomi; verified
+# end-to-end on this host 2026-08-29.
+# RMB upstream (pro CNY 3/0.025/6, ultraspeed CNY 9/0.075/18 per MTok)
+# converted at 6.5 CNY/USD (operator rate set 2026-07, spot ~6.79; buffer
+# covers FX drift), then a flat 10% markup; rounded to 4dp. Checked 2026-08-29.
+base_model = "xiaomi/mimo-v2.5-pro"
+name = "MiMo V2.5 Pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.5077
+output = 1.0154
+cache_read = 0.0042
diff --git a/providers/above/models/qwen3.8-max.toml b/providers/above/models/qwen3.8-max.toml
new file mode 100644
index 00000000000..40b109783ce
--- /dev/null
+++ b/providers/above/models/qwen3.8-max.toml
@@ -0,0 +1,27 @@
+# Passthrough to Fireworks serverless.
+# Toggle: `reasoning_effort = "none"` disables thinking (also accepts
+# `chat_template_kwargs = {"thinking": false}`); verified end-to-end on this
+# host 2026-08-29. The Alibaba first-party effort levels and budget_tokens are
+# not exposed by this host.
+# Host deltas vs the lab base: image input verified end-to-end 2026-08-23
+# (video/pdf not supported on this host); output capped at 32,768.
+base_model = "alibaba/qwen3.8-max"
+name = "Qwen 3.8 Max"
+attachment = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[modalities]
+input = ["text", "image"]
+
+[limit]
+output = 32_768
+
+[cost]
+input = 2.2
+output = 6.6
+cache_read = 0.275
diff --git a/providers/above/provider.toml b/providers/above/provider.toml
new file mode 100644
index 00000000000..8b921569ed2
--- /dev/null
+++ b/providers/above/provider.toml
@@ -0,0 +1,11 @@
+name = "above.dev"
+env = ["ABOVE_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+# OpenAI-compatible gateway selling upstream cost plus a flat 10% on every
+# token class, with the full upstream cache-hit discount passed through.
+# DeepSeek models use time-of-day pricing (peak = 2x, weekdays 01:00-04:00 and
+# 06:00-10:00 UTC; weekends off-peak); costs below are the off-peak rate.
+# Every response reports its exact cost in the `x-cost-usd` header.
+# https://above.dev/llms.txt
+api = "https://api.above.dev/v1"
+doc = "https://above.dev/docs"

From dcd590f2ac7182d0a67734f4fc9a0cde7de35d2e Mon Sep 17 00:00:00 2001
From: Lee-Si-Yoon <siyoon@friendli.ai>
Date: Sun, 30 Aug 2026 12:36:49 +0900
Subject: [PATCH 0241/2012] feat(friendli): add GLM-5.3 provider model (#5776)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(friendli): add GLM-5.3 provider model

Friendli serverless now serves zai-org/GLM-5.3, base_model'd onto the
existing zhipuai/glm-5.3 lab entry. Effort-only reasoning options per
Friendli's documented reasoning_effort=low|high|max, limit/modalities
inherited from base.

* fix(friendli): correct GLM-5.3 reasoning_options and provider.toml docs

- Removed duplicate name field (matched models/zhipuai/glm-5.3.toml
  exactly; base_model already inherits it).
- provider.toml: removed the stale "no effort or budget field is
  documented" claim. Friendli's chat-completions endpoint documents
  both reasoning_effort (enum) and reasoning_budget (integer) as real
  request fields — https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
  and #body-reasoning-budget-one-of-0.
- Added budget_tokens (min=-1, max=1_048_576) to GLM-5.3's
  reasoning_options, matching live GET /serverless/v1/models
  (max_completion_tokens=1_048_576).
- Added [limit] override (context=1_048_576, output=1_048_576): live
  context_length/max_completion_tokens exceed the inherited lab file's
  rounded 1_000_000 figure.

* fix(friendli): prove budget_tokens is a real reasoning-only cap on GLM-5.3

Verified via two live POST /chat/completions requests against
zai-org/GLM-5.3: reasoning_budget=30 and =50 (with generous max_tokens=
3000/4000) both cut reasoning_content mid-sentence at the requested cap,
while completion_tokens continued to 1553/177 respectively -- proof the
reasoning phase is capped independently of completion/context capacity.
Also clarified that min=-1/max=1_048_576 are Friendli's own reported
values from GET /serverless/v1/models, not values models.dev derived
from max_completion_tokens itself.
---
 .../friendli/models/zai-org/GLM-5.3.toml      | 42 +++++++++++++++++++
 providers/friendli/provider.toml              | 10 +++--
 2 files changed, 49 insertions(+), 3 deletions(-)
 create mode 100644 providers/friendli/models/zai-org/GLM-5.3.toml

diff --git a/providers/friendli/models/zai-org/GLM-5.3.toml b/providers/friendli/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..bf8c8ce9c16
--- /dev/null
+++ b/providers/friendli/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,42 @@
+# Friendli's chat-completions endpoint documents reasoning_effort (generic
+# enum) and reasoning_budget (integer token cap) as real request fields; GLM-5.3
+# on Z.AI's own API always reasons and only exposes low|high|max, so we mirror
+# those graded levels here instead of Friendli's full generic effort enum.
+#
+# budget_tokens is a REAL, independently-enforced reasoning-token budget, not
+# derived from max_tokens/context capacity: two live POST /chat/completions
+# requests against this exact model (reasoning_budget=30 and =50, generous
+# max_tokens=3000/4000) cut reasoning_content mid-sentence at the requested
+# cap while completion_tokens continued to 1553/177 respectively — proof the
+# reasoning phase and the completion phase are capped independently.
+# min=-1/max=1_048_576 are Friendli's own reported values, read verbatim from
+# GET /serverless/v1/models -> reasoning_options -> {type: budget_tokens} for
+# zai-org/GLM-5.3 (not computed by us from max_completion_tokens; that field
+# happens to equal this model's max_completion_tokens/context_length, but we
+# pass through the catalog's own budget_tokens object, we do not derive it).
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
+# (accessed 2026-08-29)
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = -1
+max = 1_048_576
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
diff --git a/providers/friendli/provider.toml b/providers/friendli/provider.toml
index 7b619647048..38b3260c369 100644
--- a/providers/friendli/provider.toml
+++ b/providers/friendli/provider.toml
@@ -1,7 +1,11 @@
 name = "Friendli"
-# Reasoning HTTP format (accessed 2026-06-25): POST /serverless/v1/chat/completions.
-# Toggle: chat_template_kwargs.enable_thinking = true | false; support is
-# model-specific. No effort or reasoning-token budget field is documented.
+# Reasoning HTTP format (accessed 2026-08-29): POST /serverless/v1/chat/completions.
+# Toggle: chat_template_kwargs.enable_thinking = true | false; support is model-specific.
+# Effort: reasoning_effort = minimal|low|medium|high|xhigh|max|ultracode; reasoning
+# models only, available options depend on the model.
+# Budget: reasoning_budget (positive integer reasoning-token cap); reasoning models only.
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
 # https://friendli.ai/docs/guides/reasoning
 env = ["FRIENDLI_TOKEN"]
 npm = "@ai-sdk/openai-compatible"

From 57a27f36cd0ca188a47a015c318cf9c1a5ddd965 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 22:37:09 -0500
Subject: [PATCH 0242/2012] chore(sync): update Requesty model catalog (#5770)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/glm-5.3-flash@eu.toml | 17 -----------------
 providers/requesty/models/glm-5.3@eu.toml       |  5 ++++-
 2 files changed, 4 insertions(+), 18 deletions(-)
 delete mode 100644 providers/requesty/models/glm-5.3-flash@eu.toml

diff --git a/providers/requesty/models/glm-5.3-flash@eu.toml b/providers/requesty/models/glm-5.3-flash@eu.toml
deleted file mode 100644
index d58a76089fe..00000000000
--- a/providers/requesty/models/glm-5.3-flash@eu.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "zhipuai/glm-5.3-flash"
-name = "GLM-5.3-Flash (EU)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-
-[cost]
-input = 0.2
-output = 0.5
-cache_read = 0.07
-
-[limit]
-output = 262_144
diff --git a/providers/requesty/models/glm-5.3@eu.toml b/providers/requesty/models/glm-5.3@eu.toml
index a5a072b72d9..0b46fbb34c4 100644
--- a/providers/requesty/models/glm-5.3@eu.toml
+++ b/providers/requesty/models/glm-5.3@eu.toml
@@ -8,7 +8,10 @@ name = "GLM-5.3 (EU)"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "max"]
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.75

From 1e0bb5f743acad49606704db6cf6e2a3a71bd6f6 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Sun, 30 Aug 2026 11:39:05 +0800
Subject: [PATCH 0243/2012] feat(opencode): remove Hy3 from free

---
 providers/opencode/models/hy3-free.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/opencode/models/hy3-free.toml b/providers/opencode/models/hy3-free.toml
index 7b88cc2c0a2..ebf95f43868 100644
--- a/providers/opencode/models/hy3-free.toml
+++ b/providers/opencode/models/hy3-free.toml
@@ -2,6 +2,7 @@ base_model = "tencent/hy3"
 name = "Hy3 Free"
 family = "hy3-free"
 structured_output = true
+status = "deprecated"
 
 reasoning_options = [
   { type = "toggle" },

From c9d25239649947860e080dd2537fb73656eae275 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 04:28:21 +0000
Subject: [PATCH 0244/2012] chore(sync): update Kilo model catalog (#5846)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 25dc6e6dd90..a41e9047daa 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.14
+output = 0.16
 cache_read = 0.01
 
 [limit]

From 1644b937820acd7eb09bb39e9c43a0218c4282e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 04:28:24 +0000
Subject: [PATCH 0245/2012] chore(sync): update OpenRouter model catalog
 (#5847)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-v4-flash-latest.toml   | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 364a0a34436..3ef2e65161d 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.14
+output = 0.16
 cache_read = 0.01
 
 [limit]

From 5d11a14e03f95baf539325fd889246a1983a7f00 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 04:28:27 +0000
Subject: [PATCH 0246/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#5845)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
index f8b5d01f800..5a0a6c35b26 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -20,7 +20,7 @@ output = 0.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 1_048_576
 
 [modalities]

From ff4bb8eba4fb758fd077e7e2b4ac798fd6846726 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 04:28:39 +0000
Subject: [PATCH 0247/2012] chore(sync): update NanoGPT model catalog (#5848)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 6514b80a004a3a7aaf160f33fc1742aa3489a752 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 05:24:15 +0000
Subject: [PATCH 0248/2012] chore(sync): update Eden AI model catalog (#5849)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/Muse-Glimmer-30B.toml       | 2 +-
 providers/edenai/models/mistral/codestral-latest.toml      | 1 +
 providers/edenai/models/mistral/mistral-medium-2604.toml   | 1 +
 providers/edenai/models/mistral/mistral-medium-latest.toml | 1 +
 providers/edenai/models/mistral/mistral-small-latest.toml  | 1 +
 5 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
index 1b542648bee..df75a46ae61 100644
--- a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
+++ b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
@@ -8,4 +8,4 @@ values = ["low", "medium", "high", "xhigh"]
 
 [cost]
 input = 0.3
-output = 1.2
+output = 1.1
diff --git a/providers/edenai/models/mistral/codestral-latest.toml b/providers/edenai/models/mistral/codestral-latest.toml
index 7046f48c1d3..b0b137235e1 100644
--- a/providers/edenai/models/mistral/codestral-latest.toml
+++ b/providers/edenai/models/mistral/codestral-latest.toml
@@ -4,3 +4,4 @@ structured_output = true
 [cost]
 input = 0.3
 output = 0.9
+cache_read = 0.03
diff --git a/providers/edenai/models/mistral/mistral-medium-2604.toml b/providers/edenai/models/mistral/mistral-medium-2604.toml
index 811c7ffe231..be9a882e3ab 100644
--- a/providers/edenai/models/mistral/mistral-medium-2604.toml
+++ b/providers/edenai/models/mistral/mistral-medium-2604.toml
@@ -7,3 +7,4 @@ values = ["none", "high"]
 [cost]
 input = 1.5
 output = 7.5
+cache_read = 0.15
diff --git a/providers/edenai/models/mistral/mistral-medium-latest.toml b/providers/edenai/models/mistral/mistral-medium-latest.toml
index 1f83fe33308..66bfb523c22 100644
--- a/providers/edenai/models/mistral/mistral-medium-latest.toml
+++ b/providers/edenai/models/mistral/mistral-medium-latest.toml
@@ -7,3 +7,4 @@ values = ["none", "high"]
 [cost]
 input = 1.5
 output = 7.5
+cache_read = 0.15
diff --git a/providers/edenai/models/mistral/mistral-small-latest.toml b/providers/edenai/models/mistral/mistral-small-latest.toml
index dd09544702f..59ec1c034c7 100644
--- a/providers/edenai/models/mistral/mistral-small-latest.toml
+++ b/providers/edenai/models/mistral/mistral-small-latest.toml
@@ -8,6 +8,7 @@ values = ["none", "high"]
 [cost]
 input = 0.15
 output = 0.6
+cache_read = 0.015
 
 [limit]
 context = 262_144

From da305feaaf564a2175ad895add957b778be3809e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 05:24:33 +0000
Subject: [PATCH 0249/2012] chore(sync): update Kilo model catalog (#5850)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/arcee-ai/trinity-large-thinking.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/arcee-ai/trinity-large-thinking.toml b/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
index 6640f8c04b2..9410f5822bf 100644
--- a/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
+++ b/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
@@ -12,4 +12,4 @@ cache_read = 0.06
 
 [limit]
 context = 262_144
-output = 235_929
+output = 80_000

From d595f6f351974d6b876992f4d336822df9e78a2d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 05:24:37 +0000
Subject: [PATCH 0250/2012] chore(sync): update OpenRouter model catalog
 (#5851)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/arcee-ai/trinity-large-thinking.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml b/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
index 4fcf49f5d9d..e6f6fe97d08 100644
--- a/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
+++ b/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
@@ -3,10 +3,10 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 0.22
-output = 0.85
+input = 0.25
+output = 0.8
 cache_read = 0.06
 
 [limit]
 context = 262_144
-output = 235_929
+output = 80_000
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 5bbf73f7cb8..ef7d1b3c36f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.498336
-output = 0.996672
-cache_read = 0.041528
+input = 0.488244
+output = 0.976488
+cache_read = 0.040687
 
 [limit]
 context = 1_048_576

From 24dab0f492b1f9551c89bd0312d135471802d34a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 06:36:28 +0000
Subject: [PATCH 0251/2012] chore(sync): update NanoGPT model catalog (#5853)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From fcabf0c30473b6ed6dd780d15a4ba0252688dc6b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 06:36:38 +0000
Subject: [PATCH 0252/2012] chore(sync): update OpenRouter model catalog
 (#5852)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ef7d1b3c36f..d5dbc00176a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.488244
-output = 0.976488
-cache_read = 0.040687
+input = 0.477978
+output = 0.955956
+cache_read = 0.039832
 
 [limit]
 context = 1_048_576

From bb1226f159aea6fba6c120b7fd5c6568b71de423 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 07:25:20 +0000
Subject: [PATCH 0253/2012] chore(sync): update OpenRouter model catalog
 (#5854)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index d5dbc00176a..48763b37510 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.477978
-output = 0.955956
-cache_read = 0.039832
+input = 0.467886
+output = 0.935772
+cache_read = 0.038991
 
 [limit]
 context = 1_048_576

From d32a4e537199bf430b7df5828357a50ab106302e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 08:29:02 +0000
Subject: [PATCH 0254/2012] chore(sync): update OpenRouter model catalog
 (#5855)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 48763b37510..cd9c6346855 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.467886
-output = 0.935772
-cache_read = 0.038991
+input = 0.457794
+output = 0.915588
+cache_read = 0.03815
 
 [limit]
 context = 1_048_576

From 01cab517dcc994e2beca9fed996f32d525546ad2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 09:25:20 +0000
Subject: [PATCH 0255/2012] chore(sync): update OpenRouter model catalog
 (#5856)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3081f90ea05..a94ef179f2d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08064
-output = 0.16128
-cache_read = 0.016128
+input = 0.08036
+output = 0.16072
+cache_read = 0.016072
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cd9c6346855..510c1a09706 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.457794
-output = 0.915588
-cache_read = 0.03815
+input = 0.447528
+output = 0.895056
+cache_read = 0.037294
 
 [limit]
 context = 1_048_576

From 526a966f31067f805e4239542b8a2fdbdaa7b429 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 10:22:59 +0000
Subject: [PATCH 0256/2012] chore(sync): update OpenRouter model catalog
 (#5858)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a94ef179f2d..ee4e06c18ff 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08036
-output = 0.16072
-cache_read = 0.016072
+input = 0.08008
+output = 0.16016
+cache_read = 0.016016
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 510c1a09706..f9971fdc0e5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.447528
-output = 0.895056
-cache_read = 0.037294
+input = 0.437436
+output = 0.874872
+cache_read = 0.036453
 
 [limit]
 context = 1_048_576

From 5b087ae043b7de0ef823c916ac96595c9bbd1365 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 10:23:03 +0000
Subject: [PATCH 0257/2012] chore(sync): update NanoGPT model catalog (#5859)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 39b1610e2d89c5256131e51ddb6e5cd45a5f006f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 11:21:26 +0000
Subject: [PATCH 0258/2012] chore(sync): update OpenRouter model catalog
 (#5861)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ee4e06c18ff..5ee096f845c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08008
-output = 0.16016
-cache_read = 0.016016
+input = 0.0798
+output = 0.1596
+cache_read = 0.01596
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f9971fdc0e5..4be20a12281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.437436
-output = 0.874872
-cache_read = 0.036453
+input = 0.427344
+output = 0.854688
+cache_read = 0.035612
 
 [limit]
 context = 1_048_576

From 6b3710072a6cf9ed18b603bcb079dc3431e1149a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 12:32:27 +0000
Subject: [PATCH 0259/2012] chore(sync): update Kilo model catalog (#5863)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/mistralai/devstral-2512.toml          | 6 +++---
 providers/kilo/models/mistralai/mistral-nemo.toml           | 5 ++---
 .../models/mistralai/mistral-small-3.2-24b-instruct.toml    | 5 ++---
 3 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/mistralai/devstral-2512.toml b/providers/kilo/models/mistralai/devstral-2512.toml
index 2df4113f1ab..33109af1642 100644
--- a/providers/kilo/models/mistralai/devstral-2512.toml
+++ b/providers/kilo/models/mistralai/devstral-2512.toml
@@ -4,9 +4,9 @@ attachment = true
 structured_output = true
 
 [cost]
-input = 0.44
-output = 2.2
-cache_read = 0.044
+input = 0.4
+output = 2
+cache_read = 0.04
 
 [limit]
 output = 209_715
diff --git a/providers/kilo/models/mistralai/mistral-nemo.toml b/providers/kilo/models/mistralai/mistral-nemo.toml
index 818e29c051d..05d2aff530a 100644
--- a/providers/kilo/models/mistralai/mistral-nemo.toml
+++ b/providers/kilo/models/mistralai/mistral-nemo.toml
@@ -3,9 +3,8 @@ description = "Mistral model for multilingual chat, reasoning, and tool-assisted
 structured_output = true
 
 [cost]
-input = 0.165
-output = 0.165
-cache_read = 0.0165
+input = 0.019
+output = 0.03
 
 [limit]
 context = 131_072
diff --git a/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml b/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
index 2ae0dda2905..68645a3ea09 100644
--- a/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
+++ b/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
@@ -11,9 +11,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.11
-output = 0.33
-cache_read = 0.011
+input = 0.075
+output = 0.2
 
 [limit]
 context = 128_000

From f9e6b92ef9c50aca95bc6986007c1097b9591892 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 12:32:29 +0000
Subject: [PATCH 0260/2012] chore(sync): update OpenRouter model catalog
 (#5864)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/mistralai/devstral-2512.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5ee096f845c..f56ac3460d0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0798
-output = 0.1596
-cache_read = 0.01596
+input = 0.07952
+output = 0.15904
+cache_read = 0.015904
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4be20a12281..05dce4765d6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.427344
-output = 0.854688
-cache_read = 0.035612
+input = 0.417252
+output = 0.834504
+cache_read = 0.034771
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/mistralai/devstral-2512.toml b/providers/openrouter/models/mistralai/devstral-2512.toml
index 4ac16feadc2..91948161c41 100644
--- a/providers/openrouter/models/mistralai/devstral-2512.toml
+++ b/providers/openrouter/models/mistralai/devstral-2512.toml
@@ -4,9 +4,9 @@ attachment = true
 structured_output = true
 
 [cost]
-input = 0.44
-output = 2.2
-cache_read = 0.044
+input = 0.4
+output = 2
+cache_read = 0.04
 
 [limit]
 output = 209_715

From 6b7d151c79efdd36aca4374e0cabd13c2d3704c2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 14:22:51 +0000
Subject: [PATCH 0261/2012] chore(sync): update OpenRouter model catalog
 (#5867)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f56ac3460d0..e91096ab533 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07952
-output = 0.15904
-cache_read = 0.015904
+input = 0.07924
+output = 0.15848
+cache_read = 0.015848
 
 [limit]
 context = 1_048_576

From c1b23c9984ab257d122b87ce79cefd2b1958ff97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 15:22:46 +0000
Subject: [PATCH 0262/2012] chore(sync): update Eden AI model catalog (#5869)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/mistral/devstral-2512.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/edenai/models/mistral/devstral-2512.toml b/providers/edenai/models/mistral/devstral-2512.toml
index 53cfd98aa8f..dc55a17b584 100644
--- a/providers/edenai/models/mistral/devstral-2512.toml
+++ b/providers/edenai/models/mistral/devstral-2512.toml
@@ -3,9 +3,9 @@ attachment = true
 structured_output = true
 
 [cost]
-input = 0.44
-output = 2.2
-cache_read = 0.044
+input = 0.4
+output = 2
+cache_read = 0.04
 
 [modalities]
 input = ["text", "pdf"]

From ea7674463ad849be3655bf605bc32077554849a4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 15:22:52 +0000
Subject: [PATCH 0263/2012] chore(sync): update OpenRouter model catalog
 (#5868)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e91096ab533..3e8301fafee 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07924
-output = 0.15848
-cache_read = 0.015848
+input = 0.07896
+output = 0.15792
+cache_read = 0.015792
 
 [limit]
 context = 1_048_576

From 47a67a2149464b40e9958b69cb22fb38a9f88f06 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 16:24:29 +0000
Subject: [PATCH 0264/2012] chore(sync): update OpenRouter model catalog
 (#5870)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 7d66223d4311abed783b3d1e5bcda70455723f77 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 16:24:44 +0000
Subject: [PATCH 0265/2012] chore(sync): update Kilo model catalog (#5871)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 103f738879ec8614cafa9c632d86a5fe51355e76 Mon Sep 17 00:00:00 2001
From: Rubin Simons <me@rubin55.org>
Date: Sun, 30 Aug 2026 18:53:01 +0200
Subject: [PATCH 0266/2012] Update neuralwatt catalog, GLM-5.3 officially open
 weights (#5860)

* fix(models): mark GLM-5.3 as open weights

Sources:
  * https://z.ai/blog/glm-5.3
  * https://news.ycombinator.com/item?id=49479878
  * https://twitter.com/Zai_org/status/2093354097122455713

* feat(neuralwatt): add GLM 5.3

Neuralwatt serves glm-5.3 at 1.45 input and 4.50 output per million
tokens, with effort levels low, high and max, and it accepts
thinking_token_budget. The entry inherits the rest from
models/zhipuai/glm-5.3.toml.

* chore(neuralwatt): normalise existing catalog entries

Rewritten by the upcoming sync provider. The resolved catalog is
unchanged: reasoning_options move to table syntax, fields that
already match the base model are dropped, and the default text
output modality is left implicit.

* fix(neuralwatt): mark GLM 5.3 as a gated preview

The API description flags glm-5.3 as a gated preview whose price is a
GLM 5.2 parity default pending review at launch, matching how
deepseek-v4-pro and qwen-3.8-27b are already recorded. Adds the source
header and drops the output cap, which the API does not state, so it
inherits 131072 from the lab entry.

* fix(neuralwatt): restore the GLM 5.3 output cap

Neuralwatt reports no max_output_tokens for glm-5.3, which the provider
README records as capped only by the context window and cataloged with
output equal to context. Lists glm-5.3 in that README and notes it as
the one model with no none effort level.
---
 models/zhipuai/glm-5.3.toml                   |  2 +-
 providers/neuralwatt/README.md                |  4 ++-
 providers/neuralwatt/models/glm-5.2-fast.toml | 27 +++++++---------
 providers/neuralwatt/models/glm-5.2-flex.toml | 27 +++++++---------
 .../models/glm-5.2-short-fast-flex.toml       | 27 +++++++---------
 .../neuralwatt/models/glm-5.2-short-fast.toml | 27 +++++++---------
 .../neuralwatt/models/glm-5.2-short-flex.toml | 27 +++++++---------
 .../neuralwatt/models/glm-5.2-short.toml      | 27 +++++++---------
 providers/neuralwatt/models/glm-5.2.toml      | 27 +++++++---------
 providers/neuralwatt/models/glm-5.3.toml      | 32 +++++++++++++++++++
 .../models/kimi-k2.7-code-fast.toml           |  8 +++--
 .../models/kimi-k2.7-code-flex.toml           |  8 +++--
 .../neuralwatt/models/kimi-k2.7-code.toml     |  8 +++--
 providers/neuralwatt/models/kimi-k3-fast.toml |  5 ++-
 providers/neuralwatt/models/kimi-k3-flex.toml | 15 +++++----
 providers/neuralwatt/models/kimi-k3.toml      | 17 ++++++----
 .../neuralwatt/models/qwen3.6-35b-fast.toml   |  7 +---
 providers/neuralwatt/models/qwen3.6-35b.toml  | 15 +++++----
 18 files changed, 159 insertions(+), 151 deletions(-)
 create mode 100644 providers/neuralwatt/models/glm-5.3.toml

diff --git a/models/zhipuai/glm-5.3.toml b/models/zhipuai/glm-5.3.toml
index 948bc6a3882..bb71386ebb3 100644
--- a/models/zhipuai/glm-5.3.toml
+++ b/models/zhipuai/glm-5.3.toml
@@ -8,7 +8,7 @@ reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_000_000
diff --git a/providers/neuralwatt/README.md b/providers/neuralwatt/README.md
index f27507bb677..84a3b5f5cc7 100644
--- a/providers/neuralwatt/README.md
+++ b/providers/neuralwatt/README.md
@@ -14,6 +14,7 @@ Standard Models
 - gemma-4-31b: Gemma 4 31B, image input, 262K context
 - glm-5.2: GLM 5.2, 1M context
 - glm-5.2-short: GLM 5.2 Short, 200K context, 32K output cap
+- glm-5.3: GLM 5.3, 1M context, gated preview
 - kimi-k2.7-code: Kimi K2.7 Code, image input, 262K context
 - kimi-k3: Kimi K3, image input, 1M context
 - qwen-3.8-27b: Qwen 3.8 27B, image input, 262K context, private preview
@@ -43,7 +44,8 @@ Pricing
 
 Reasoning Controls
 - reasoning_effort accepts the levels in metadata.reasoning.supported_efforts.
-  Every list that offers reasoning also offers none, which is the off switch.
+  Most lists include none, which is the off switch. glm-5.3 is the exception:
+  it reports reasoning.mandatory and offers only low, high and max.
 - accepted_efforts is wider than supported_efforts. The extra values are
   aliases the API folds into a supported level, not distinct levels.
 - thinking_token_budget is accepted on every model except deepseek-v4-flash and
diff --git a/providers/neuralwatt/models/glm-5.2-fast.toml b/providers/neuralwatt/models/glm-5.2-fast.toml
index f43574adc89..a61cf125702 100644
--- a/providers/neuralwatt/models/glm-5.2-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-fast.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Fast"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.45
@@ -24,7 +23,3 @@ cache_read = 0.145
 [limit]
 context = 1_048_560
 output = 1_048_560
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/glm-5.2-flex.toml b/providers/neuralwatt/models/glm-5.2-flex.toml
index 47465b4023f..e42b7fb1cc5 100644
--- a/providers/neuralwatt/models/glm-5.2-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-flex.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Flex"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 0.9425
@@ -24,7 +23,3 @@ cache_read = 0.09425
 [limit]
 context = 1_048_560
 output = 1_048_560
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
index 36151854e7c..eea25d5f58e 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Fast Flex"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 0.9425
@@ -24,7 +23,3 @@ cache_read = 0.09425
 [limit]
 context = 199_984
 output = 32_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast.toml b/providers/neuralwatt/models/glm-5.2-short-fast.toml
index 12d05d328dc..fa48e039e59 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Fast"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.45
@@ -24,7 +23,3 @@ cache_read = 0.145
 [limit]
 context = 199_984
 output = 32_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/glm-5.2-short-flex.toml b/providers/neuralwatt/models/glm-5.2-short-flex.toml
index 1e96b70f2ef..604df60b11d 100644
--- a/providers/neuralwatt/models/glm-5.2-short-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-flex.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Flex"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 0.9425
@@ -24,7 +23,3 @@ cache_read = 0.09425
 [limit]
 context = 199_984
 output = 32_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2-short.toml b/providers/neuralwatt/models/glm-5.2-short.toml
index 857cc0d70cd..26127431e9c 100644
--- a/providers/neuralwatt/models/glm-5.2-short.toml
+++ b/providers/neuralwatt/models/glm-5.2-short.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.45
@@ -24,7 +23,3 @@ cache_read = 0.145
 [limit]
 context = 199_984
 output = 32_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.2.toml b/providers/neuralwatt/models/glm-5.2.toml
index 0f3f23cfd75..0b7b1191535 100644
--- a/providers/neuralwatt/models/glm-5.2.toml
+++ b/providers/neuralwatt/models/glm-5.2.toml
@@ -1,20 +1,19 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
+base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
-attachment = false
-reasoning = true
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high", "max"] },
-  { type = "budget_tokens" },
-]
-interleaved = true
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.45
@@ -24,7 +23,3 @@ cache_read = 0.145
 [limit]
 context = 1_048_560
 output = 1_048_560
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.3.toml b/providers/neuralwatt/models/glm-5.3.toml
new file mode 100644
index 00000000000..700f54f026e
--- /dev/null
+++ b/providers/neuralwatt/models/glm-5.3.toml
@@ -0,0 +1,32 @@
+# Preview rollout: glm-5.3 is a gated preview on Neuralwatt, visible to
+# accounts granted access per the Preview Models guide. Its list price is
+# Neuralwatt's GLM 5.2 parity default and is marked for review at launch,
+# per the model description in GET /v1/models (checked 2026-08-30).
+# Reasoning cannot be disabled (reasoning.mandatory = true), so there is no
+# none level and no toggle. The API supports low, high and max as distinct
+# efforts; minimal, medium and xhigh are accepted as aliases onto them.
+# Budget: thinking_token_budget (integer reasoning tokens)
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/docs/guides/preview-models
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3"
+structured_output = false
+status = "beta"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 1.45
+output = 4.5
+cache_read = 0.145
+
+[limit]
+context = 1_048_560
+output = 1_048_560
diff --git a/providers/neuralwatt/models/kimi-k2.7-code-fast.toml b/providers/neuralwatt/models/kimi-k2.7-code-fast.toml
index 5bb7381dad9..0799b2e3f23 100644
--- a/providers/neuralwatt/models/kimi-k2.7-code-fast.toml
+++ b/providers/neuralwatt/models/kimi-k2.7-code-fast.toml
@@ -1,10 +1,13 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "moonshotai/kimi-k2.7-code"
 name = "Kimi K2.7 Code Fast"
 description = "Kimi K2.7 Code with reasoning capped to a short budget for lower latency; reasoning cannot be disabled on this model"
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [{ type = "budget_tokens" }]
+
 interleaved = true
 
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.95
 output = 4
@@ -16,4 +19,3 @@ output = 262_128
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k2.7-code-flex.toml b/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
index cf9cefa44aa..6df952380fc 100644
--- a/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
+++ b/providers/neuralwatt/models/kimi-k2.7-code-flex.toml
@@ -1,9 +1,12 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "moonshotai/kimi-k2.7-code"
 name = "Kimi K2.7 Code Flex"
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [{ type = "budget_tokens" }]
+
 interleaved = true
 
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.6175
 output = 2.6
@@ -15,4 +18,3 @@ output = 262_128
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k2.7-code.toml b/providers/neuralwatt/models/kimi-k2.7-code.toml
index 099d845c043..13152a3a491 100644
--- a/providers/neuralwatt/models/kimi-k2.7-code.toml
+++ b/providers/neuralwatt/models/kimi-k2.7-code.toml
@@ -1,8 +1,11 @@
-base_model = "moonshotai/kimi-k2.7-code"
 # Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [{ type = "budget_tokens" }]
+base_model = "moonshotai/kimi-k2.7-code"
+
 interleaved = true
 
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.95
 output = 4
@@ -14,4 +17,3 @@ output = 262_128
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k3-fast.toml b/providers/neuralwatt/models/kimi-k3-fast.toml
index 0c1ebe32ec3..68cacc816ff 100644
--- a/providers/neuralwatt/models/kimi-k3-fast.toml
+++ b/providers/neuralwatt/models/kimi-k3-fast.toml
@@ -4,8 +4,8 @@ description = "Kimi K3 with thinking disabled for low-latency tool calling, visi
 reasoning = false
 
 [cost]
-input = 3.0
-output = 15.0
+input = 3
+output = 15
 cache_read = 0.3
 
 [limit]
@@ -14,4 +14,3 @@ output = 1_048_560
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k3-flex.toml b/providers/neuralwatt/models/kimi-k3-flex.toml
index 23752742a08..96234c9c8d1 100644
--- a/providers/neuralwatt/models/kimi-k3-flex.toml
+++ b/providers/neuralwatt/models/kimi-k3-flex.toml
@@ -1,13 +1,17 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "moonshotai/kimi-k3"
 name = "Kimi K3 Flex"
 description = "Kimi K3 on the flex tier: discounted, best-effort latency, requests may be held under load"
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "low", "high", "max"] },
-  { type = "budget_tokens" },
-]
+
 interleaved = true
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 1.95
 output = 9.75
@@ -19,4 +23,3 @@ output = 1_048_560
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/kimi-k3.toml b/providers/neuralwatt/models/kimi-k3.toml
index ce4d8d81577..06cfeff6fc6 100644
--- a/providers/neuralwatt/models/kimi-k3.toml
+++ b/providers/neuralwatt/models/kimi-k3.toml
@@ -1,15 +1,19 @@
 # Effort: reasoning_effort = none|low|high|max (none disables thinking)
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "moonshotai/kimi-k3"
-reasoning_options = [
-  { type = "effort", values = ["none", "low", "high", "max"] },
-  { type = "budget_tokens" },
-]
+
 interleaved = true
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
-input = 3.0
-output = 15.0
+input = 3
+output = 15
 cache_read = 0.3
 
 [limit]
@@ -18,4 +22,3 @@ output = 1_048_560
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/neuralwatt/models/qwen3.6-35b-fast.toml b/providers/neuralwatt/models/qwen3.6-35b-fast.toml
index c0536ac3b67..46634b0e8e7 100644
--- a/providers/neuralwatt/models/qwen3.6-35b-fast.toml
+++ b/providers/neuralwatt/models/qwen3.6-35b-fast.toml
@@ -1,14 +1,10 @@
+base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B Fast"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 family = "qwen3.6"
 release_date = "2026-04-01"
 last_updated = "2026-04-01"
-attachment = true
 reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.29
@@ -21,4 +17,3 @@ output = 131_056
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/neuralwatt/models/qwen3.6-35b.toml b/providers/neuralwatt/models/qwen3.6-35b.toml
index 3daca1e8ce2..efe4eb1711b 100644
--- a/providers/neuralwatt/models/qwen3.6-35b.toml
+++ b/providers/neuralwatt/models/qwen3.6-35b.toml
@@ -1,12 +1,16 @@
+# Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B"
-# Budget: thinking_token_budget (integer reasoning tokens)
-reasoning_options = [
-  { type = "effort", values = ["none", "high"] },
-  { type = "budget_tokens" },
-]
+
 interleaved = true
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.29
 output = 1.15
@@ -18,4 +22,3 @@ output = 131_056
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]

From 167dda913d01658e17be4522b4d4ba8000cfb6ce Mon Sep 17 00:00:00 2001
From: Christian Landgren <christian@landgren.nu>
Date: Sun, 30 Aug 2026 18:53:12 +0200
Subject: [PATCH 0267/2012] feat(berget): add GLM-5.3 and GLM-5.3-Flash (#5797)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(berget): add GLM-5.3 and GLM-5.3-Flash

Both served on Berget AI's Swedish infrastructure (NVIDIA B300, SGLang
TP4, 4 GPUs each on a shared node):

- zai-org/GLM-5.3 (753B, text): extends zai/glm-5.3, 1.5/5.0 EUR/M
- zai-org/GLM-5.3-Flash (320B/18B active, text+image): extends
  zai/glm-5.3-flash, 0.25/0.5 EUR/M

Context advertised at 327K (needle-validated; native 1M). Reasoning
always on, effort low/high/max, returned in reasoning_content — inherited
from the base models.

* fix(berget): use base_model + explicit reasoning_options

The extends target (zai/glm-5.3) is itself only a cost-override entry
without the required name/description/attachment/reasoning/tool_call/
open_weights fields, so validation failed. Switch to base_model =
zhipuai/glm-5.3 (the canonical model definition) and add the
reasoning_options + interleaved fields explicitly, matching the Baseten
entries for the same models.

* fix(berget): USD costs, modalities, release dates per review

- Costs converted to USD at 1.1643 (ECB, 2026-08-28) with EUR source
  prices + rate/date in leading comments: GLM-5.3 1.75/5.82, Flash
  0.29/0.58 USD/MTok.
- Flash modalities: text + image + video (video verified against the
  live endpoint; PDF not supported by the SGLang serving stack).
- release_date omitted so lab dates inherit; last_updated kept as the
  host entry date.

* fix(berget): drop redundant modalities override on GLM-5.3

The [modalities] block restated input/output = text, identical to the
lab base (zhipuai/glm-5.3) — override-only rule violation. Let it
inherit. (GLM-5.3-Flash keeps its override: it drops pdf, which the
Berget serving stack does not support.)
---
 .../berget/models/zai-org/GLM-5.3-Flash.toml  | 33 +++++++++++++++++++
 providers/berget/models/zai-org/GLM-5.3.toml  | 26 +++++++++++++++
 2 files changed, 59 insertions(+)
 create mode 100644 providers/berget/models/zai-org/GLM-5.3-Flash.toml
 create mode 100644 providers/berget/models/zai-org/GLM-5.3.toml

diff --git a/providers/berget/models/zai-org/GLM-5.3-Flash.toml b/providers/berget/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..6550f13b057
--- /dev/null
+++ b/providers/berget/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,33 @@
+# GLM-5.3-Flash served on Berget AI's Swedish infrastructure (NVIDIA B300,
+# SGLang TP4). Multimodal (text + image), reasoning on by default with effort
+# levels low/high/max. Needle-validated to 327K context.
+# https://api.berget.ai/openapi.json
+#
+# Berget bills in EUR: 0.25 EUR/M input, 0.5 EUR/M output.
+# Converted to USD at 1.1643 (ECB rate, 2026-08-28):
+#   0.25 EUR/M -> 0.29 USD/M input
+#   0.5 EUR/M  -> 0.58 USD/M output
+base_model = "zhipuai/glm-5.3-flash"
+last_updated = "2026-08-29"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.29
+output = 0.58
+
+[limit]
+context = 327_680
+output = 16_384
+
+# The lab model accepts text/image/video/pdf; the Berget SGLang deployment
+# serves text + image + video (video verified against the live endpoint
+# 2026-08-29). PDF input is NOT supported by the serving stack.
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/berget/models/zai-org/GLM-5.3.toml b/providers/berget/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..355bc31b861
--- /dev/null
+++ b/providers/berget/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,26 @@
+# GLM-5.3 served on Berget AI's Swedish infrastructure (NVIDIA B300, SGLang
+# TP4). Reasoning always on, effort levels low/high/max (default max), returned
+# in message.reasoning_content. Needle-validated to 327K context.
+# https://api.berget.ai/openapi.json
+#
+# Berget bills in EUR: 1.5 EUR/M input, 5.0 EUR/M output.
+# Converted to USD at 1.1643 (ECB rate, 2026-08-28):
+#   1.5 EUR/M  -> 1.75 USD/M input
+#   5.0 EUR/M  -> 5.82 USD/M output
+base_model = "zhipuai/glm-5.3"
+last_updated = "2026-08-29"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.75
+output = 5.82
+
+[limit]
+context = 327_680
+output = 32_768

From 3e481492d092fd885b4e6c9879278d81cd819177 Mon Sep 17 00:00:00 2001
From: MicroHEROX <140800253+MicroHEROX@users.noreply.github.com>
Date: Mon, 31 Aug 2026 00:53:23 +0800
Subject: [PATCH 0268/2012] feat: add SenseNova (China) provider (#5787)

* test write access

* feat: add SenseNova (China) provider

* fix: align relay reasoning_options to lab-effective levels (none|high)
---
 providers/sensenova/logo.svg                  | 18 ++++++++++
 .../sensenova/models/deepseek-v4-flash.toml   | 23 ++++++++++++
 providers/sensenova/models/glm-5.2.toml       | 18 ++++++++++
 .../models/sensenova-6.8-flash-lite.toml      | 36 +++++++++++++++++++
 providers/sensenova/provider.toml             |  5 +++
 5 files changed, 100 insertions(+)
 create mode 100644 providers/sensenova/logo.svg
 create mode 100644 providers/sensenova/models/deepseek-v4-flash.toml
 create mode 100644 providers/sensenova/models/glm-5.2.toml
 create mode 100644 providers/sensenova/models/sensenova-6.8-flash-lite.toml
 create mode 100644 providers/sensenova/provider.toml

diff --git a/providers/sensenova/logo.svg b/providers/sensenova/logo.svg
new file mode 100644
index 00000000000..e9df39bda41
--- /dev/null
+++ b/providers/sensenova/logo.svg
@@ -0,0 +1,18 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 25.72 25.72" fill="currentColor">
+  <rect x="0.6" y="0.6" width="24.52" height="24.52" fill="none" stroke="currentColor" stroke-width="1.2"/>
+  <path d="M6.365503787994385,6.430479049682617L6.365503787994385,6.860865049682618L12.731003787994386,13.291349049682617L12.731003787994386,12.862159049682617L13.157043787994386,12.862159049682617L6.791541787994385,6.430479049682617L6.365503787994385,6.430479049682617Z"/>
+  <path d="M8.412155151367188,6.430479049682617L14.777655151367188,12.862159049682617L15.629735151367187,12.862159049682617L9.263037151367188,6.430479049682617L8.412155151367188,6.430479049682617Z"/>
+  <path d="M10.883649826049805,6.430479049682617L17.250349826049806,12.862159049682617L18.101229826049803,12.862159049682617L11.735724826049804,6.430479049682617L10.883649826049805,6.430479049682617Z"/>
+  <path d="M13.356339454650879,6.430479049682617L19.72183945465088,12.862159049682617L20.57272945465088,12.862159049682617L14.20722145465088,6.430479049682617L13.356339454650879,6.430479049682617Z"/>
+  <path d="M15.829028129577637,6.430479049682617L22.194528129577638,12.862159049682617L23.045418129577637,12.862159049682617L16.679910129577635,6.430479049682617L15.829028129577637,6.430479049682617Z"/>
+  <path d="M18.300525665283203,6.430479049682617L24.666025665283204,12.862159049682617L25.463205665283205,12.862159049682617L25.463205665283205,12.806709049682617L19.151406665283204,6.430479049682617L18.300525665283203,6.430479049682617Z"/>
+  <path d="M20.77202033996582,6.430479049682617L25.46321033996582,11.169549049682617L25.46321033996582,10.309979049682617L21.62290233996582,6.430479049682617L20.77202033996582,6.430479049682617Z"/>
+  <path d="M25.463209014892577,7.812059049682617L24.095591014892577,6.430479049682617L23.244709014892578,6.430479049682617L25.463209014892577,8.672829049682617L25.463209014892577,7.812059049682617Z"/>
+  <path d="M6.365503787994385,24.341545462768554L7.733123787994385,25.723117462768556L8.584003787994385,25.723117462768556L6.365503787994385,23.481977462768555L6.365503787994385,24.341545462768554Z"/>
+  <path d="M6.365503787994385,21.84482184338379L10.204613787994385,25.723118843383787L11.056693787994384,25.723118843383787L6.365503787994385,20.98404884338379L6.365503787994385,21.84482184338379Z"/>
+  <path d="M6.365503787994385,19.346893575683595L12.677303787994385,25.723117575683595L12.731003787994386,25.723117575683595L12.731003787994386,24.917807575683593L6.365503787994385,18.487327575683594L6.365503787994385,19.346893575683595Z"/>
+  <path d="M6.365503787994385,16.850171002624514L12.731003787994386,23.28064800262451L12.731003787994386,22.41987800262451L6.365503787994385,15.989398002624512L6.365503787994385,16.850171002624514Z"/>
+  <path d="M6.365503787994385,14.352241873901367L12.731003787994386,20.782723873901368L12.731003787994386,19.923153873901366L6.365503787994385,13.492673873901367L6.365503787994385,14.352241873901367Z"/>
+  <path d="M6.365503787994385,11.85551769885254L12.731003787994386,18.28600069885254L12.731003787994386,17.42643069885254L6.365503787994385,10.995950698852539L6.365503787994385,11.85551769885254Z"/>
+  <path d="M6.365503787994385,9.357589125793456L12.731003787994386,15.789271125793457L12.731003787994386,14.928501125793456L6.365503787994385,8.498021125793457L6.365503787994385,9.357589125793456Z"/>
+</svg>
diff --git a/providers/sensenova/models/deepseek-v4-flash.toml b/providers/sensenova/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..8fd30e82d9e
--- /dev/null
+++ b/providers/sensenova/models/deepseek-v4-flash.toml
@@ -0,0 +1,23 @@
+# Sources (accessed 2026-08-28):
+# https://platform.sensenova.cn/docs
+# Reasoning: host documents reasoning_effort none|low|medium|high (default
+# medium), but live calls show no grading: none disables reasoning and
+# low|medium|high produce indistinguishable reasoning volume (same
+# reasoning_content length within run variance). Aligned to the lab-effective
+# level (DeepSeek lab maps low/medium -> high); no toggle and no max on this
+# host.
+base_model = "deepseek/deepseek-v4-flash"
+reasoning_options = [
+  { type = "effort", values = ["none", "high"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+output = 65_536
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/sensenova/models/glm-5.2.toml b/providers/sensenova/models/glm-5.2.toml
new file mode 100644
index 00000000000..1cefde124b0
--- /dev/null
+++ b/providers/sensenova/models/glm-5.2.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-08-28):
+# https://platform.sensenova.cn/docs
+# Reasoning: host documents reasoning_effort none|low|medium|high (default
+# medium), but live calls show no grading: none disables reasoning and
+# low|medium|high produce indistinguishable reasoning volume. Aligned to the
+# lab-effective level (Z.AI GLM-5.2: high/max; max absent on this host).
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [
+  { type = "effort", values = ["none", "high"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/sensenova/models/sensenova-6.8-flash-lite.toml b/providers/sensenova/models/sensenova-6.8-flash-lite.toml
new file mode 100644
index 00000000000..ecf16a424ab
--- /dev/null
+++ b/providers/sensenova/models/sensenova-6.8-flash-lite.toml
@@ -0,0 +1,36 @@
+# Sources (accessed 2026-08-28):
+# https://platform.sensenova.cn/docs
+# GET /v1/models: context 262144, max output 65536, fp8, text+image -> text,
+# supported features tools+json_mode+reasoning; pricing all zero (public beta
+# free on TokenPlan).
+# Preview launched 2026-08-11: https://www.ithome.com/0/988/416.htm
+# Reasoning: reasoning_effort documented enum none|low|medium|high (default
+# medium). A reasoning side channel is returned as message.reasoning, which is
+# not a schema-supported interleaved field, so no interleaved is authored.
+
+name = "SenseNova 6.8 Flash Lite"
+description = "SenseNova lightweight multimodal agent model for real-world complex tasks, data analysis, and complex information presentation"
+release_date = "2026-08-11"
+last_updated = "2026-08-28"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = [
+  { type = "effort", values = ["none", "low", "medium", "high"] },
+]
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/sensenova/provider.toml b/providers/sensenova/provider.toml
new file mode 100644
index 00000000000..ed10eeab7f0
--- /dev/null
+++ b/providers/sensenova/provider.toml
@@ -0,0 +1,5 @@
+name = "SenseNova (China)"
+npm = "@ai-sdk/openai-compatible"
+env = ["SENSENOVA_API_KEY"]
+api = "https://token.sensenova.cn/v1"
+doc = "https://platform.sensenova.cn/docs"

From 23ff9fcfef5ef7498302b67757cca2d37fda50c7 Mon Sep 17 00:00:00 2001
From: realtimegb <support@realtimecomms.co.uk>
Date: Sun, 30 Aug 2026 17:54:03 +0100
Subject: [PATCH 0269/2012] above: add GLM 5.3 Flash (#5862)

* above: add glm-5.3-flash

Fireworks-served, sold at the Fireworks rate plus a flat 10%
(0.165/0.0319/0.55). Reasoning toggle verified (reasoning_effort
low|high; none rejected by this host), image input verified, video/pdf
excluded as host deltas.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

* above/glm-5.3-flash: match Fireworks effort surface (high|max)

Reviewer was right: earlier 'low disables thinking' was a sampling
artifact on trivial prompts. Re-tested with 3 trials per level on a
proof prompt: this is a thinking-only hybrid model; low/medium are
accepted on the wire but collapse per Fireworks docs, none is a 400.
Encoded as effort [high, max] matching the Fireworks peer, with the
wire behavior documented in the comment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014VuCev2YmeAs4WchNqw7uz

---------

Co-authored-by: realtimegb <pkreil3@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
---
 providers/above/models/glm-5.3-flash.toml | 24 +++++++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/above/models/glm-5.3-flash.toml

diff --git a/providers/above/models/glm-5.3-flash.toml b/providers/above/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..13710139108
--- /dev/null
+++ b/providers/above/models/glm-5.3-flash.toml
@@ -0,0 +1,24 @@
+# Passthrough to Fireworks serverless (accounts/fireworks/models/glm-5p3-flash).
+# Thinking-only model: reasoning cannot be disabled. Same surface as the
+# Fireworks peer entry: effort high|max (Fireworks collapses requested
+# low/medium to high; the wire accepts them but they select no distinct tier,
+# and "none" is rejected with a 400 — verified live 2026-08-30). Image input
+# verified; video/pdf are not supported on this host. Costs are the Fireworks
+# rate ($0.15/$0.029/$0.50) plus a flat 10%.
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.165
+output = 0.55
+cache_read = 0.0319

From 08b47b243faa18bc309413acf89e784b07eff047 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 17:20:49 +0000
Subject: [PATCH 0270/2012] chore(sync): update OpenRouter model catalog
 (#5873)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3e8301fafee..eed63a2025b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07896
-output = 0.15792
-cache_read = 0.015792
+input = 0.07868
+output = 0.15736
+cache_read = 0.015736
 
 [limit]
 context = 1_048_576

From ec8bc2ea05646c0513f39f41e71e74ca674102a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 17:21:03 +0000
Subject: [PATCH 0271/2012] chore(sync): update NanoGPT model catalog (#5875)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml     | 6 +++---
 .../nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml  | 6 +++---
 providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml      | 6 +++---
 .../nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml   | 6 +++---
 .../nano-gpt/models/abacusai/Dracarys-72B-Instruct.toml     | 4 ++--
 providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml  | 6 +++---
 providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml  | 6 +++---
 providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml  | 6 +++---
 providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml | 6 +++---
 providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml    | 6 +++---
 .../nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml     | 6 +++---
 .../nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml     | 6 +++---
 providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml   | 6 +++---
 .../models/gemma-4-26b-a4b-uncensored:thinking.toml         | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml      | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-fabled.toml        | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-garnet.toml        | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml      | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml      | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-isometry.toml      | 6 +++---
 providers/nano-gpt/models/gemma-4-31b-it-novelist.toml      | 6 +++---
 providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml    | 6 +++---
 providers/nano-gpt/models/google/gemma-4-31b-it.toml        | 6 +++---
 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml      | 6 +++---
 .../nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml   | 6 +++---
 .../nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml    | 2 +-
 .../models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml    | 2 +-
 providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml       | 6 +++---
 providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml | 6 +++---
 .../models/qwen/qwen3.8-27b-obliterated:thinking.toml       | 6 +++---
 providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml  | 6 +++---
 .../models/qwen/qwen3.8-27b-uncensored:thinking.toml        | 6 +++---
 .../nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml      | 6 +++---
 33 files changed, 94 insertions(+), 94 deletions(-)

diff --git a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
index 1c6aa9516ee..085bb305a2c 100644
--- a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
+++ b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
index 1e636658f07..bc7e93d87e3 100644
--- a/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
+++ b/providers/nano-gpt/models/Gemma-4-26B-A4B-MeroMero:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
index fd2ec3d29f6..c848256aa1a 100644
--- a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
+++ b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
index d7df31bd6ff..04d1ed8eadb 100644
--- a/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
+++ b/providers/nano-gpt/models/Gemma-4-31B-MeroMero-v2:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/abacusai/Dracarys-72B-Instruct.toml b/providers/nano-gpt/models/abacusai/Dracarys-72B-Instruct.toml
index fda08f61986..d00235b0748 100644
--- a/providers/nano-gpt/models/abacusai/Dracarys-72B-Instruct.toml
+++ b/providers/nano-gpt/models/abacusai/Dracarys-72B-Instruct.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 8_192
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
index 0869768f349..1b47787ffc4 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-chimerax.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
index 7b9406d64bb..126c906a1dc 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-darksoul.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
index 3fe6fa6df43..80a00f3714d 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-luminous.toml
@@ -10,9 +10,9 @@ structured_output = false
 open_weights = true
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
index 1128d4e0ca1..81a4963069c 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-moonlight.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
index 51a0a4560e2..210b4a169db 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-musica.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
index d28838392ec..dfc80e77d94 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-opusdistill.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
index df248b80165..d7ce2f13bd1 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-it-shadowsiren.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
index 4743aaa8aa7..e6d38df0eb8 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
index bda36ed39c3..f1ac4a30667 100644
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml b/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
index 808caed4c24..48214e90f31 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-darkidol.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml b/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
index 747250c6fbe..a307993afc4 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-fabled.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml b/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
index aee93f31708..5ae3faf4915 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-garnet.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml b/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
index 1f190959f1d..51fc85a0841 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-gembrain.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml b/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
index aa9c62abf42..06532309930 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-gemsicle.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml b/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
index 8f9c876eb5d..e5db9cc6bda 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-isometry.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml b/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
index 5379ce74bf6..f5ebe442424 100644
--- a/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
+++ b/providers/nano-gpt/models/gemma-4-31b-it-novelist.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
index bb5c61bc250..0603997c0b8 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
@@ -6,9 +6,9 @@ name = "Gemma 4 26B A4B"
 type = "toggle"
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.12
+output = 0.38
+cache_read = 0.06
 
 [limit]
 input = 262_144
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it.toml b/providers/nano-gpt/models/google/gemma-4-31b-it.toml
index 8fa48e25be1..e4866ce9cc0 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it.toml
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it.toml
@@ -6,9 +6,9 @@ name = "Gemma 4 31B"
 type = "toggle"
 
 [cost]
-input = 0.08
-output = 0.33
-cache_read = 0.04
+input = 0.1
+output = 0.45
+cache_read = 0.05
 
 [limit]
 input = 262_144
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
index 1c1e75a810c..43cf8a042d3 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.025
+input = 0.1
+output = 0.2
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
index cccf569b7dd..64843cc0f5d 100644
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.025
+input = 0.1
+output = 0.2
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
index 7760a7632c0..b9e509e8e09 100644
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
@@ -12,7 +12,7 @@ reasoning_options = []
 
 [cost]
 input = 0.15
-output = 0.5
+output = 0.95
 cache_read = 0.075
 
 [limit]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
index 0695f700ce0..de3e48de51a 100644
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
@@ -12,7 +12,7 @@ reasoning_options = []
 
 [cost]
 input = 0.15
-output = 0.5
+output = 0.95
 cache_read = 0.075
 
 [limit]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
index 4796ed09894..64b901ef6c3 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.2
-output = 1.4
-cache_read = 0.04
+input = 0.25
+output = 1.5
+cache_read = 0.125
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
index 80546d7654a..214e4ccf3b3 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.18
-output = 0.5
-cache_read = 0.075
+input = 0.25
+output = 1.5
+cache_read = 0.125
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
index d3bdfe9f86d..29772075d7d 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.18
-output = 0.5
-cache_read = 0.075
+input = 0.25
+output = 1.5
+cache_read = 0.125
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index efd6402590e..88a6536a1da 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.18
-output = 0.5
-cache_read = 0.075
+input = 0.25
+output = 1.5
+cache_read = 0.125
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 654490ff569..8dfeb5f15a4 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.18
-output = 0.5
-cache_read = 0.075
+input = 0.25
+output = 1.5
+cache_read = 0.125
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 031db2e4bbf..753a724d11b 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high"]
 
 [cost]
-input = 0.125
-output = 0.5
-cache_read = 0.0625
+input = 0.35
+output = 1.4
+cache_read = 0.175
 
 [limit]
 context = 262_144

From 2541b33acefadbf50cb1147fe53038355f06b8f5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 17:21:09 +0000
Subject: [PATCH 0272/2012] chore(sync): update Requesty model catalog (#5876)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/gpt-5.3-chat.toml | 30 ---------------------
 1 file changed, 30 deletions(-)
 delete mode 100644 providers/requesty/models/gpt-5.3-chat.toml

diff --git a/providers/requesty/models/gpt-5.3-chat.toml b/providers/requesty/models/gpt-5.3-chat.toml
deleted file mode 100644
index c2069c739f1..00000000000
--- a/providers/requesty/models/gpt-5.3-chat.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "gpt-5.3-chat"
-description = "GPT-5.3 Chat is an update to ChatGPT's most-used model that makes everyday conversations smoother, more useful, and more directly helpful. It delivers more accurate answers with better contextualization and significantly reduces unnecessary refusals, caveats, and overly cautious phrasing that can interrupt conversational flow."
-family = "gpt"
-release_date = "2026-03-18"
-last_updated = "2026-03-18"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From e152986657485e9f5ba9682c19fdfc84964e7146 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 17:21:12 +0000
Subject: [PATCH 0273/2012] chore(sync): update Venice model catalog (#5874)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/z-ai-glm-5-3.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/venice/models/z-ai-glm-5-3.toml b/providers/venice/models/z-ai-glm-5-3.toml
index 4d41a6a336e..2f84aeb13ab 100644
--- a/providers/venice/models/z-ai-glm-5-3.toml
+++ b/providers/venice/models/z-ai-glm-5-3.toml
@@ -3,6 +3,7 @@ name = "GLM 5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 release_date = "2026-08-18"
 last_updated = "2026-08-18"
+open_weights = false
 reasoning_options = []
 
 [cost]

From acf0c72328aa8294c4765c6e08756daedf405c7c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 18:26:55 +0000
Subject: [PATCH 0274/2012] chore(sync): update NanoGPT model catalog (#5878)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/nvidia/nemotron-3.5-lightning.toml     | 1 +
 .../nano-gpt/models/nvidia/nemotron-3.5-lightning:thinking.toml  | 1 +
 2 files changed, 2 insertions(+)

diff --git a/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning.toml b/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning.toml
index 3aece5779f7..e2f97741108 100644
--- a/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning.toml
@@ -1,5 +1,6 @@
 base_model = "nvidia/nemotron-3.5-lightning"
 name = "Nvidia Nemotron 3.5 Lightning"
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning:thinking.toml b/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning:thinking.toml
index b594cd96cdd..360ccdb1329 100644
--- a/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning:thinking.toml
+++ b/providers/nano-gpt/models/nvidia/nemotron-3.5-lightning:thinking.toml
@@ -1,5 +1,6 @@
 base_model = "nvidia/nemotron-3.5-lightning"
 name = "Nvidia Nemotron 3.5 Lightning Thinking"
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"

From d887749a8255fb1323b63a95fca41ac0f9ad29ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 18:27:00 +0000
Subject: [PATCH 0275/2012] chore(sync): update OpenRouter model catalog
 (#5877)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index eed63a2025b..3e8301fafee 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07868
-output = 0.15736
-cache_read = 0.015736
+input = 0.07896
+output = 0.15792
+cache_read = 0.015792
 
 [limit]
 context = 1_048_576

From 14ef007b85bf94afe92c48c558680bb7fc836897 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 19:20:05 +0000
Subject: [PATCH 0276/2012] chore(sync): update OpenRouter model catalog
 (#5880)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3e8301fafee..eed63a2025b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07896
-output = 0.15792
-cache_read = 0.015792
+input = 0.07868
+output = 0.15736
+cache_read = 0.015736
 
 [limit]
 context = 1_048_576

From 450479839cfee32024c70d2a0bd0dfa03866f5ea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 20:23:44 +0000
Subject: [PATCH 0277/2012] chore(sync): update Kilo model catalog (#5882)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-4.1-nano.toml | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-4.1-nano.toml b/providers/kilo/models/openai/gpt-4.1-nano.toml
index e2c5942f03b..c35fd0eb40b 100644
--- a/providers/kilo/models/openai/gpt-4.1-nano.toml
+++ b/providers/kilo/models/openai/gpt-4.1-nano.toml
@@ -6,8 +6,5 @@ input = 0.1
 output = 0.4
 cache_read = 0.025
 
-[limit]
-output = 942_818
-
 [modalities]
 input = ["image", "text", "pdf"]

From 314de3376e0b9b78f82ab182b0ad9d6bf70f205b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 20:23:46 +0000
Subject: [PATCH 0278/2012] chore(sync): update NanoGPT model catalog (#5881)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/claude-opus-4-1-thinking:32000.toml       | 12 ------------
 .../models/claude-opus-4-thinking:32000.toml         | 12 ------------
 2 files changed, 24 deletions(-)
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-1-thinking:32000.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-thinking:32000.toml

diff --git a/providers/nano-gpt/models/claude-opus-4-1-thinking:32000.toml b/providers/nano-gpt/models/claude-opus-4-1-thinking:32000.toml
deleted file mode 100644
index eed01381084..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-1-thinking:32000.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-1"
-name = "Claude 4.1 Opus Thinking (32K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-opus-4-thinking:32000.toml b/providers/nano-gpt/models/claude-opus-4-thinking:32000.toml
deleted file mode 100644
index ef6450e1339..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-thinking:32000.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-name = "Claude 4 Opus Thinking (32K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000

From 2d5ad65edad2d493672d8e7091afa6fb14a6c627 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 20:23:56 +0000
Subject: [PATCH 0279/2012] chore(sync): update OpenRouter model catalog
 (#5883)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/openai/gpt-4.1-nano.toml        | 5 +----
 2 files changed, 4 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index eed63a2025b..3e8301fafee 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07868
-output = 0.15736
-cache_read = 0.015736
+input = 0.07896
+output = 0.15792
+cache_read = 0.015792
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-4.1-nano.toml b/providers/openrouter/models/openai/gpt-4.1-nano.toml
index 8db2a7fea71..fb48681a171 100644
--- a/providers/openrouter/models/openai/gpt-4.1-nano.toml
+++ b/providers/openrouter/models/openai/gpt-4.1-nano.toml
@@ -3,10 +3,7 @@ base_model = "openai/gpt-4.1-nano"
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.03
-
-[limit]
-output = 942_818
+cache_read = 0.025
 
 [modalities]
 input = ["image", "text", "pdf"]

From c5f47314c6284b3bc540e9a0201a4837b8057c04 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 21:21:51 +0000
Subject: [PATCH 0280/2012] chore(sync): update Kilo model catalog (#5887)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index a41e9047daa..bd7a4ebaa04 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -17,7 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.16
-cache_read = 0.01
+cache_read = 0.013
 
 [limit]
 context = 1_048_576

From ca8c9740d8ad225f4e098755446b02e6ba2ed33d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 21:21:54 +0000
Subject: [PATCH 0281/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5885)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/minimax/minimax-h3-max.toml   | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/vercel/models/minimax/minimax-h3-max.toml

diff --git a/providers/vercel/models/minimax/minimax-h3-max.toml b/providers/vercel/models/minimax/minimax-h3-max.toml
new file mode 100644
index 00000000000..7523209ec56
--- /dev/null
+++ b/providers/vercel/models/minimax/minimax-h3-max.toml
@@ -0,0 +1,17 @@
+name = "MiniMax H3 Max"
+description = "Video model for prompt-guided generation, editing, and motion workflows"
+family = "minimax"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = true
+reasoning = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["video"]

From aa36bf036f0306bbaaed34c276ac13d493545afa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 21:22:13 +0000
Subject: [PATCH 0282/2012] chore(sync): update OpenRouter model catalog
 (#5886)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3e8301fafee..eed63a2025b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07896
-output = 0.15792
-cache_read = 0.015792
+input = 0.07868
+output = 0.15736
+cache_read = 0.015736
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 3ef2e65161d..36d6f5b0521 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.16
-cache_read = 0.01
+cache_read = 0.013
 
 [limit]
 context = 1_310_720

From b9bb273a2093e0f8fe954a61aca10c4bdfb5ce44 Mon Sep 17 00:00:00 2001
From: Joseph Perla <jperla@users.noreply.github.com>
Date: Sun, 30 Aug 2026 14:26:06 -0700
Subject: [PATCH 0283/2012] fix(trustedrouter): use fully-prefixed model IDs
 (trustedrouter/auto) (#5884)

Co-authored-by: Joseph Perla <claude@jperla.com>
---
 providers/trustedrouter/models/{ => trustedrouter}/auto.toml      | 0
 providers/trustedrouter/models/{ => trustedrouter}/cheap.toml     | 0
 providers/trustedrouter/models/{ => trustedrouter}/e2e.toml       | 0
 providers/trustedrouter/models/{ => trustedrouter}/fast.toml      | 0
 .../trustedrouter/models/{ => trustedrouter}/synth-code.toml      | 0
 providers/trustedrouter/models/{ => trustedrouter}/synth.toml     | 0
 providers/trustedrouter/models/{ => trustedrouter}/zdr.toml       | 0
 7 files changed, 0 insertions(+), 0 deletions(-)
 rename providers/trustedrouter/models/{ => trustedrouter}/auto.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/cheap.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/e2e.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/fast.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/synth-code.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/synth.toml (100%)
 rename providers/trustedrouter/models/{ => trustedrouter}/zdr.toml (100%)

diff --git a/providers/trustedrouter/models/auto.toml b/providers/trustedrouter/models/trustedrouter/auto.toml
similarity index 100%
rename from providers/trustedrouter/models/auto.toml
rename to providers/trustedrouter/models/trustedrouter/auto.toml
diff --git a/providers/trustedrouter/models/cheap.toml b/providers/trustedrouter/models/trustedrouter/cheap.toml
similarity index 100%
rename from providers/trustedrouter/models/cheap.toml
rename to providers/trustedrouter/models/trustedrouter/cheap.toml
diff --git a/providers/trustedrouter/models/e2e.toml b/providers/trustedrouter/models/trustedrouter/e2e.toml
similarity index 100%
rename from providers/trustedrouter/models/e2e.toml
rename to providers/trustedrouter/models/trustedrouter/e2e.toml
diff --git a/providers/trustedrouter/models/fast.toml b/providers/trustedrouter/models/trustedrouter/fast.toml
similarity index 100%
rename from providers/trustedrouter/models/fast.toml
rename to providers/trustedrouter/models/trustedrouter/fast.toml
diff --git a/providers/trustedrouter/models/synth-code.toml b/providers/trustedrouter/models/trustedrouter/synth-code.toml
similarity index 100%
rename from providers/trustedrouter/models/synth-code.toml
rename to providers/trustedrouter/models/trustedrouter/synth-code.toml
diff --git a/providers/trustedrouter/models/synth.toml b/providers/trustedrouter/models/trustedrouter/synth.toml
similarity index 100%
rename from providers/trustedrouter/models/synth.toml
rename to providers/trustedrouter/models/trustedrouter/synth.toml
diff --git a/providers/trustedrouter/models/zdr.toml b/providers/trustedrouter/models/trustedrouter/zdr.toml
similarity index 100%
rename from providers/trustedrouter/models/zdr.toml
rename to providers/trustedrouter/models/trustedrouter/zdr.toml

From c48229f94ae8d9e129bb376f8e284b38d580ccee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 30 Aug 2026 23:22:02 +0000
Subject: [PATCH 0284/2012] chore(sync): update OpenRouter model catalog
 (#5888)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index eed63a2025b..95593cbe6b9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07868
-output = 0.15736
-cache_read = 0.015736
+input = 0.0784
+output = 0.1568
+cache_read = 0.01568
 
 [limit]
 context = 1_048_576

From 85f9d90a76aea0fd5a6da81c4c0f9f25e3680d0d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 01:05:13 +0000
Subject: [PATCH 0285/2012] chore(sync): update NanoGPT model catalog (#5894)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From a2e60e3a4bc3ad1079a863039eaaaa0de17fa85d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 01:05:15 +0000
Subject: [PATCH 0286/2012] chore(sync): update OpenRouter model catalog
 (#5892)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 7 ++++---
 .../openrouter/models/meta-llama/llama-4-maverick.toml     | 4 ++--
 providers/openrouter/models/moonshotai/kimi-k2.5.toml      | 6 +++---
 providers/openrouter/models/tencent/hy3.toml               | 6 +++---
 5 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 95593cbe6b9..57d98266566 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0784
-output = 0.1568
-cache_read = 0.01568
+input = 0.089866
+output = 0.179732
+cache_read = 0.017973
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 05dce4765d6..95460b6f5f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.417252
-output = 0.834504
-cache_read = 0.034771
+input = 1.6
+output = 3.2
+cache_read = 0.135
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index de6e5240ff9..f4b36c39886 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -13,11 +13,11 @@ open_weights = true
 
 [cost]
 input = 0.2
-output = 0.8
+output = 0.696
 
 [limit]
 context = 1_048_576
-output = 16_384
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.5.toml b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
index 7943f5c3aaa..ebb15bee06b 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.5.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.5.toml
@@ -10,9 +10,9 @@ field = "reasoning_details"
 type = "toggle"
 
 [cost]
-input = 0.6
-output = 3
-cache_read = 0.1
+input = 0.45
+output = 2.25
+cache_read = 0.07
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 248e19b9e12735dd7fdf5d3ffb985c7ecf92cd5b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 01:05:50 +0000
Subject: [PATCH 0287/2012] chore(sync): update Kilo model catalog (#5890)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml    | 3 ++-
 providers/kilo/models/meta-llama/llama-4-maverick.toml | 4 ++--
 providers/kilo/models/tencent/hy3.toml                 | 6 +++---
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index 88500bd7dd8..e58e24281d4 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,4 +11,5 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 3db8291511b..5c1df37c375 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -15,8 +15,8 @@ input = 0.2
 output = 0.696
 
 [limit]
-context = 1_048_576
-output = 16_384
+context = 128_000
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144

From 466b5a57f000876fa45ea77e922ad05afafd55c8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 01:39:53 +0000
Subject: [PATCH 0288/2012] chore(sync): update OpenRouter model catalog
 (#5895)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576

From fd02f6fff829f38b4839917ac39d000d2a81125a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 01:40:09 +0000
Subject: [PATCH 0289/2012] chore(sync): update Kilo model catalog (#5896)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..e79a3173757 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 41b0e848c965e51bc05e265f62054b22bd673e70 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 03:32:17 +0000
Subject: [PATCH 0290/2012] chore(sync): update OpenRouter model catalog
 (#5899)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 57d98266566..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.089866
-output = 0.179732
-cache_read = 0.017973
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576

From 228579f0ead5fc3b68b8b1884496814528ea8528 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 03:32:22 +0000
Subject: [PATCH 0291/2012] chore(sync): update Kilo model catalog (#5898)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 4fbd2317736..2e32cc624ce 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,12 +15,11 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1875
-output = 4.18
-cache_read = 0.247
+input = 0
+output = 0
 
 [limit]
-context = 262_144
+context = 1_048_576
 output = 131_072
 
 [modalities]

From 592f8f85ab355b8c68dd939b2b8382d27575de36 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 03:32:37 +0000
Subject: [PATCH 0292/2012] chore(sync): update CrossModel model catalog
 (#5897)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/tencent/hy4-preview.toml    | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/crossmodel/models/tencent/hy4-preview.toml

diff --git a/providers/crossmodel/models/tencent/hy4-preview.toml b/providers/crossmodel/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..c7f6b89ed74
--- /dev/null
+++ b/providers/crossmodel/models/tencent/hy4-preview.toml
@@ -0,0 +1,15 @@
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.96
+output = 2.88
+cache_read = 0.048
+cache_write = 0.96
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 9c90ec7bc998632f6604629818ff9448bbd32bc2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 04:31:42 +0000
Subject: [PATCH 0293/2012] chore(sync): update OpenRouter model catalog
 (#5903)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From c3352299972c045c0c7bc88e05c642dc8915d1c9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 04:31:44 +0000
Subject: [PATCH 0294/2012] chore(sync): update Kilo model catalog (#5904)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml                | 7 ++++---
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index e79a3173757..6f876cf5199 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2e32cc624ce..4fbd2317736 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,11 +15,12 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0
-output = 0
+input = 1.1875
+output = 4.18
+cache_read = 0.247
 
 [limit]
-context = 1_048_576
+context = 262_144
 output = 131_072
 
 [modalities]

From c00c192726b4059d4fa8f7e5cf4e4140b5993e7a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 04:32:10 +0000
Subject: [PATCH 0295/2012] chore(sync): update NanoGPT model catalog (#5905)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From b6f92d1084becca2e66f2870c357244454810a44 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 06:49:32 +0000
Subject: [PATCH 0296/2012] chore(sync): update Kilo model catalog (#5909)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..e79a3173757 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 64268cc29207fd88c784eea3e64a3b04f7b6f9ee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 06:49:47 +0000
Subject: [PATCH 0297/2012] chore(sync): update NanoGPT model catalog (#5908)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 35a156d23e61300c077b8bd77f6188f54fbac741 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 06:50:27 +0000
Subject: [PATCH 0298/2012] chore(sync): update OpenRouter model catalog
 (#5907)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576

From 8834d2cf3e5636228e846aebf1d8f9e2d550f737 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 08:36:46 +0000
Subject: [PATCH 0299/2012] chore(sync): update EmpirioLabs AI model catalog
 (#5911)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/mistral-medium-3.toml | 10 ----------
 1 file changed, 10 deletions(-)
 delete mode 100644 providers/empiriolabs/models/mistral-medium-3.toml

diff --git a/providers/empiriolabs/models/mistral-medium-3.toml b/providers/empiriolabs/models/mistral-medium-3.toml
deleted file mode 100644
index f868fad65db..00000000000
--- a/providers/empiriolabs/models/mistral-medium-3.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "mistral/mistral-medium-2505"
-structured_output = true
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 130_000
-output = 40_000

From 5e4089efba59bc00b784bb2f5640005bcd98bc1c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 08:37:38 +0000
Subject: [PATCH 0300/2012] chore(sync): update NanoGPT model catalog (#5912)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/glm-5.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/nano-gpt/models/TEE/glm-5.3.toml

diff --git a/providers/nano-gpt/models/TEE/glm-5.3.toml b/providers/nano-gpt/models/TEE/glm-5.3.toml
new file mode 100644
index 00000000000..33b553ac7b5
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 TEE"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+input = 1_048_576

From 283c9c8618190a02f6ed1ff42da3b205d0998efd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 10:27:53 +0000
Subject: [PATCH 0301/2012] chore(sync): update CrossModel model catalog
 (#5917)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/crossmodel/models/tencent/hy3.toml         | 2 +-
 providers/crossmodel/models/tencent/hy4-preview.toml | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/crossmodel/models/tencent/hy3.toml b/providers/crossmodel/models/tencent/hy3.toml
index 0aca4d6ee70..dba2194c99c 100644
--- a/providers/crossmodel/models/tencent/hy3.toml
+++ b/providers/crossmodel/models/tencent/hy3.toml
@@ -13,4 +13,4 @@ cache_write = 0.16
 
 [limit]
 context = 262_144
-output = 262_144
+output = 131_072
diff --git a/providers/crossmodel/models/tencent/hy4-preview.toml b/providers/crossmodel/models/tencent/hy4-preview.toml
index c7f6b89ed74..3fbba5da035 100644
--- a/providers/crossmodel/models/tencent/hy4-preview.toml
+++ b/providers/crossmodel/models/tencent/hy4-preview.toml
@@ -12,4 +12,4 @@ cache_write = 0.96
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 65_536

From b063951aab01ade2e6c22b3d2b908dedd9961d35 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 10:28:31 +0000
Subject: [PATCH 0302/2012] chore(sync): update OpenRouter model catalog
 (#5915)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From 3bbe7bd5a8cbdb4314fb3d8de4fa5f72a4c832f4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 10:28:36 +0000
Subject: [PATCH 0303/2012] chore(sync): update NanoGPT model catalog (#5918)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 786336f085cfb427d27007a9aa4ffd508f0ded88 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 10:28:49 +0000
Subject: [PATCH 0304/2012] chore(sync): update Kilo model catalog (#5916)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index e79a3173757..6f876cf5199 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From fa73fb758638fe119686ade087d689d12dbab985 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 11:24:54 +0000
Subject: [PATCH 0305/2012] chore(sync): update OpenRouter model catalog
 (#5920)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 95460b6f5f3..66443ade4c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.6
-output = 3.2
-cache_read = 0.135
+input = 1.04226
+output = 2.08452
+cache_read = 0.086855
 
 [limit]
 context = 1_048_576
-output = 393_216

From f09af0fa76ac2f08f38aabd4b0f5e5f81b8ba64d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 11:24:59 +0000
Subject: [PATCH 0306/2012] chore(sync): update Kilo model catalog (#5919)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index e58e24281d4..88500bd7dd8 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,5 +11,4 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000

From 05d212c32c4cc9cd4463e65a4247111403f09e67 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 12:37:47 +0000
Subject: [PATCH 0307/2012] chore(sync): update Kilo model catalog (#5921)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml  | 6 +++---
 providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml       | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 6f876cf5199..4626ba1427a 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.028
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml b/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml
index 8c14b589e85..8e2499749ae 100644
--- a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml
+++ b/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml
@@ -16,8 +16,8 @@ output = 1.2
 cache_read = 0.06
 
 [limit]
-context = 256_000
-output = 80_000
+context = 262_144
+output = 144_000
 
 [modalities]
 input = ["text"]

From bea920f3186a95e6663dabd67f3916ced987c0b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 12:37:53 +0000
Subject: [PATCH 0308/2012] chore(sync): update OpenRouter model catalog
 (#5922)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml
index afdeb4228d2..80b0f00e52d 100644
--- a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml
+++ b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml
@@ -17,7 +17,7 @@ cache_read = 0.06
 
 [limit]
 context = 262_144
-output = 80_000
+output = 144_000
 
 [modalities]
 input = ["text"]

From c6980a5928714560e2b57d90e37e798ce163f534 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 13:29:22 +0000
Subject: [PATCH 0309/2012] chore(sync): update Charm Hyper model catalog
 (#5924)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml  |  6 +++---
 providers/hyper/models/glm-5.1.toml             |  6 +++---
 providers/hyper/models/glm-5.3-flash.toml       | 17 +++++++++++++++++
 providers/hyper/models/glm-5.3.toml             | 16 ++++++++++++++++
 providers/hyper/models/glm-5.toml               |  6 +++---
 providers/hyper/models/gpt-oss-120b.toml        |  6 +++---
 .../llama-4-maverick-17b-128e-instruct-fp8.toml |  6 +++---
 providers/hyper/models/minimax-m2.7.toml        |  6 +++---
 8 files changed, 51 insertions(+), 18 deletions(-)
 create mode 100644 providers/hyper/models/glm-5.3-flash.toml
 create mode 100644 providers/hyper/models/glm-5.3.toml

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 443e26fe9ac..85218385ac6 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.11
-output = 0.408
-cache_write = 0.055
+input = 0.12
+output = 0.42
+cache_write = 0.06
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 78d701cc4e0..a1b201fde20 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.314
-output = 4.268
-cache_write = 0.657
+input = 1.332
+output = 4.312
+cache_write = 0.666
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.3-flash.toml b/providers/hyper/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..8a41dc80ef6
--- /dev/null
+++ b/providers/hyper/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+last_updated = "2026-08-31"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.16332
+output = 0.5444
+cache_read = 0.031575
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/hyper/models/glm-5.3.toml b/providers/hyper/models/glm-5.3.toml
new file mode 100644
index 00000000000..bc974016adc
--- /dev/null
+++ b/providers/hyper/models/glm-5.3.toml
@@ -0,0 +1,16 @@
+base_model = "zhipuai/glm-5.3"
+release_date = "2026-08-28"
+last_updated = "2026-08-31"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.52432
+output = 4.79072
+cache_read = 0.283088
+
+[limit]
+context = 1_048_576
+output = 262_144
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 9e565da68e5..01e0d21e4fb 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.9
-output = 2.804
-cache_write = 0.45
+input = 0.91
+output = 2.934
+cache_write = 0.455
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index b8abd9a4a66..ad08deed2d7 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.178
-output = 0.68
-cache_write = 0.089
+input = 0.188
+output = 0.7
+cache_write = 0.094
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
index b36adf9ee14..9b588de13d6 100644
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.284
-output = 0.934
-cache_write = 0.142
+input = 0.274
+output = 0.8992
+cache_write = 0.137
 
 [limit]
 context = 430_000
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index de3e51335c2..de482c56069 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.404
-output = 1.496
-cache_write = 0.202
+input = 0.424
+output = 1.612
+cache_write = 0.212
 
 [limit]
 context = 262_100

From 819936141188e3a1ec15444d6e65cddc9bcc3ecc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 13:29:42 +0000
Subject: [PATCH 0310/2012] chore(sync): update OpenRouter model catalog
 (#5923)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 66443ade4c4..4bf670bbd00 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.04226
-output = 2.08452
-cache_read = 0.086855
+input = 1.039302
+output = 2.078604
+cache_read = 0.086609
 
 [limit]
 context = 1_048_576

From 27e103e33714bf47e3e16425026f9cb9bba3d8ff Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 14:30:00 +0000
Subject: [PATCH 0311/2012] chore(sync): update OpenRouter model catalog
 (#5926)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576

From 4a3a072b45d6d79611b6d1ccddf23f22a7b4cfc2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 14:30:04 +0000
Subject: [PATCH 0312/2012] chore(sync): update NanoGPT model catalog (#5925)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../abliterated-model-large-v2.toml           | 27 +++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml

diff --git a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
new file mode 100644
index 00000000000..0333560b0f1
--- /dev/null
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
@@ -0,0 +1,27 @@
+name = "Abliterated Model Large V2"
+description = "Abliteration.ai's default large text reasoning model is derived from GLM-5.3 for harder reasoning and evaluation workloads, with automatic prompt caching and a one-million-token context window."
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 5
+output = 5
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 999_990
+
+[modalities]
+input = ["text"]
+output = ["text"]

From d14ec223a1fd982e31c78899ce4c212143ed4d4a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 15:26:39 +0000
Subject: [PATCH 0313/2012] chore(sync): update Eden AI model catalog (#5930)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 5 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 8307fda46b6..48c5c557a03 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.756795
-output = 0.756795
+input = 0.75374
+output = 0.75374
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 384c2737624..8b46b24c3fd 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174645
-output = 0.756795
+input = 0.17394
+output = 0.75374
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 31f9aa58835..5359e6b91b0 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46572
-output = 0.93144
+input = 0.46384
+output = 0.92768
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 75b35e954bc..315161c8eea 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174645
-output = 0.69858
+input = 0.17394
+output = 0.69576
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 9304c96d6a1..9f6e0059657 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04787
-output = 1.04787
+input = 1.04364
+output = 1.04364

From ed143882e2ec2070b244da1beb32bc6ebe900697 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 15:26:52 +0000
Subject: [PATCH 0314/2012] chore(sync): update Kilo model catalog (#5931)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-chat-v3-0324.toml       |  5 ++--
 .../kilo/models/deepseek/deepseek-chat.toml   |  4 ++--
 .../models/kwaipilot/kat-coder-air-v2.5.toml  | 24 -------------------
 .../models/kwaipilot/kat-coder-pro-v2.5.toml  |  4 ++--
 providers/kilo/models/tencent/hy3:free.toml   | 18 --------------
 5 files changed, 6 insertions(+), 49 deletions(-)
 delete mode 100644 providers/kilo/models/kwaipilot/kat-coder-air-v2.5.toml
 delete mode 100644 providers/kilo/models/tencent/hy3:free.toml

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
index dd52e88cbb2..764c5575e7c 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
@@ -11,9 +11,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.27
-output = 1.12
-cache_read = 0.135
+input = 0.25
+output = 1
 
 [limit]
 context = 163_840
diff --git a/providers/kilo/models/deepseek/deepseek-chat.toml b/providers/kilo/models/deepseek/deepseek-chat.toml
index 9284b65fb41..f67bbe80e1f 100644
--- a/providers/kilo/models/deepseek/deepseek-chat.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat.toml
@@ -4,8 +4,8 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.4
-output = 1.3
+input = 0.2574
+output = 1.0287
 
 [limit]
 context = 128_000
diff --git a/providers/kilo/models/kwaipilot/kat-coder-air-v2.5.toml b/providers/kilo/models/kwaipilot/kat-coder-air-v2.5.toml
deleted file mode 100644
index d891c2b7b64..00000000000
--- a/providers/kilo/models/kwaipilot/kat-coder-air-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kwaipilot: KAT-Coder-Air V2.5"
-description = "KAT-Coder-Air V2.5 is a flagship-level Agentic Coding model that can directly hand over an entire issue or an entire business workflow to it, allowing it to autonomously locate and make..."
-family = "kat-coder"
-release_date = "2026-07-10"
-last_updated = "2026-07-10"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.03
-
-[limit]
-context = 256_000
-output = 80_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.5.toml b/providers/kilo/models/kwaipilot/kat-coder-pro-v2.5.toml
index a7851a2fa50..38b3e5d6d0e 100644
--- a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.5.toml
+++ b/providers/kilo/models/kwaipilot/kat-coder-pro-v2.5.toml
@@ -16,8 +16,8 @@ output = 2.96
 cache_read = 0.15
 
 [limit]
-context = 256_000
-output = 80_000
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/tencent/hy3:free.toml b/providers/kilo/models/tencent/hy3:free.toml
deleted file mode 100644
index 6106b0b26a5..00000000000
--- a/providers/kilo/models/tencent/hy3:free.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "tencent/hy3"
-name = "Tencent: Hy3 (free)"
-description = "Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent, activating 21B parameters per token. It supports configurable reasoning effort, agentic workflows, reliable tool calling, and long-context tasks across coding, document processing, financial analysis, and frontend development."
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high"]
-
-[cost]
-input = 0
-output = 0
-reasoning = 0
-cache_read = 0
-
-[limit]
-context = 262_144
-output = 128_000

From 04bf43c80fe86b810790acb1bc7a7d50be9b8075 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 15:27:16 +0000
Subject: [PATCH 0315/2012] chore(sync): update Merge Gateway model catalog
 (#5928)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/openai/gpt-5.6-luna.toml    | 1 +
 providers/merge-gateway/models/openai/gpt-5.6-sol.toml     | 1 +
 providers/merge-gateway/models/openai/gpt-5.6-terra.toml   | 1 +
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 5 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/providers/merge-gateway/models/openai/gpt-5.6-luna.toml b/providers/merge-gateway/models/openai/gpt-5.6-luna.toml
index 0cbbb124685..9c9f16a4eeb 100644
--- a/providers/merge-gateway/models/openai/gpt-5.6-luna.toml
+++ b/providers/merge-gateway/models/openai/gpt-5.6-luna.toml
@@ -12,3 +12,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 0.2
 output = 1.2
 cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/merge-gateway/models/openai/gpt-5.6-sol.toml b/providers/merge-gateway/models/openai/gpt-5.6-sol.toml
index 6acd06e8bb8..762cacdb749 100644
--- a/providers/merge-gateway/models/openai/gpt-5.6-sol.toml
+++ b/providers/merge-gateway/models/openai/gpt-5.6-sol.toml
@@ -11,3 +11,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 5
 output = 30
 cache_read = 0.5
+cache_write = 5
diff --git a/providers/merge-gateway/models/openai/gpt-5.6-terra.toml b/providers/merge-gateway/models/openai/gpt-5.6-terra.toml
index 1e8e3952eaf..a8492821213 100644
--- a/providers/merge-gateway/models/openai/gpt-5.6-terra.toml
+++ b/providers/merge-gateway/models/openai/gpt-5.6-terra.toml
@@ -12,3 +12,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 2
 output = 12
 cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..faa70b7c1e0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -11,8 +12,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index f30ae74cdb3..35e52350151 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 3ecc6ada2166467c5b13ce8322e693e342e7adbf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 15:27:41 +0000
Subject: [PATCH 0316/2012] chore(sync): update NanoGPT model catalog (#5932)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/gemma-4-12b-it.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/gemma-4-12b-it.toml b/providers/nano-gpt/models/gemma-4-12b-it.toml
index 1bde1b6c0cc..7c2890a9c37 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it.toml
@@ -10,9 +10,9 @@ structured_output = true
 open_weights = true
 
 [cost]
-input = 0.06
-output = 0.3
-cache_read = 0.03
+input = 0.05
+output = 0.25
+cache_read = 0.025
 
 [limit]
 context = 262_144

From 5abd0a2ed03a75232b3fb5d210b151ed1eb09ca6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 15:28:18 +0000
Subject: [PATCH 0317/2012] chore(sync): update OpenRouter model catalog
 (#5929)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++---
 .../models/kwaipilot/kat-coder-air-v2.5.toml  | 24 -------------------
 .../models/kwaipilot/kat-coder-pro-v2.5.toml  |  2 +-
 3 files changed, 4 insertions(+), 28 deletions(-)
 delete mode 100644 providers/openrouter/models/kwaipilot/kat-coder-air-v2.5.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4bf670bbd00..97904bfc0b8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.039302
-output = 2.078604
-cache_read = 0.086609
+input = 1.035996
+output = 2.071992
+cache_read = 0.086333
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/kwaipilot/kat-coder-air-v2.5.toml b/providers/openrouter/models/kwaipilot/kat-coder-air-v2.5.toml
deleted file mode 100644
index 2c513fd50ba..00000000000
--- a/providers/openrouter/models/kwaipilot/kat-coder-air-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "KAT-Coder-Air V2.5"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-07-10"
-last_updated = "2026-07-10"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.03
-
-[limit]
-context = 256_000
-output = 80_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
index 9f459a478b0..3a631336627 100644
--- a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
+++ b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.5.toml
@@ -17,7 +17,7 @@ cache_read = 0.15
 
 [limit]
 context = 262_144
-output = 80_000
+output = 235_929
 
 [modalities]
 input = ["text"]

From 583ffb32324b98e567aeedd212320d3cad50e2ad Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:27:29 +0000
Subject: [PATCH 0318/2012] chore(sync): update OpenRouter model catalog
 (#5934)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 378cd0676d8ca31a1dce34f0080e219441ffacf2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:27:39 +0000
Subject: [PATCH 0319/2012] chore(sync): update CrossModel model catalog
 (#5935)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/qwen/qwen3.7-max.toml      |  8 ++++----
 .../crossmodel/models/qwen/qwen3.7-plus.toml     | 16 ++++++++--------
 2 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/crossmodel/models/qwen/qwen3.7-max.toml b/providers/crossmodel/models/qwen/qwen3.7-max.toml
index 1f610e24e7e..66e5926e909 100644
--- a/providers/crossmodel/models/qwen/qwen3.7-max.toml
+++ b/providers/crossmodel/models/qwen/qwen3.7-max.toml
@@ -5,7 +5,7 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 1.504
-output = 4.504
-cache_read = 0.3
-cache_write = 1.88
+input = 1.88
+output = 5.63
+cache_read = 0.375
+cache_write = 2.35
diff --git a/providers/crossmodel/models/qwen/qwen3.7-plus.toml b/providers/crossmodel/models/qwen/qwen3.7-plus.toml
index 749ed99d2c4..2c9e669358a 100644
--- a/providers/crossmodel/models/qwen/qwen3.7-plus.toml
+++ b/providers/crossmodel/models/qwen/qwen3.7-plus.toml
@@ -5,14 +5,14 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.288
-output = 1.125
-cache_read = 0.0288
-cache_write = 0.36
+input = 0.32
+output = 1.25
+cache_read = 0.032
+cache_write = 0.4
 
 [[cost.tiers]]
 tier = { type = "context", size = 256_000 }
-input = 0.864
-output = 3.375
-cache_read = 0.0864
-cache_write = 1.08
+input = 0.96
+output = 3.75
+cache_read = 0.096
+cache_write = 1.2

From bf38988bb2a8ff71b26b6b8c83fb78dbd83a1da4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:28:05 +0000
Subject: [PATCH 0320/2012] chore(sync): update Merge Gateway model catalog
 (#5933)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-sonnet-5.toml           | 13 +++++++++----
 providers/merge-gateway/models/xai/grok-4.6.toml    |  6 +++---
 2 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/providers/merge-gateway/models/anthropic/claude-sonnet-5.toml b/providers/merge-gateway/models/anthropic/claude-sonnet-5.toml
index 4ba9d259a9d..bb662561fba 100644
--- a/providers/merge-gateway/models/anthropic/claude-sonnet-5.toml
+++ b/providers/merge-gateway/models/anthropic/claude-sonnet-5.toml
@@ -1,13 +1,18 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=anthropic%2Fclaude-sonnet-5 (accessed 2026-07-21)
 # Merge Gateway GET /v1/models?model=anthropic/claude-sonnet-5 (accessed 2026-07-14):
 # the Anthropic route advertises configurable `reasoning.effort` and disable support.
-
 base_model = "anthropic/claude-sonnet-5"
 structured_output = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2
-output = 10
+input = 3
+output = 15
 cache_read = 0.2
 cache_write = 2.5
diff --git a/providers/merge-gateway/models/xai/grok-4.6.toml b/providers/merge-gateway/models/xai/grok-4.6.toml
index 7bd903104ea..1dee37c146c 100644
--- a/providers/merge-gateway/models/xai/grok-4.6.toml
+++ b/providers/merge-gateway/models/xai/grok-4.6.toml
@@ -5,6 +5,6 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.5
-output = 4.5
-cache_read = 0.375
+input = 2
+output = 6
+cache_read = 0.5

From 3260c2b27046d2997eb54491f6b2fa75b77582dd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:28:36 +0000
Subject: [PATCH 0321/2012] chore(sync): update Kilo model catalog (#5936)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 473af465dce64ff796cf276bde238f2e6544a742 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 17:23:40 +0000
Subject: [PATCH 0322/2012] chore(sync): update OpenRouter model catalog
 (#5938)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml     | 3 +--
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 97904bfc0b8..b239680e290 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.035996
-output = 2.071992
-cache_read = 0.086333
+input = 1.034256
+output = 2.068512
+cache_read = 0.086188
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index abacdd68c88..51225871db8 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,8 +14,7 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.25
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
-output = 262_144

From ad3ee31a26aec7159ba3abae2cff802fc3d769f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 17:24:05 +0000
Subject: [PATCH 0323/2012] chore(sync): update Kilo model catalog (#5939)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index c79bdfce589..ad993c8f794 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,7 +13,3 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
-
-[limit]
-context = 1_000_000
-output = 262_144

From 9322c13a0985d337780094272d97ac2fa26bd82a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 18:31:25 +0000
Subject: [PATCH 0324/2012] chore(sync): update OpenRouter model catalog
 (#5940)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/google/gemini-3.5-flash.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/google/gemini-3.5-flash.toml b/providers/openrouter/models/google/gemini-3.5-flash.toml
index 007d477d87b..e81d5758bd5 100644
--- a/providers/openrouter/models/google/gemini-3.5-flash.toml
+++ b/providers/openrouter/models/google/gemini-3.5-flash.toml
@@ -5,8 +5,8 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 9
-reasoning = 9
-cache_read = 0.15
+input = 1.65
+output = 9.9
+reasoning = 9.9
+cache_read = 0.165
 cache_write = 0.083333

From 92c4c46cd0c797508fff7f17e77434ec8ea6b99b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 18:31:29 +0000
Subject: [PATCH 0325/2012] chore(sync): update Kilo model catalog (#5941)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/google/gemini-2.5-flash-image.toml     |  8 ++++----
 .../kilo/models/google/gemini-3-flash-preview.toml     | 10 +++++-----
 .../kilo/models/google/gemini-3-pro-image-preview.toml | 10 +++++-----
 .../models/google/gemini-3.1-flash-lite-preview.toml   | 10 +++++-----
 .../kilo/models/google/gemini-3.1-flash-lite.toml      | 10 +++++-----
 .../kilo/models/google/gemini-3.1-pro-preview.toml     | 10 +++++-----
 .../kilo/models/google/gemini-3.5-flash-lite.toml      | 10 +++++-----
 providers/kilo/models/google/gemini-3.5-flash.toml     | 10 +++++-----
 providers/kilo/models/google/gemini-3.6-flash.toml     | 10 +++++-----
 providers/kilo/models/google/gemini-3.7-flash.toml     | 10 +++++-----
 10 files changed, 49 insertions(+), 49 deletions(-)

diff --git a/providers/kilo/models/google/gemini-2.5-flash-image.toml b/providers/kilo/models/google/gemini-2.5-flash-image.toml
index cec711d9189..1ac49c4d12d 100644
--- a/providers/kilo/models/google/gemini-2.5-flash-image.toml
+++ b/providers/kilo/models/google/gemini-2.5-flash-image.toml
@@ -4,10 +4,10 @@ reasoning = false
 structured_output = true
 
 [cost]
-input = 0.3
-output = 2.5
-cache_read = 0.03
-cache_write = 0.083333
+input = 0.15
+output = 1.25
+cache_read = 0.015
+cache_write = 0.041667
 
 [limit]
 output = 8_192
diff --git a/providers/kilo/models/google/gemini-3-flash-preview.toml b/providers/kilo/models/google/gemini-3-flash-preview.toml
index c36634cee03..53969153e81 100644
--- a/providers/kilo/models/google/gemini-3-flash-preview.toml
+++ b/providers/kilo/models/google/gemini-3-flash-preview.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.5
-output = 3
-reasoning = 3
-cache_read = 0.05
-cache_write = 0.083333
+input = 0.25
+output = 1.5
+reasoning = 1.5
+cache_read = 0.025
+cache_write = 0.041667
diff --git a/providers/kilo/models/google/gemini-3-pro-image-preview.toml b/providers/kilo/models/google/gemini-3-pro-image-preview.toml
index fbc8273e7e5..187901bba6f 100644
--- a/providers/kilo/models/google/gemini-3-pro-image-preview.toml
+++ b/providers/kilo/models/google/gemini-3-pro-image-preview.toml
@@ -7,11 +7,11 @@ type = "effort"
 values = ["high"]
 
 [cost]
-input = 2
-output = 12
-reasoning = 12
-cache_read = 0.2
-cache_write = 0.375
+input = 1
+output = 6
+reasoning = 6
+cache_read = 0.1
+cache_write = 0.1875
 
 [modalities]
 input = ["image", "text", "pdf"]
diff --git a/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml b/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
index 3ed3174dbdc..3cc639df076 100644
--- a/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
+++ b/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 1.5
-reasoning = 1.5
-cache_read = 0.025
-cache_write = 0.083333
+input = 0.125
+output = 0.75
+reasoning = 0.75
+cache_read = 0.0125
+cache_write = 0.041667
diff --git a/providers/kilo/models/google/gemini-3.1-flash-lite.toml b/providers/kilo/models/google/gemini-3.1-flash-lite.toml
index d01a3b46f5d..5de09a65c91 100644
--- a/providers/kilo/models/google/gemini-3.1-flash-lite.toml
+++ b/providers/kilo/models/google/gemini-3.1-flash-lite.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 1.5
-reasoning = 1.5
-cache_read = 0.025
-cache_write = 0.083333
+input = 0.125
+output = 0.75
+reasoning = 0.75
+cache_read = 0.0125
+cache_write = 0.041667
diff --git a/providers/kilo/models/google/gemini-3.1-pro-preview.toml b/providers/kilo/models/google/gemini-3.1-pro-preview.toml
index b6940362b26..9504d134abe 100644
--- a/providers/kilo/models/google/gemini-3.1-pro-preview.toml
+++ b/providers/kilo/models/google/gemini-3.1-pro-preview.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 2
-output = 12
-reasoning = 12
-cache_read = 0.2
-cache_write = 0.375
+input = 1
+output = 6
+reasoning = 6
+cache_read = 0.1
+cache_write = 0.1875
diff --git a/providers/kilo/models/google/gemini-3.5-flash-lite.toml b/providers/kilo/models/google/gemini-3.5-flash-lite.toml
index 9d7dc172272..7df3fa1b11d 100644
--- a/providers/kilo/models/google/gemini-3.5-flash-lite.toml
+++ b/providers/kilo/models/google/gemini-3.5-flash-lite.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.3
-output = 2.5
-reasoning = 2.5
-cache_read = 0.03
-cache_write = 0.083333
+input = 0.15
+output = 1.25
+reasoning = 1.25
+cache_read = 0.015
+cache_write = 0.041667
diff --git a/providers/kilo/models/google/gemini-3.5-flash.toml b/providers/kilo/models/google/gemini-3.5-flash.toml
index 636e645a59e..924de6b8409 100644
--- a/providers/kilo/models/google/gemini-3.5-flash.toml
+++ b/providers/kilo/models/google/gemini-3.5-flash.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 9
-reasoning = 9
-cache_read = 0.15
-cache_write = 0.083333
+input = 0.75
+output = 4.5
+reasoning = 4.5
+cache_read = 0.075
+cache_write = 0.041667
diff --git a/providers/kilo/models/google/gemini-3.6-flash.toml b/providers/kilo/models/google/gemini-3.6-flash.toml
index 0cfea2d6c4d..a0edcbc465b 100644
--- a/providers/kilo/models/google/gemini-3.6-flash.toml
+++ b/providers/kilo/models/google/gemini-3.6-flash.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.75
-output = 3.75
-reasoning = 3.75
-cache_read = 0.075
-cache_write = 0.041667
+input = 0.375
+output = 1.875
+reasoning = 1.875
+cache_read = 0.0375
+cache_write = 0.020833
diff --git a/providers/kilo/models/google/gemini-3.7-flash.toml b/providers/kilo/models/google/gemini-3.7-flash.toml
index 912cb550f9a..c15dde8744f 100644
--- a/providers/kilo/models/google/gemini-3.7-flash.toml
+++ b/providers/kilo/models/google/gemini-3.7-flash.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667

From 88f66ceacc25ae0639db380de81e9e44060dade7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 18:33:17 +0000
Subject: [PATCH 0326/2012] chore(sync): update Deep Infra model catalog
 (#5942)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/zai-org/GLM-5.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/deepinfra/models/zai-org/GLM-5.3.toml b/providers/deepinfra/models/zai-org/GLM-5.3.toml
index 16174adddac..b9e733f0b6f 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.3.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.3.toml
@@ -12,7 +12,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 1.2
 output = 4
-cache_read = 0.24
+cache_read = 0.12
 
 [limit]
 context = 1_048_576

From a55cea59e863410fd1efa7008266c71c6ee2513d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 19:21:26 +0000
Subject: [PATCH 0327/2012] chore(sync): update OpenRouter model catalog
 (#5943)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b239680e290..07235ccdb08 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.034256
-output = 2.068512
-cache_read = 0.086188
+input = 1.030776
+output = 2.061552
+cache_read = 0.085898
 
 [limit]
 context = 1_048_576

From c276895618c5886fdb150799525d827117502e29 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 20:26:37 +0000
Subject: [PATCH 0328/2012] chore(sync): update NanoGPT model catalog (#5945)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/z-ai/glm-latest.toml b/providers/nano-gpt/models/z-ai/glm-latest.toml
index d1e853a3a90..7151629fb90 100644
--- a/providers/nano-gpt/models/z-ai/glm-latest.toml
+++ b/providers/nano-gpt/models/z-ai/glm-latest.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.42
-output = 1.32
-cache_read = 0.078
+input = 1
+output = 3.2
+cache_read = 0.2
 
 [limit]
 context = 1_048_576

From 5de341ca599941191338da3830a2ee07ec79bccd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 20:26:54 +0000
Subject: [PATCH 0329/2012] chore(sync): update Kilo model catalog (#5944)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/arcee-ai/trinity-large-thinking.toml  | 6 +++---
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml           | 4 ++++
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 2 +-
 3 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/arcee-ai/trinity-large-thinking.toml b/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
index 9410f5822bf..aab2adaa0f5 100644
--- a/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
+++ b/providers/kilo/models/arcee-ai/trinity-large-thinking.toml
@@ -1,13 +1,13 @@
 base_model = "arcee-ai/trinity-large-thinking"
-structured_output = true
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["high"]
 
 [cost]
-input = 0.22
-output = 0.85
+input = 0.25
+output = 0.8
 cache_read = 0.06
 
 [limit]
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index ad993c8f794..c79bdfce589 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,3 +13,7 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
+
+[limit]
+context = 1_000_000
+output = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index bd7a4ebaa04..80dc12b4ce8 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,7 +15,7 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.03
+input = 0.05
 output = 0.16
 cache_read = 0.013
 

From 35a746466d06c59a88c1f458f8481ab7c4c1312d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 20:26:57 +0000
Subject: [PATCH 0330/2012] chore(sync): update Merge Gateway model catalog
 (#5948)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731-fast.toml     | 9 +++++++++
 providers/merge-gateway/models/moonshot/kimi-k3.toml     | 4 ++--
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml     | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml | 7 +++----
 4 files changed, 17 insertions(+), 10 deletions(-)
 create mode 100644 providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731-fast.toml

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731-fast.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731-fast.toml
new file mode 100644
index 00000000000..4fe94f12c6b
--- /dev/null
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731-fast.toml
@@ -0,0 +1,9 @@
+base_model = "deepseek/deepseek-v4-flash-0731"
+tool_call = false
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 0.28
+output = 0.56
+cache_read = 0.07
diff --git a/providers/merge-gateway/models/moonshot/kimi-k3.toml b/providers/merge-gateway/models/moonshot/kimi-k3.toml
index 1e611952389..7e536155c8c 100644
--- a/providers/merge-gateway/models/moonshot/kimi-k3.toml
+++ b/providers/merge-gateway/models/moonshot/kimi-k3.toml
@@ -8,8 +8,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 3
-output = 15
+input = 2.9
+output = 14
 cache_read = 0.3
 
 [limit]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index faa70b7c1e0..3ef83d21d9d 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 0a7a2254ee8..b4bdd294444 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From bd46820ec59d21b6eac6ceb316b91ca3d3410adc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 20:27:01 +0000
Subject: [PATCH 0331/2012] chore(sync): update Venice model catalog (#5946)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/deepseek-v4-pro-0813.toml | 1 -
 providers/venice/models/kimi-k2-5.toml            | 1 -
 providers/venice/models/qwen3-6-27b.toml          | 2 --
 providers/venice/models/z-ai-glm-5-3.toml         | 1 -
 4 files changed, 5 deletions(-)

diff --git a/providers/venice/models/deepseek-v4-pro-0813.toml b/providers/venice/models/deepseek-v4-pro-0813.toml
index 76f63a75fb9..84bd6185cc1 100644
--- a/providers/venice/models/deepseek-v4-pro-0813.toml
+++ b/providers/venice/models/deepseek-v4-pro-0813.toml
@@ -2,7 +2,6 @@ base_model = "deepseek/deepseek-v4-pro-0813"
 description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
 release_date = "2026-08-14"
 last_updated = "2026-08-14"
-open_weights = false
 reasoning_options = []
 
 [cost]
diff --git a/providers/venice/models/kimi-k2-5.toml b/providers/venice/models/kimi-k2-5.toml
index 3e15964592c..cd7cb32ce01 100644
--- a/providers/venice/models/kimi-k2-5.toml
+++ b/providers/venice/models/kimi-k2-5.toml
@@ -4,7 +4,6 @@ base_model = "moonshotai/kimi-k2.5"
 release_date = "2026-01-27"
 last_updated = "2026-06-11"
 knowledge = "2024-04"
-open_weights = false
 reasoning_options = []
 
 [interleaved]
diff --git a/providers/venice/models/qwen3-6-27b.toml b/providers/venice/models/qwen3-6-27b.toml
index b57c8e4b878..27cdef2bf5c 100644
--- a/providers/venice/models/qwen3-6-27b.toml
+++ b/providers/venice/models/qwen3-6-27b.toml
@@ -1,11 +1,9 @@
 # Live /models (2026-06-25): none|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "alibaba/qwen3.6-27b"
 name = "Qwen 3.6 27B"
 release_date = "2026-04-24"
 last_updated = "2026-06-11"
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/venice/models/z-ai-glm-5-3.toml b/providers/venice/models/z-ai-glm-5-3.toml
index 2f84aeb13ab..4d41a6a336e 100644
--- a/providers/venice/models/z-ai-glm-5-3.toml
+++ b/providers/venice/models/z-ai-glm-5-3.toml
@@ -3,7 +3,6 @@ name = "GLM 5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 release_date = "2026-08-18"
 last_updated = "2026-08-18"
-open_weights = false
 reasoning_options = []
 
 [cost]

From 413f78011ef0af50035a65edb497e757433d64fa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 20:28:05 +0000
Subject: [PATCH 0332/2012] chore(sync): update OpenRouter model catalog
 (#5947)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../arcee-ai/trinity-large-thinking.toml      |  2 +-
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/google/gemini-3.5-flash.toml       |  8 +++---
 .../models/ibm-granite/granite-4.2-8b.toml    | 28 +++++++++++++++++++
 .../models/qwen/qwen3.8-2.4t-a95b.toml        |  3 +-
 .../~deepseek/deepseek-v4-flash-latest.toml   |  2 +-
 6 files changed, 39 insertions(+), 10 deletions(-)
 create mode 100644 providers/openrouter/models/ibm-granite/granite-4.2-8b.toml

diff --git a/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml b/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
index e6f6fe97d08..303c0f58cee 100644
--- a/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
+++ b/providers/openrouter/models/arcee-ai/trinity-large-thinking.toml
@@ -1,5 +1,5 @@
 base_model = "arcee-ai/trinity-large-thinking"
-structured_output = true
+structured_output = false
 reasoning_options = []
 
 [cost]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemini-3.5-flash.toml b/providers/openrouter/models/google/gemini-3.5-flash.toml
index e81d5758bd5..007d477d87b 100644
--- a/providers/openrouter/models/google/gemini-3.5-flash.toml
+++ b/providers/openrouter/models/google/gemini-3.5-flash.toml
@@ -5,8 +5,8 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 1.65
-output = 9.9
-reasoning = 9.9
-cache_read = 0.165
+input = 1.5
+output = 9
+reasoning = 9
+cache_read = 0.15
 cache_write = 0.083333
diff --git a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
new file mode 100644
index 00000000000..873558d6ea7
--- /dev/null
+++ b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
@@ -0,0 +1,28 @@
+name = "Granite 4.2 8B"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+family = "granite"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = false
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.1
+output = 0.15
+cache_read = 0.05
+
+[limit]
+context = 131_072
+output = 117_964
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index 51225871db8..abacdd68c88 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,7 +14,8 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.2
+cache_read = 0.25
 
 [limit]
 context = 1_048_576
+output = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 36d6f5b0521..27d5c88cb9e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,7 +20,7 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.03
+input = 0.05
 output = 0.16
 cache_read = 0.013
 

From de7f1b57dab95a7e45314680994b950dbc101fd3 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:04:12 -0500
Subject: [PATCH 0333/2012] fix: [missing-model] Groq: qwen/qwen3.8-27b (#5725)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/groq/models/qwen/qwen3.8-27b.toml | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/groq/models/qwen/qwen3.8-27b.toml

diff --git a/providers/groq/models/qwen/qwen3.8-27b.toml b/providers/groq/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..623715d4605
--- /dev/null
+++ b/providers/groq/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,21 @@
+# Sources (accessed 2026-08-28):
+# https://console.groq.com/docs/model/qwen/qwen3.8-27b
+# https://console.groq.com/docs/models
+# https://console.groq.com/docs/reasoning
+# Effort: reasoning_effort = none|default|low|medium|high
+# Pricing: $0.80/M input, $4.00/M output. Context 131_042, max output 16_384.
+# Modalities on Groq: text + images in, text out (no video). Preview model.
+base_model = "alibaba/qwen3.8-27b"
+reasoning_options = [{ type = "effort", values = ["none", "default", "low", "medium", "high"] }]
+
+[cost]
+input = 0.80
+output = 4.00
+
+[limit]
+context = 131_042
+output = 16_384
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From cca4a3015d97a40b62474a3b98bc18740156938d Mon Sep 17 00:00:00 2001
From: Abliteration AI <help@abliteration.ai>
Date: Mon, 31 Aug 2026 14:04:36 -0700
Subject: [PATCH 0334/2012] Add abliterated-model-large-v2 (#5927)

* feat(abliteration-ai): add abliterated-model-large-v2

GLM-5.3-based successor to abliterated-model-large. Reasoning runs in
three distinct modes (low, high, max) and cannot be disabled, so the
entry declares effort only - no toggle.

* Cite docs pricing page for abliterated-model-large-v2 rates
---
 .../models/abliterated-model-large-v2.toml    | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)
 create mode 100644 providers/abliteration-ai/models/abliterated-model-large-v2.toml

diff --git a/providers/abliteration-ai/models/abliterated-model-large-v2.toml b/providers/abliteration-ai/models/abliterated-model-large-v2.toml
new file mode 100644
index 00000000000..9d7d47eb436
--- /dev/null
+++ b/providers/abliteration-ai/models/abliterated-model-large-v2.toml
@@ -0,0 +1,42 @@
+# Sources (accessed 2026-08-31):
+# - Model card, limits (1M context, 999,990 max output), capabilities, GLM-5.3 base:
+#   https://docs.abliteration.ai/models
+# - Reasoning modes and per-endpoint request fields:
+#   https://docs.abliteration.ai/capabilities/thinking
+# - Pricing ($5 per 1M tokens input/output, $0.50 per 1M cached input):
+#   https://docs.abliteration.ai/pricing
+#   https://abliteration.ai/pricing
+# - Launch announcement (GLM-5.3 base, FP8 hosting, benchmarks):
+#   https://abliteration.ai/blog/introducing-abliterated-model-large-v2
+name = "Abliterated Model Large V2"
+description = "GLM-5.3 model abliterated and finetuned for cyber, ML red teaming, and agent testing"
+release_date = "2026-08-29"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+temperature = true
+open_weights = false
+
+# Three distinct reasoning depths: low, high, max (default: max).
+# Reasoning cannot be disabled on this model: "none" and {"thinking": false}
+# map to low and hide the trace, so there is no toggle option. Other ladder
+# values are accepted aliases: minimal -> low, medium -> high, xhigh -> max.
+[[reasoning_options]]
+type = "effort" # API: {"reasoning_effort": "<value>"} on /v1/chat/completions
+values = ["low", "high", "max"]
+
+[cost]
+input = 5.00
+output = 5.00
+cache_read = 0.50
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 999_990
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 45b840218cf89ecbb58565ceab9f112638103901 Mon Sep 17 00:00:00 2001
From: Divy <divy@coralbricks.ai>
Date: Mon, 31 Aug 2026 14:05:05 -0700
Subject: [PATCH 0335/2012] feat(coralbricks): add GLM 5.3, remove deprecated
 GLM 5.2 (#5902)

* feat(coralbricks): add GLM 5.3 (glm-5.3-fp4)

New model live on the CoralBricks gateway: 1M context, $1.12/M in,
$4.40/M out, cached input free (cache_read=0), reasoning toggle +
effort. Verified in the live /v1/models and a chat smoke.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(coralbricks): align GLM 5.3 reasoning_options to the lab baseline

Bot review: GLM-5.3 always reasons with effort low|high|max (per the
first-party zhipuai/zai entries and our own models.json routing). Drop
the copied-from-5.2 toggle + minimal/medium levels, and drop the
redundant reasoning=true (inherited from base_model zhipuai/glm-5.3).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* chore(coralbricks): remove GLM 5.2 (deprecated on prod)

glm-5.2-fp4 is deprecated in favor of GLM 5.3 (dropped from the prod tier
on the CoralBricks gateway; returns 404 model_not_public). Remove its
model entry. Provider now lists glm-5.3-fp4, kimi-k3, gpt-oss-120b.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
---
 providers/coralbricks/models/glm-5.2-fp4.toml | 15 ---------------
 providers/coralbricks/models/glm-5.3-fp4.toml | 14 ++++++++++++++
 2 files changed, 14 insertions(+), 15 deletions(-)
 delete mode 100644 providers/coralbricks/models/glm-5.2-fp4.toml
 create mode 100644 providers/coralbricks/models/glm-5.3-fp4.toml

diff --git a/providers/coralbricks/models/glm-5.2-fp4.toml b/providers/coralbricks/models/glm-5.2-fp4.toml
deleted file mode 100644
index 575d37719c6..00000000000
--- a/providers/coralbricks/models/glm-5.2-fp4.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2 FP4"
-
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["minimal", "low", "medium", "high"] }]
-interleaved = true
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[cost]
-input = 1.12
-output = 4.4
-cache_read = 0
diff --git a/providers/coralbricks/models/glm-5.3-fp4.toml b/providers/coralbricks/models/glm-5.3-fp4.toml
new file mode 100644
index 00000000000..2488afc97cb
--- /dev/null
+++ b/providers/coralbricks/models/glm-5.3-fp4.toml
@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 FP4"
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+interleaved = true
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[cost]
+input = 1.12
+output = 4.4
+cache_read = 0

From 1ad2635b6267a5478078fbc4172d047167cedec8 Mon Sep 17 00:00:00 2001
From: d2bz <139675120+d2bz@users.noreply.github.com>
Date: Tue, 1 Sep 2026 05:05:16 +0800
Subject: [PATCH 0336/2012] chore(aihubmix): sync model catalog (#5901)

* chore(aihubmix): sync model catalog

* fix(aihubmix): address catalog review

* fix(aihubmix): expose deepseek effort levels

* chore(ci): retry review workflow
---
 .../models/deepseek-v4-flash-0731.toml        | 13 +++++++++++
 .../aihubmix/models/deepseek-v4-pro-0813.toml | 13 +++++++++++
 providers/aihubmix/models/glm-5.3-flash.toml  | 17 ++++++++++++++
 providers/aihubmix/models/glm-5.3.toml        | 13 +++++++++++
 providers/aihubmix/models/grok-4.6.toml       |  7 ++++++
 providers/aihubmix/models/qwen3.7-flash.toml  | 22 +++++++++++++++++++
 .../aihubmix/models/qwen3.8-2.4t-a95b.toml    | 21 ++++++++++++++++++
 7 files changed, 106 insertions(+)
 create mode 100644 providers/aihubmix/models/deepseek-v4-flash-0731.toml
 create mode 100644 providers/aihubmix/models/deepseek-v4-pro-0813.toml
 create mode 100644 providers/aihubmix/models/glm-5.3-flash.toml
 create mode 100644 providers/aihubmix/models/glm-5.3.toml
 create mode 100644 providers/aihubmix/models/grok-4.6.toml
 create mode 100644 providers/aihubmix/models/qwen3.7-flash.toml
 create mode 100644 providers/aihubmix/models/qwen3.8-2.4t-a95b.toml

diff --git a/providers/aihubmix/models/deepseek-v4-flash-0731.toml b/providers/aihubmix/models/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..2659d18ff9a
--- /dev/null
+++ b/providers/aihubmix/models/deepseek-v4-flash-0731.toml
@@ -0,0 +1,13 @@
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort = low|high|max
+# AIHubMix effort levels returned HTTP 200 (validated 2026-08-31T04:17:24Z).
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.142
+output = 0.284
+cache_read = 0.0284
diff --git a/providers/aihubmix/models/deepseek-v4-pro-0813.toml b/providers/aihubmix/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..fbe29803f4c
--- /dev/null
+++ b/providers/aihubmix/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,13 @@
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort = high|max
+# AIHubMix effort levels returned HTTP 200 (validated 2026-08-31T04:17:24Z).
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.6918
+output = 2.0754
+cache_read = 0.023058
diff --git a/providers/aihubmix/models/glm-5.3-flash.toml b/providers/aihubmix/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..58250ec9ab4
--- /dev/null
+++ b/providers/aihubmix/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.11268
+output = 0.39438
+cache_read = 0.02817
+
+[limit]
+output = 128_000
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/aihubmix/models/glm-5.3.toml b/providers/aihubmix/models/glm-5.3.toml
new file mode 100644
index 00000000000..f9de29a855e
--- /dev/null
+++ b/providers/aihubmix/models/glm-5.3.toml
@@ -0,0 +1,13 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.1268
+output = 3.9438
+cache_read = 0.2817
+
+[limit]
+output = 128_000
diff --git a/providers/aihubmix/models/grok-4.6.toml b/providers/aihubmix/models/grok-4.6.toml
new file mode 100644
index 00000000000..5264268a96f
--- /dev/null
+++ b/providers/aihubmix/models/grok-4.6.toml
@@ -0,0 +1,7 @@
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
diff --git a/providers/aihubmix/models/qwen3.7-flash.toml b/providers/aihubmix/models/qwen3.7-flash.toml
new file mode 100644
index 00000000000..3f34f7fd74f
--- /dev/null
+++ b/providers/aihubmix/models/qwen3.7-flash.toml
@@ -0,0 +1,22 @@
+# Toggle: enable_thinking = true|false
+# Budget: thinking_budget = integer reasoning tokens
+base_model = "alibaba/qwen3.7-flash"
+attachment = false
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", max = 262_144 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0282
+output = 0.1128
+cache_read = 0.00564
+cache_write = 0.03525
+
+[limit]
+context = 991_000
+output = 64_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/aihubmix/models/qwen3.8-2.4t-a95b.toml b/providers/aihubmix/models/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..717b12b3b19
--- /dev/null
+++ b/providers/aihubmix/models/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,21 @@
+# AIHubMix Models API reports text,image input (queried 2026-08-31T04:06:29Z).
+# OpenAI-compatible HTTPS image input returned HTTP 200 (validated 2026-08-31T04:08:19Z).
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+attachment = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "xhigh"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[limit]
+context = 262_000
+output = 262_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 5f9d2d43ecb64544749e21692eb9bab9606bf5a3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:05:46 -0500
Subject: [PATCH 0337/2012] chore(sync): update LLM Gateway model catalog
 (#5893)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nebius/cosmos3-super-reasoner.toml | 29 ------------------
 .../models/nebius/deepseek-v4-pro.toml        | 17 -----------
 .../models/nebius/gemma-3-27b.toml            | 23 --------------
 .../models/nebius/glm-5.1.toml                | 17 -----------
 .../models/nebius/glm-5.2.toml                | 17 -----------
 .../models/nebius/gpt-oss-120b.toml           | 11 -------
 .../models/nebius/hermes-4-405b.toml          | 30 -------------------
 .../models/nebius/hermes-4-70b.toml           | 30 -------------------
 .../models/nebius/kimi-k2.6.toml              | 14 ---------
 .../models/nebius/kimi-k2.7-code.toml         | 18 -----------
 .../models/nebius/kimi-k3.toml                | 14 ---------
 .../nebius/llama-3.1-nemotron-ultra-253b.toml | 23 --------------
 .../models/nebius/llama-3.3-70b-instruct.toml |  8 -----
 .../models/nebius/minicpm-v-4.5.toml          | 22 --------------
 .../models/nebius/minimax-m2.5.toml           | 15 ----------
 .../models/nebius/minimax-m3.toml             | 15 ----------
 .../models/nebius/nemotron-3-nano-30b.toml    | 30 -------------------
 .../models/nebius/nemotron-3-nano-omni.toml   | 30 -------------------
 .../models/nebius/nemotron-3-super-120b.toml  | 30 -------------------
 .../models/nebius/nemotron-3-ultra-550b.toml  | 30 -------------------
 .../nebius/qwen2-5-vl-72b-instruct.toml       | 12 --------
 .../nebius/qwen3-235b-a22b-instruct-2507.toml | 11 -------
 .../nebius/qwen3-30b-a3b-instruct-2507.toml   | 22 --------------
 .../models/nebius/qwen3-32b.toml              | 12 --------
 .../nebius/qwen3-next-80b-a3b-thinking.toml   | 11 -------
 .../models/runware/glm-5.3-flash.toml         | 11 +++++++
 .../models/zai/glm-5.3-flash.toml             |  4 ---
 27 files changed, 11 insertions(+), 495 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/nebius/cosmos3-super-reasoner.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/deepseek-v4-pro.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/gemma-3-27b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/glm-5.1.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/glm-5.2.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/gpt-oss-120b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/hermes-4-405b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/hermes-4-70b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/kimi-k2.6.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/kimi-k2.7-code.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/kimi-k3.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/llama-3.1-nemotron-ultra-253b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/llama-3.3-70b-instruct.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/minicpm-v-4.5.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/minimax-m2.5.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/minimax-m3.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/nemotron-3-nano-30b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/nemotron-3-nano-omni.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/nemotron-3-super-120b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/nemotron-3-ultra-550b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/qwen2-5-vl-72b-instruct.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/qwen3-235b-a22b-instruct-2507.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/qwen3-30b-a3b-instruct-2507.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/qwen3-32b.toml
 delete mode 100644 providers/llmgateway-providers/models/nebius/qwen3-next-80b-a3b-thinking.toml
 create mode 100644 providers/llmgateway-providers/models/runware/glm-5.3-flash.toml

diff --git a/providers/llmgateway-providers/models/nebius/cosmos3-super-reasoner.toml b/providers/llmgateway-providers/models/nebius/cosmos3-super-reasoner.toml
deleted file mode 100644
index 826aa36feab..00000000000
--- a/providers/llmgateway-providers/models/nebius/cosmos3-super-reasoner.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Cosmos 3 Super Reasoner (Nebius AI)"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-release_date = "2026-06-01"
-last_updated = "2026-06-01"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/deepseek-v4-pro.toml b/providers/llmgateway-providers/models/nebius/deepseek-v4-pro.toml
deleted file mode 100644
index e6af2f69621..00000000000
--- a/providers/llmgateway-providers/models/nebius/deepseek-v4-pro.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "deepseek/deepseek-v4-pro"
-name = "DeepSeek V4 Pro (Nebius AI)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 1.75
-output = 3.5
-
-[limit]
-context = 1_048_576
diff --git a/providers/llmgateway-providers/models/nebius/gemma-3-27b.toml b/providers/llmgateway-providers/models/nebius/gemma-3-27b.toml
deleted file mode 100644
index 2919cf2f4a8..00000000000
--- a/providers/llmgateway-providers/models/nebius/gemma-3-27b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Gemma 3 27B (Nebius AI)"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 110_000
-output = 110_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/glm-5.1.toml b/providers/llmgateway-providers/models/nebius/glm-5.1.toml
deleted file mode 100644
index c2c273568be..00000000000
--- a/providers/llmgateway-providers/models/nebius/glm-5.1.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "zhipuai/glm-5.1"
-name = "GLM-5.1 (Nebius AI)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 1.4
-output = 4.4
-
-[limit]
-context = 202_752
diff --git a/providers/llmgateway-providers/models/nebius/glm-5.2.toml b/providers/llmgateway-providers/models/nebius/glm-5.2.toml
deleted file mode 100644
index e6abb09ba21..00000000000
--- a/providers/llmgateway-providers/models/nebius/glm-5.2.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM-5.2 (Nebius AI)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 1.4
-output = 4.4
-
-[limit]
-context = 1_048_576
diff --git a/providers/llmgateway-providers/models/nebius/gpt-oss-120b.toml b/providers/llmgateway-providers/models/nebius/gpt-oss-120b.toml
deleted file mode 100644
index ff5c0f3cc92..00000000000
--- a/providers/llmgateway-providers/models/nebius/gpt-oss-120b.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "openai/gpt-oss-120b"
-name = "GPT OSS 120B (Nebius AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.15
-output = 0.6
diff --git a/providers/llmgateway-providers/models/nebius/hermes-4-405b.toml b/providers/llmgateway-providers/models/nebius/hermes-4-405b.toml
deleted file mode 100644
index 3d3b4ab3858..00000000000
--- a/providers/llmgateway-providers/models/nebius/hermes-4-405b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Hermes 4 405B (Nebius AI)"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "hermes"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 1
-output = 3
-
-[limit]
-context = 131_072
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/hermes-4-70b.toml b/providers/llmgateway-providers/models/nebius/hermes-4-70b.toml
deleted file mode 100644
index 0dbe173405e..00000000000
--- a/providers/llmgateway-providers/models/nebius/hermes-4-70b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Hermes 4 70B (Nebius AI)"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "hermes"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.13
-output = 0.4
-
-[limit]
-context = 131_072
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/kimi-k2.6.toml b/providers/llmgateway-providers/models/nebius/kimi-k2.6.toml
deleted file mode 100644
index c51b499d7c2..00000000000
--- a/providers/llmgateway-providers/models/nebius/kimi-k2.6.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "moonshotai/kimi-k2.6"
-name = "Kimi K2.6 (Nebius AI)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.95
-output = 4
diff --git a/providers/llmgateway-providers/models/nebius/kimi-k2.7-code.toml b/providers/llmgateway-providers/models/nebius/kimi-k2.7-code.toml
deleted file mode 100644
index e2a4830879b..00000000000
--- a/providers/llmgateway-providers/models/nebius/kimi-k2.7-code.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "moonshotai/kimi-k2.7-code"
-name = "Kimi K2.7 Code (Nebius AI)"
-attachment = false
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.95
-output = 4
-
-[modalities]
-input = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/kimi-k3.toml b/providers/llmgateway-providers/models/nebius/kimi-k3.toml
deleted file mode 100644
index 90d46512e7e..00000000000
--- a/providers/llmgateway-providers/models/nebius/kimi-k3.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "moonshotai/kimi-k3"
-name = "Kimi K3 (Nebius AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 3
-output = 15
-
-[limit]
-output = 1_048_576
diff --git a/providers/llmgateway-providers/models/nebius/llama-3.1-nemotron-ultra-253b.toml b/providers/llmgateway-providers/models/nebius/llama-3.1-nemotron-ultra-253b.toml
deleted file mode 100644
index 56575dab8ea..00000000000
--- a/providers/llmgateway-providers/models/nebius/llama-3.1-nemotron-ultra-253b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Llama 3.1 Nemotron Ultra 253B (Nebius AI)"
-description = "Flagship Nemotron model for high-throughput reasoning and complex agents"
-family = "nemotron"
-release_date = "2025-04-07"
-last_updated = "2025-04-07"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.6
-output = 1.8
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/llama-3.3-70b-instruct.toml b/providers/llmgateway-providers/models/nebius/llama-3.3-70b-instruct.toml
deleted file mode 100644
index ad45104cca9..00000000000
--- a/providers/llmgateway-providers/models/nebius/llama-3.3-70b-instruct.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "meta/llama-3.3-70b-instruct"
-name = "Llama 3.3 70B Instruct (Nebius AI)"
-attachment = false
-structured_output = false
-
-[cost]
-input = 0.13
-output = 0.4
diff --git a/providers/llmgateway-providers/models/nebius/minicpm-v-4.5.toml b/providers/llmgateway-providers/models/nebius/minicpm-v-4.5.toml
deleted file mode 100644
index b6759d8b6c5..00000000000
--- a/providers/llmgateway-providers/models/nebius/minicpm-v-4.5.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "MiniCPM-V 4.5 (Nebius AI)"
-description = "Multimodal model for analyzing text, images, documents, and rich media"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.658
-output = 1.11
-
-[limit]
-context = 32_000
-output = 32_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/minimax-m2.5.toml b/providers/llmgateway-providers/models/nebius/minimax-m2.5.toml
deleted file mode 100644
index 1cea2927551..00000000000
--- a/providers/llmgateway-providers/models/nebius/minimax-m2.5.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "minimax/MiniMax-M2.5"
-name = "MiniMax M2.5 (Nebius AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.3
-output = 1.2
-
-[limit]
-context = 196_608
-output = 131_100
diff --git a/providers/llmgateway-providers/models/nebius/minimax-m3.toml b/providers/llmgateway-providers/models/nebius/minimax-m3.toml
deleted file mode 100644
index 69e984b52dd..00000000000
--- a/providers/llmgateway-providers/models/nebius/minimax-m3.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "minimax/MiniMax-M3"
-name = "MiniMax M3 (Nebius AI)"
-attachment = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.3
-output = 1.2
-
-[modalities]
-input = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/nemotron-3-nano-30b.toml b/providers/llmgateway-providers/models/nebius/nemotron-3-nano-30b.toml
deleted file mode 100644
index 760c83f5f05..00000000000
--- a/providers/llmgateway-providers/models/nebius/nemotron-3-nano-30b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Nemotron 3 Nano 30B (Nebius AI)"
-description = "Compact Nemotron model for efficient reasoning and deployable AI agents"
-family = "nemotron"
-release_date = "2025-12-15"
-last_updated = "2025-12-15"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.06
-output = 0.24
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/nemotron-3-nano-omni.toml b/providers/llmgateway-providers/models/nebius/nemotron-3-nano-omni.toml
deleted file mode 100644
index 6dd6ae2ca16..00000000000
--- a/providers/llmgateway-providers/models/nebius/nemotron-3-nano-omni.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Nemotron 3 Nano Omni (Nebius AI)"
-description = "Omni-modal model for text, vision, audio, and multimodal agent tasks"
-family = "nemotron"
-release_date = "2026-04-28"
-last_updated = "2026-04-28"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.06
-output = 0.24
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/nemotron-3-super-120b.toml b/providers/llmgateway-providers/models/nebius/nemotron-3-super-120b.toml
deleted file mode 100644
index ac876940067..00000000000
--- a/providers/llmgateway-providers/models/nebius/nemotron-3-super-120b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Nemotron 3 Super 120B (Nebius AI)"
-description = "Nemotron model for efficient reasoning, coding, and specialized AI agents"
-family = "nemotron"
-release_date = "2026-03-11"
-last_updated = "2026-03-11"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.3
-output = 0.9
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/nemotron-3-ultra-550b.toml b/providers/llmgateway-providers/models/nebius/nemotron-3-ultra-550b.toml
deleted file mode 100644
index 4e533ea48bb..00000000000
--- a/providers/llmgateway-providers/models/nebius/nemotron-3-ultra-550b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-name = "Nemotron 3 Ultra 550B (Nebius AI)"
-description = "Flagship Nemotron model for high-throughput reasoning and complex agents"
-family = "nemotron"
-release_date = "2026-06-01"
-last_updated = "2026-06-01"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 1
-output = 3
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/qwen2-5-vl-72b-instruct.toml b/providers/llmgateway-providers/models/nebius/qwen2-5-vl-72b-instruct.toml
deleted file mode 100644
index 21c4e2d187f..00000000000
--- a/providers/llmgateway-providers/models/nebius/qwen2-5-vl-72b-instruct.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "alibaba/qwen2-5-vl-72b-instruct"
-name = "Qwen2.5 VL 72B Instruct (Nebius AI)"
-attachment = true
-tool_call = false
-structured_output = false
-
-[cost]
-input = 0.25
-output = 0.75
-
-[limit]
-context = 32_000
diff --git a/providers/llmgateway-providers/models/nebius/qwen3-235b-a22b-instruct-2507.toml b/providers/llmgateway-providers/models/nebius/qwen3-235b-a22b-instruct-2507.toml
deleted file mode 100644
index 51c2c748850..00000000000
--- a/providers/llmgateway-providers/models/nebius/qwen3-235b-a22b-instruct-2507.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "alibaba/qwen3-235b-a22b-instruct-2507"
-name = "Qwen3 235B A22B Instruct 2507 (Nebius AI)"
-structured_output = false
-
-[cost]
-input = 0.2
-output = 0.6
-
-[limit]
-context = 262_000
-output = 8_192
diff --git a/providers/llmgateway-providers/models/nebius/qwen3-30b-a3b-instruct-2507.toml b/providers/llmgateway-providers/models/nebius/qwen3-30b-a3b-instruct-2507.toml
deleted file mode 100644
index e4b4eefbf6b..00000000000
--- a/providers/llmgateway-providers/models/nebius/qwen3-30b-a3b-instruct-2507.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Qwen3 30B A3B Instruct 2507 (Nebius AI)"
-description = "Tool-capable chat model for instruction following and agentic application workflows"
-release_date = "2025-07-30"
-last_updated = "2025-07-30"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 262_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway-providers/models/nebius/qwen3-32b.toml b/providers/llmgateway-providers/models/nebius/qwen3-32b.toml
deleted file mode 100644
index e8507cc1b85..00000000000
--- a/providers/llmgateway-providers/models/nebius/qwen3-32b.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "alibaba/qwen3-32b"
-name = "Qwen3 32B (Nebius AI)"
-reasoning = false
-structured_output = false
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 40_960
-output = 8_192
diff --git a/providers/llmgateway-providers/models/nebius/qwen3-next-80b-a3b-thinking.toml b/providers/llmgateway-providers/models/nebius/qwen3-next-80b-a3b-thinking.toml
deleted file mode 100644
index 8cbde7bf0d6..00000000000
--- a/providers/llmgateway-providers/models/nebius/qwen3-next-80b-a3b-thinking.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "alibaba/qwen3-next-80b-a3b-thinking"
-name = "Qwen3 Next 80B A3B Thinking (Nebius AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.15
-output = 1.2
diff --git a/providers/llmgateway-providers/models/runware/glm-5.3-flash.toml b/providers/llmgateway-providers/models/runware/glm-5.3-flash.toml
new file mode 100644
index 00000000000..caac83ee8bf
--- /dev/null
+++ b/providers/llmgateway-providers/models/runware/glm-5.3-flash.toml
@@ -0,0 +1,11 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (Runware)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
diff --git a/providers/llmgateway-providers/models/zai/glm-5.3-flash.toml b/providers/llmgateway-providers/models/zai/glm-5.3-flash.toml
index bfadb0fa4c9..f88b263f11c 100644
--- a/providers/llmgateway-providers/models/zai/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/zai/glm-5.3-flash.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM-5.3 Flash (Z AI)"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,6 +13,3 @@ cache_read = 0.03
 
 [limit]
 context = 1_048_576
-
-[modalities]
-input = ["text"]

From afb7ece0c3ce163998d90609161007ce62522d04 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:05:55 -0500
Subject: [PATCH 0338/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#5891)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/cosmos3-super-reasoner.toml        | 29 ------------------
 providers/llmgateway/models/gemma-3-27b.toml  | 23 --------------
 .../llmgateway/models/glm-5.3-flash.toml      |  2 +-
 .../llmgateway/models/hermes-4-405b.toml      | 30 -------------------
 providers/llmgateway/models/hermes-4-70b.toml | 30 -------------------
 .../models/llama-3.1-nemotron-ultra-253b.toml |  8 -----
 .../models/llama-3.3-70b-instruct.toml        |  2 +-
 .../llmgateway/models/minicpm-v-4.5.toml      | 22 --------------
 .../models/nemotron-3-nano-30b.toml           | 30 -------------------
 .../models/nemotron-3-nano-omni.toml          | 30 -------------------
 .../models/nemotron-3-super-120b.toml         | 30 -------------------
 .../models/qwen2-5-vl-72b-instruct.toml       |  8 -----
 .../models/qwen3-30b-a3b-instruct-2507.toml   | 23 --------------
 providers/llmgateway/models/qwen3-32b.toml    |  6 ++--
 14 files changed, 5 insertions(+), 268 deletions(-)
 delete mode 100644 providers/llmgateway/models/cosmos3-super-reasoner.toml
 delete mode 100644 providers/llmgateway/models/gemma-3-27b.toml
 delete mode 100644 providers/llmgateway/models/hermes-4-405b.toml
 delete mode 100644 providers/llmgateway/models/hermes-4-70b.toml
 delete mode 100644 providers/llmgateway/models/llama-3.1-nemotron-ultra-253b.toml
 delete mode 100644 providers/llmgateway/models/minicpm-v-4.5.toml
 delete mode 100644 providers/llmgateway/models/nemotron-3-nano-30b.toml
 delete mode 100644 providers/llmgateway/models/nemotron-3-nano-omni.toml
 delete mode 100644 providers/llmgateway/models/nemotron-3-super-120b.toml
 delete mode 100644 providers/llmgateway/models/qwen2-5-vl-72b-instruct.toml
 delete mode 100644 providers/llmgateway/models/qwen3-30b-a3b-instruct-2507.toml

diff --git a/providers/llmgateway/models/cosmos3-super-reasoner.toml b/providers/llmgateway/models/cosmos3-super-reasoner.toml
deleted file mode 100644
index 7a78bfe25ff..00000000000
--- a/providers/llmgateway/models/cosmos3-super-reasoner.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Cosmos 3 Super Reasoner"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-release_date = "2026-06-01"
-last_updated = "2026-06-01"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway/models/gemma-3-27b.toml b/providers/llmgateway/models/gemma-3-27b.toml
deleted file mode 100644
index 67fa0c85a7a..00000000000
--- a/providers/llmgateway/models/gemma-3-27b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Gemma 3 27B"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 110_000
-output = 110_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index e11959adcb4..5a002179e53 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -2,7 +2,7 @@ base_model = "zhipuai/glm-5.3-flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "max"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.13
diff --git a/providers/llmgateway/models/hermes-4-405b.toml b/providers/llmgateway/models/hermes-4-405b.toml
deleted file mode 100644
index b525acfa4c9..00000000000
--- a/providers/llmgateway/models/hermes-4-405b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Hermes 4 405B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "hermes"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 1
-output = 3
-
-[limit]
-context = 131_072
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway/models/hermes-4-70b.toml b/providers/llmgateway/models/hermes-4-70b.toml
deleted file mode 100644
index 8dfe55419f1..00000000000
--- a/providers/llmgateway/models/hermes-4-70b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Hermes 4 70B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "hermes"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 0.13
-output = 0.4
-
-[limit]
-context = 131_072
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway/models/llama-3.1-nemotron-ultra-253b.toml b/providers/llmgateway/models/llama-3.1-nemotron-ultra-253b.toml
deleted file mode 100644
index c27f84b351e..00000000000
--- a/providers/llmgateway/models/llama-3.1-nemotron-ultra-253b.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "nvidia/llama-3.1-nemotron-ultra-253b"
-reasoning = false
-tool_call = false
-structured_output = true
-
-[cost]
-input = 0.6
-output = 1.8
diff --git a/providers/llmgateway/models/llama-3.3-70b-instruct.toml b/providers/llmgateway/models/llama-3.3-70b-instruct.toml
index f4694d9b75a..1609c1b2b94 100644
--- a/providers/llmgateway/models/llama-3.3-70b-instruct.toml
+++ b/providers/llmgateway/models/llama-3.3-70b-instruct.toml
@@ -1,7 +1,7 @@
 base_model = "meta/llama-3.3-70b-instruct"
 
 [cost]
-input = 0.13
+input = 0.135
 output = 0.4
 
 [limit]
diff --git a/providers/llmgateway/models/minicpm-v-4.5.toml b/providers/llmgateway/models/minicpm-v-4.5.toml
deleted file mode 100644
index 2b50f6bcbd8..00000000000
--- a/providers/llmgateway/models/minicpm-v-4.5.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "MiniCPM-V 4.5"
-description = "Multimodal model for analyzing text, images, documents, and rich media"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.658
-output = 1.11
-
-[limit]
-context = 32_000
-output = 32_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway/models/nemotron-3-nano-30b.toml b/providers/llmgateway/models/nemotron-3-nano-30b.toml
deleted file mode 100644
index 8b8d3c92db6..00000000000
--- a/providers/llmgateway/models/nemotron-3-nano-30b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Nemotron 3 Nano 30B"
-description = "Compact Nemotron model for efficient reasoning and deployable AI agents"
-family = "nemotron"
-release_date = "2025-12-15"
-last_updated = "2025-12-15"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 0.06
-output = 0.24
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway/models/nemotron-3-nano-omni.toml b/providers/llmgateway/models/nemotron-3-nano-omni.toml
deleted file mode 100644
index 81a7a15fbc4..00000000000
--- a/providers/llmgateway/models/nemotron-3-nano-omni.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Nemotron 3 Nano Omni"
-description = "Omni-modal model for text, vision, audio, and multimodal agent tasks"
-family = "nemotron"
-release_date = "2026-04-28"
-last_updated = "2026-04-28"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 0.06
-output = 0.24
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/llmgateway/models/nemotron-3-super-120b.toml b/providers/llmgateway/models/nemotron-3-super-120b.toml
deleted file mode 100644
index 57a9d188262..00000000000
--- a/providers/llmgateway/models/nemotron-3-super-120b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Gateway-wide $.reasoning_effort = "none" (or $.reasoning.effort = "none")
-# disables non-OpenAI reasoning; this model has no documented effort tiers
-# or exact reasoning-token budget.
-# https://docs.llmgateway.io/features/reasoning (accessed 2026-07-23)
-name = "Nemotron 3 Super 120B"
-description = "Nemotron model for efficient reasoning, coding, and specialized AI agents"
-family = "nemotron"
-release_date = "2026-03-11"
-last_updated = "2026-03-11"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
-
-[cost]
-input = 0.3
-output = 0.9
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway/models/qwen2-5-vl-72b-instruct.toml b/providers/llmgateway/models/qwen2-5-vl-72b-instruct.toml
deleted file mode 100644
index 823e43d3ec7..00000000000
--- a/providers/llmgateway/models/qwen2-5-vl-72b-instruct.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "alibaba/qwen2-5-vl-72b-instruct"
-
-[cost]
-input = 0.25
-output = 0.75
-
-[limit]
-context = 32_000
diff --git a/providers/llmgateway/models/qwen3-30b-a3b-instruct-2507.toml b/providers/llmgateway/models/qwen3-30b-a3b-instruct-2507.toml
deleted file mode 100644
index 732be1543c5..00000000000
--- a/providers/llmgateway/models/qwen3-30b-a3b-instruct-2507.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen3 30B A3B Instruct (2507)"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-07-08"
-last_updated = "2025-07-08"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 262_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/llmgateway/models/qwen3-32b.toml b/providers/llmgateway/models/qwen3-32b.toml
index 7eec83a099a..31f2b85bd38 100644
--- a/providers/llmgateway/models/qwen3-32b.toml
+++ b/providers/llmgateway/models/qwen3-32b.toml
@@ -5,11 +5,11 @@
 base_model = "alibaba/qwen3-32b"
 
 [[reasoning_options]]
-type = "toggle" # API: {"reasoning_effort": "none"} or {"reasoning": {"effort": "none"}}
+type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.3
+input = 0.36
+output = 0.87
 reasoning = 8.4
 
 [limit]

From 29b2c5fce24a173fc9a681d5e39d40770426e807 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:06:05 -0500
Subject: [PATCH 0339/2012] chore(sync): update Requesty model catalog (#5879)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../requesty/models/deepseek-v4-flash-0731.toml   | 15 ++++-----------
 providers/requesty/models/glm-5.2.toml            |  6 +++---
 providers/requesty/models/glm-5.3-flash.toml      |  9 ++++-----
 providers/requesty/models/glm-5.3.toml            |  8 +++++---
 providers/requesty/models/glm-5.3@eu.toml         |  7 ++++---
 5 files changed, 20 insertions(+), 25 deletions(-)

diff --git a/providers/requesty/models/deepseek-v4-flash-0731.toml b/providers/requesty/models/deepseek-v4-flash-0731.toml
index d73d297b44e..d5b37a5c0f8 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731.toml
@@ -1,17 +1,10 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
+reasoning = false
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.07
+input = 0.076
+output = 0.153
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
-output = 131_072
diff --git a/providers/requesty/models/glm-5.2.toml b/providers/requesty/models/glm-5.2.toml
index bab0f699fff..b2d4e10b216 100644
--- a/providers/requesty/models/glm-5.2.toml
+++ b/providers/requesty/models/glm-5.2.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.2
-output = 4.2
-cache_read = 0.26
+input = 0.8
+output = 2.55
+cache_read = 0.16
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
index 394d0fef42b..f918675673a 100644
--- a/providers/requesty/models/glm-5.3-flash.toml
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -1,5 +1,4 @@
 base_model = "zhipuai/glm-5.3-flash"
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -9,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
-output = 128_000
+context = 131_072
diff --git a/providers/requesty/models/glm-5.3.toml b/providers/requesty/models/glm-5.3.toml
index baa2d5f5503..57ec57604a4 100644
--- a/providers/requesty/models/glm-5.3.toml
+++ b/providers/requesty/models/glm-5.3.toml
@@ -1,4 +1,5 @@
 base_model = "zhipuai/glm-5.3"
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -8,9 +9,10 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.4
-output = 4.4
+input = 1.2
+output = 4.2
 cache_read = 0.26
 
 [limit]
-output = 128_000
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/requesty/models/glm-5.3@eu.toml b/providers/requesty/models/glm-5.3@eu.toml
index 0b46fbb34c4..7e0a95b8ea3 100644
--- a/providers/requesty/models/glm-5.3@eu.toml
+++ b/providers/requesty/models/glm-5.3@eu.toml
@@ -5,6 +5,7 @@
 # https://docs.requesty.ai/features/reasoning
 base_model = "zhipuai/glm-5.3"
 name = "GLM-5.3 (EU)"
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -14,9 +15,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 1.75
-output = 4.5
-cache_read = 0.44
+input = 1.2
+output = 4.2
+cache_read = 0.26
 
 [limit]
 context = 1_048_576

From 620f871d634d78f71b4e3b98a321187a2613d285 Mon Sep 17 00:00:00 2001
From: MickeyXin <49945543+fx247562340@users.noreply.github.com>
Date: Tue, 1 Sep 2026 05:06:36 +0800
Subject: [PATCH 0340/2012] feat(provider): add four Vancine models (#5866)

Co-authored-by: fx247562340 <fx247562340@users.noreply.github.com>
---
 providers/vancine/models/LongCat-2.0.toml     | 19 +++++++++++++
 .../models/deepseek-v4-flash-vision-exp.toml  | 28 +++++++++++++++++++
 providers/vancine/models/hy4-preview.toml     | 17 +++++++++++
 providers/vancine/models/mimo-v2.5-pro.toml   | 20 +++++++++++++
 4 files changed, 84 insertions(+)
 create mode 100644 providers/vancine/models/LongCat-2.0.toml
 create mode 100644 providers/vancine/models/deepseek-v4-flash-vision-exp.toml
 create mode 100644 providers/vancine/models/hy4-preview.toml
 create mode 100644 providers/vancine/models/mimo-v2.5-pro.toml

diff --git a/providers/vancine/models/LongCat-2.0.toml b/providers/vancine/models/LongCat-2.0.toml
new file mode 100644
index 00000000000..3958f79bf1a
--- /dev/null
+++ b/providers/vancine/models/LongCat-2.0.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-08-30):
+# models/meituan/longcat-2.0.toml
+# providers/longcat/models/LongCat-2.0.toml
+# Toggle: thinking.type = "enabled" | "disabled".
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.75 / output 2.95 / cache_read 0.015.
+# Vancine public pricing (/api/pricing) is 0.24 / 0.96 / n/a / 0.0048 (USD/MTok).
+base_model = "meituan/longcat-2.0"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.24
+output = 0.96
+cache_read = 0.0048
diff --git a/providers/vancine/models/deepseek-v4-flash-vision-exp.toml b/providers/vancine/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..9a7df2b3b8c
--- /dev/null
+++ b/providers/vancine/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,28 @@
+# Sources (accessed 2026-08-30):
+# https://api-docs.deepseek.com/quick_start/pricing/
+# https://api-docs.deepseek.com/guides/vision
+# models/deepseek/deepseek-v4-flash-vision-exp.toml
+# providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
+# providers/vancine/models/deepseek-v4-flash.toml
+# Toggle: thinking.type = "enabled" | "disabled".
+# Effort: reasoning_effort = "low" | "high" | "max". Priced the same as
+# DeepSeek V4 Flash; reasoning tokens are billed at the output rate.
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.14 / output 0.28 / reasoning 0.28 / cache_read 0.0028.
+# Vancine public pricing (/api/pricing) is 0.22 / 0.66 / n/a / 0.007 (USD/MTok).
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
diff --git a/providers/vancine/models/hy4-preview.toml b/providers/vancine/models/hy4-preview.toml
new file mode 100644
index 00000000000..6a2c758be98
--- /dev/null
+++ b/providers/vancine/models/hy4-preview.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-08-30):
+# models/tencent/hy4-preview.toml
+# providers/opencode-go/models/hy4-preview.toml
+# Effort: reasoning_effort = "none" | "high". Off is effort = "none", so
+# no toggle is authored. No interleaved field evidence on this host.
+# Lab cost (Tencent list): input 0.834 / output 2.501 / cache_read 0.042.
+# Vancine public pricing (/api/pricing) is 0.67 / 2.00 / 0.034 (USD/MTok).
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.67
+output = 2.00
+cache_read = 0.034
diff --git a/providers/vancine/models/mimo-v2.5-pro.toml b/providers/vancine/models/mimo-v2.5-pro.toml
new file mode 100644
index 00000000000..29b4040f1af
--- /dev/null
+++ b/providers/vancine/models/mimo-v2.5-pro.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-08-30):
+# https://platform.xiaomimimo.com/static/docs/api/chat/openai-api.md
+# models/xiaomi/mimo-v2.5-pro.toml
+# providers/xiaomi/models/mimo-v2.5-pro.toml
+# Toggle: thinking.type = "enabled" | "disabled".
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.435 / output 0.87 / cache_read 0.0036.
+# Vancine public pricing (/api/pricing) is 0.348 / 0.696 / n/a / 0.00288 (USD/MTok).
+base_model = "xiaomi/mimo-v2.5-pro"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.348
+output = 0.696
+cache_read = 0.00288

From d9c9e0fa336ce2b8eaf42c889f910b13f6b58e54 Mon Sep 17 00:00:00 2001
From: Pessimist228 <artemburej5@gmail.com>
Date: Mon, 31 Aug 2026 23:06:45 +0200
Subject: [PATCH 0341/2012] llmtech: enable vision on Qwen3.8-27B-NVFP4 (#5865)

* llmtech: enable vision (attachment + image modality)

* llmtech: drop redundant attachment, keep modality delta
---
 .../models/unsloth/Qwen3.8-27B-NVFP4.toml     | 45 +++++++++----------
 1 file changed, 22 insertions(+), 23 deletions(-)

diff --git a/providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml b/providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml
index dbad694e9bb..ac3503d02a3 100644
--- a/providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml
+++ b/providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml
@@ -1,23 +1,22 @@
-# Reasoning wire (OpenAI-compatible chat completions at https://api.llmtech.eu/v1):
-# Toggle: chat_template_kwargs.enable_thinking = true | false.
-# Effort: chat_template_kwargs.reasoning_effort = "low" | "medium" | "xhigh".
-# Non-streaming responses carry reasoning text in the `reasoning` field;
-# streaming deltas use `reasoning_content`.
-
-base_model = "alibaba/qwen3.8-27b"
-attachment = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "xhigh"]
-
-[cost]
-input = 0.25
-output = 2.09
-cache_read = 0.04
-
-[modalities]
-input = ["text"]
+# Reasoning wire (OpenAI-compatible chat completions at https://api.llmtech.eu/v1):
+# Toggle: chat_template_kwargs.enable_thinking = true | false.
+# Effort: chat_template_kwargs.reasoning_effort = "low" | "medium" | "xhigh".
+# Non-streaming responses carry reasoning text in the `reasoning` field;
+# streaming deltas use `reasoning_content`.
+
+base_model = "alibaba/qwen3.8-27b"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.25
+output = 2.09
+cache_read = 0.04
+
+[modalities]
+input = ["text", "image"]

From c65ab7d2b35e275b77a90b5664a8cdcc484ce12b Mon Sep 17 00:00:00 2001
From: jarvislee <jarvislee90s@gmail.com>
Date: Tue, 1 Sep 2026 05:07:28 +0800
Subject: [PATCH 0342/2012] fix(catalog): align attachment flag with modalities
 for kimi-k2.5 and grok-code (#5793)

* fix(moonshotai): align kimi-k2.5 attachment flag with its own modalities

The provider entry sets attachment=false while its [modalities] declares
image+video input, its lab entry (models/moonshotai/kimi-k2.5.toml) sets
attachment=true, and Moonshot's official model list documents kimi-k2.5 as
supporting 'visual and text input'. Per the catalog convention (attachment
derived from non-text input modalities) and the provider-file override-only
rule, attachment must be true.

Consumers (e.g. OpenCode) read this flag to gate image attachments; the
current value blocks image input for a vision-capable model (see
anomalyco/opencode#23427).

* fix(opencode): disable attachment flag for text-only grok-code

grok-code (xAI Grok Code Fast 1) accepts text input only, but its entry
sets attachment=true. Per the catalog's derivation convention (attachment
from non-text input modalities) the flag must be false; the current value
makes consumers offer attachment flows that cannot succeed. Entry is
already status=deprecated; no other fields touched.

---------

Co-authored-by: jarvislee90s-dot <266288279+jarvislee90s-dot@users.noreply.github.com>
---
 providers/moonshotai/models/kimi-k2.5.toml | 2 +-
 providers/opencode/models/grok-code.toml   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/moonshotai/models/kimi-k2.5.toml b/providers/moonshotai/models/kimi-k2.5.toml
index cb64df827ce..9db35770f6b 100644
--- a/providers/moonshotai/models/kimi-k2.5.toml
+++ b/providers/moonshotai/models/kimi-k2.5.toml
@@ -3,7 +3,7 @@ description = "Earlier Kimi frontier model for long-context agents, coding, and
 family = "kimi-k2"
 release_date = "2026-01"
 last_updated = "2026-01"
-attachment = false
+attachment = true
 reasoning = true
 structured_output = true
 temperature = false
diff --git a/providers/opencode/models/grok-code.toml b/providers/opencode/models/grok-code.toml
index c3a1f2fac08..caf8e8c27ae 100644
--- a/providers/opencode/models/grok-code.toml
+++ b/providers/opencode/models/grok-code.toml
@@ -3,7 +3,7 @@ description = "Legacy model retained for compatibility with older integrations"
 family = "grok"
 release_date = "2025-08-20"
 last_updated = "2025-08-20"
-attachment = true
+attachment = false
 reasoning = true
 reasoning_options = []
 temperature = true

From 893cea19a08d91aebfa3ac5aaa28ed2ec18118da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:07:58 -0500
Subject: [PATCH 0343/2012] chore(sync): update Cortecs model catalog (#5914)

* chore(sync): update Cortecs model catalog

* fix(cortecs): add DeepSeek V4 Pro reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden <rekram1-node@users.noreply.github.com>
---
 providers/cortecs/models/deepseek-v4-pro-0813.toml | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/cortecs/models/deepseek-v4-pro-0813.toml

diff --git a/providers/cortecs/models/deepseek-v4-pro-0813.toml b/providers/cortecs/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..181ae5d1396
--- /dev/null
+++ b/providers/cortecs/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,11 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2
+output = 3.999
+cache_read = 0.5
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From c56ca3c369ef8f4ad98d23a0e977fafdba9aa846 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 16:08:30 -0500
Subject: [PATCH 0344/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5900)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add Qwen 3.8 Flash Next reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/alibaba/qwen3.8-flash-next.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/vercel/models/alibaba/qwen3.8-flash-next.toml

diff --git a/providers/vercel/models/alibaba/qwen3.8-flash-next.toml b/providers/vercel/models/alibaba/qwen3.8-flash-next.toml
new file mode 100644
index 00000000000..a905f1db633
--- /dev/null
+++ b/providers/vercel/models/alibaba/qwen3.8-flash-next.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-flash-next"
+name = "Qwen 3.8 Flash Next"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "xhigh"] }]
+
+[cost]
+input = 0.12
+output = 0.4
+cache_read = 0.01
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]

From 978eadabcf949068075f32956f15729a2861fe52 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 21:23:28 +0000
Subject: [PATCH 0345/2012] chore(sync): update OpenRouter model catalog
 (#5949)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/ibm-granite/granite-4.2-8b.toml     | 4 ++--
 .../models/qwen/qwen3-next-80b-a3b-instruct.toml          | 5 +++--
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 5 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 07235ccdb08..92d05b69dd5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.030776
-output = 2.061552
-cache_read = 0.085898
+input = 1.02747
+output = 2.05494
+cache_read = 0.085623
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
index 873558d6ea7..1b683dd536e 100644
--- a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
+++ b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
@@ -6,8 +6,8 @@ last_updated = "2026-08-31"
 attachment = false
 reasoning = true
 temperature = true
-tool_call = false
-structured_output = false
+tool_call = true
+structured_output = true
 open_weights = true
 
 [[reasoning_options]]
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 64deac24f44..31bb83cc8bb 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -2,9 +2,10 @@ base_model = "alibaba/qwen3-next-80b-a3b-instruct"
 structured_output = true
 
 [cost]
-input = 0.09
+input = 0.1
 output = 1.1
+cache_read = 0.07
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0e324e4270a..d6026834e83 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1875
-output = 4.18
-cache_read = 0.247
+input = 1.17
+output = 3.96
+cache_read = 0.234
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From c0a40a57bba893d8fbe6cf708123d07d346f1110 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 21:23:35 +0000
Subject: [PATCH 0346/2012] chore(sync): update Kilo model catalog (#5950)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ibm-granite/granite-4.2-8b.toml    | 28 +++++++++++++++++++
 .../qwen/qwen3-next-80b-a3b-instruct.toml     |  2 +-
 providers/kilo/models/~z-ai/glm-latest.toml   | 10 +++----
 3 files changed, 34 insertions(+), 6 deletions(-)
 create mode 100644 providers/kilo/models/ibm-granite/granite-4.2-8b.toml

diff --git a/providers/kilo/models/ibm-granite/granite-4.2-8b.toml b/providers/kilo/models/ibm-granite/granite-4.2-8b.toml
new file mode 100644
index 00000000000..c99650d60ce
--- /dev/null
+++ b/providers/kilo/models/ibm-granite/granite-4.2-8b.toml
@@ -0,0 +1,28 @@
+name = "IBM: Granite 4.2 8B"
+description = "Granite 4.2 8B is a dense reasoning model from IBM. It is suited for mathematics, code generation, multilingual dialogue, and agentic workflows that need multi-step reasoning. It supports full, low-effort,..."
+family = "granite"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.1
+output = 0.15
+cache_read = 0.05
+
+[limit]
+context = 131_072
+output = 117_964
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 8071bb6d3c0..8247cd4ca01 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.78
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 4fbd2317736..a2f41897231 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1875
-output = 4.18
-cache_read = 0.247
+input = 1.17
+output = 3.96
+cache_read = 0.234
 
 [limit]
-context = 262_144
-output = 131_072
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 1578453e018aa8cee694b53c6b3f8fb535d60507 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 21:23:41 +0000
Subject: [PATCH 0347/2012] chore(sync): update Merge Gateway model catalog
 (#5951)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 35e52350151..f30ae74cdb3 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From abe614b9a73dce8b1719a3fd230a94409851d774 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 22:24:54 +0000
Subject: [PATCH 0348/2012] chore(sync): update NanoGPT model catalog (#5952)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ibm-granite/granite-4.2-8b.toml    | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/nano-gpt/models/ibm-granite/granite-4.2-8b.toml

diff --git a/providers/nano-gpt/models/ibm-granite/granite-4.2-8b.toml b/providers/nano-gpt/models/ibm-granite/granite-4.2-8b.toml
new file mode 100644
index 00000000000..21c84dcc421
--- /dev/null
+++ b/providers/nano-gpt/models/ibm-granite/granite-4.2-8b.toml
@@ -0,0 +1,28 @@
+name = "Granite 4.2 8B"
+description = "IBM Granite 4.2 8B is an Apache 2.0-licensed dense model with native step-by-step reasoning and specialized training for agentic work. It can plan before acting, sequence tools, navigate codebases, work in terminals, and verify results across coding, search, mathematics, science, and complex instruction-following tasks."
+family = "granite"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.1
+output = 0.15
+cache_read = 0.05
+
+[limit]
+context = 131_072
+input = 131_072
+output = 117_964
+
+[modalities]
+input = ["text"]
+output = ["text"]

From f0856b6c96167c63f2fb0f17ac3def896176ae5e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 22:25:47 +0000
Subject: [PATCH 0349/2012] chore(sync): update OpenRouter model catalog
 (#5953)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92d05b69dd5..a010dd38ba2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02747
-output = 2.05494
-cache_read = 0.085623
+input = 0.87
+output = 1.74
+cache_read = 0.0725
 
 [limit]
 context = 1_048_576

From eda98d420a05fbc3d7c87a8805985d24b2c2b79b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 22:30:30 +0000
Subject: [PATCH 0350/2012] chore(sync): update Google model catalog (#5954)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gemini-robotics-er-1.6-preview.toml | 18 ------------------
 1 file changed, 18 deletions(-)
 delete mode 100644 providers/google/models/gemini-robotics-er-1.6-preview.toml

diff --git a/providers/google/models/gemini-robotics-er-1.6-preview.toml b/providers/google/models/gemini-robotics-er-1.6-preview.toml
deleted file mode 100644
index 826ca678aae..00000000000
--- a/providers/google/models/gemini-robotics-er-1.6-preview.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-# Sources:
-# - https://ai.google.dev/gemini-api/docs/pricing (Gemini Robotics-ER 1.6 Preview)
-# - https://ai.google.dev/gemini-api/docs/robotics-overview (thinking_budget)
-# - https://ai.google.dev/gemini-api/docs/models/gemini-robotics-er-1.6-preview
-
-base_model = "google/gemini-robotics-er-1.6-preview"
-
-[[reasoning_options]]
-type = "toggle" # API: thinking_config.thinking_budget = 0 disables thinking
-
-[[reasoning_options]]
-type = "budget_tokens" # API: thinking_config.thinking_budget
-min = 0
-
-[cost]
-input = 1.0
-output = 5.0
-input_audio = 2.0

From a7e25b37ea7e7a363d6d9c4e068c872ef77b7d7e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 23:23:21 +0000
Subject: [PATCH 0351/2012] chore(sync): update Eden AI model catalog (#5956)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/flexai/deepseek-v4-flash-0731.toml       |  4 ++--
 .../models/moonshot/kimi-k2.7-code-highspeed.toml   | 13 +++++++++++++
 2 files changed, 15 insertions(+), 2 deletions(-)
 create mode 100644 providers/edenai/models/moonshot/kimi-k2.7-code-highspeed.toml

diff --git a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml b/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
index 8d7dbd5484c..7a18dfdfacb 100644
--- a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.08
-output = 0.18
+input = 0.03
+output = 0.1
 
 [limit]
 context = 786_432
diff --git a/providers/edenai/models/moonshot/kimi-k2.7-code-highspeed.toml b/providers/edenai/models/moonshot/kimi-k2.7-code-highspeed.toml
new file mode 100644
index 00000000000..3153080377b
--- /dev/null
+++ b/providers/edenai/models/moonshot/kimi-k2.7-code-highspeed.toml
@@ -0,0 +1,13 @@
+base_model = "moonshotai/kimi-k2.7-code-highspeed"
+attachment = false
+tool_call = false
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 1.9
+output = 8
+cache_read = 0.38
+
+[modalities]
+input = ["text"]

From 0c257d337edf8b611c81f2c08878e43c56faaa73 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 23:23:24 +0000
Subject: [PATCH 0352/2012] chore(sync): update OpenRouter model catalog
 (#5955)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576

From ed419f7047fdba5cffd0539c4081f94a9b6c138e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:10:34 +0000
Subject: [PATCH 0353/2012] chore(sync): update Requesty model catalog (#5960)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/gemini-3-pro-image.toml       | 7 -------
 providers/requesty/models/gemini-3.1-flash-image.toml   | 5 -----
 providers/requesty/models/gemini-3.1-flash-lite.toml    | 7 -------
 providers/requesty/models/gemini-3.1-flash-lite@eu.toml | 7 -------
 providers/requesty/models/gemini-3.1-pro-preview.toml   | 4 ++--
 5 files changed, 2 insertions(+), 28 deletions(-)

diff --git a/providers/requesty/models/gemini-3-pro-image.toml b/providers/requesty/models/gemini-3-pro-image.toml
index 21eb5f8c6ce..804c54fa8e9 100644
--- a/providers/requesty/models/gemini-3-pro-image.toml
+++ b/providers/requesty/models/gemini-3-pro-image.toml
@@ -15,12 +15,5 @@ output = 12
 cache_read = 0.2
 cache_write = 4.5
 
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 18
-cache_read = 0.2
-cache_write = 4.5
-
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/gemini-3.1-flash-image.toml b/providers/requesty/models/gemini-3.1-flash-image.toml
index c3b869b378a..e79038dbf6a 100644
--- a/providers/requesty/models/gemini-3.1-flash-image.toml
+++ b/providers/requesty/models/gemini-3.1-flash-image.toml
@@ -12,8 +12,3 @@ type = "budget_tokens"
 [cost]
 input = 0.5
 output = 2
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 0.5
-output = 2
diff --git a/providers/requesty/models/gemini-3.1-flash-lite.toml b/providers/requesty/models/gemini-3.1-flash-lite.toml
index f45173eeab0..09bc70069c5 100644
--- a/providers/requesty/models/gemini-3.1-flash-lite.toml
+++ b/providers/requesty/models/gemini-3.1-flash-lite.toml
@@ -13,12 +13,5 @@ output = 1.5
 cache_read = 0.025
 cache_write = 0.08333
 
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 0.5
-output = 2.25
-cache_read = 0.025
-cache_write = 0.08333
-
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.1-flash-lite@eu.toml b/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
index a5170790571..01e801b06ac 100644
--- a/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
+++ b/providers/requesty/models/gemini-3.1-flash-lite@eu.toml
@@ -14,12 +14,5 @@ output = 1.65
 cache_read = 0.0275
 cache_write = 0.091663
 
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 0.55
-output = 2.475
-cache_read = 0.0275
-cache_write = 0.091663
-
 [limit]
 output = 65_535
diff --git a/providers/requesty/models/gemini-3.1-pro-preview.toml b/providers/requesty/models/gemini-3.1-pro-preview.toml
index 45392c2634d..994be938c6c 100644
--- a/providers/requesty/models/gemini-3.1-pro-preview.toml
+++ b/providers/requesty/models/gemini-3.1-pro-preview.toml
@@ -17,8 +17,8 @@ cache_write = 4.5
 tier = { type = "context", size = 200_000 }
 input = 4
 output = 18
-cache_read = 0.2
-cache_write = 4.5
+cache_read = 0.4
+cache_write = 9
 
 [limit]
 output = 65_535

From 1a9d57b459fbfdc71a54fbe8a26129e736935354 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:10:50 +0000
Subject: [PATCH 0354/2012] chore(sync): update OpenRouter model catalog
 (#5961)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 7 ++++---
 providers/openrouter/models/gryphe/mythomax-l2-13b.toml    | 6 +++---
 providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml | 7 +++----
 providers/openrouter/models/tencent/hy3.toml               | 6 +++---
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a010dd38ba2..95460b6f5f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.0725
+input = 1.6
+output = 3.2
+cache_read = 0.135
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
index 0ddba1dbce0..8a0b69511e3 100644
--- a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
@@ -11,12 +11,12 @@ knowledge = "2023-06-30"
 open_weights = true
 
 [cost]
-input = 0.06
-output = 0.06
+input = 0.4
+output = 0.6
 
 [limit]
 context = 8_192
-output = 3_686
+output = 7_372
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index 0d691163001..aa706cc488b 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,13 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.0875
-output = 0.35
-cache_read = 0.0175
+input = 0.09
+output = 0.55
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 22777cc7821819b472110f49de85b47fd688ef30 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:11:13 +0000
Subject: [PATCH 0355/2012] chore(sync): update NanoGPT model catalog (#5962)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From eda76b28005ab4fe5038681a590ab393357d16fb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:11:21 +0000
Subject: [PATCH 0356/2012] chore(sync): update Venice model catalog (#5963)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/claude-sonnet-5.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/venice/models/claude-sonnet-5.toml b/providers/venice/models/claude-sonnet-5.toml
index c3b9665f1b1..04e34721ec0 100644
--- a/providers/venice/models/claude-sonnet-5.toml
+++ b/providers/venice/models/claude-sonnet-5.toml
@@ -8,10 +8,10 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 2
-output = 10.000000000000002
-cache_read = 0.2
-cache_write = 2.5000000000000004
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [limit]
 output = 64_000

From 1717d6792d83c8648005daa548dbbb8902797769 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:11:32 +0000
Subject: [PATCH 0357/2012] chore(sync): update Kilo model catalog (#5959)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml  | 3 ++-
 providers/kilo/models/gryphe/mythomax-l2-13b.toml    | 4 ++--
 providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml | 2 +-
 providers/kilo/models/tencent/hy3.toml               | 6 +++---
 4 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index 88500bd7dd8..e58e24281d4 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,4 +11,5 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/gryphe/mythomax-l2-13b.toml b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
index a7751171cf6..e402d2dc4a0 100644
--- a/providers/kilo/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
@@ -14,8 +14,8 @@ input = 0.06
 output = 0.06
 
 [limit]
-context = 4_096
-output = 3_686
+context = 8_192
+output = 7_372
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
index b356342358d..3f37ded3629 100644
--- a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
@@ -16,7 +16,7 @@ output = 0.598
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144

From 9a47a070e22e2d3b9ded7f4add89cf6d3de7ba8f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:42:42 +0000
Subject: [PATCH 0358/2012] chore(sync): update Kilo model catalog (#5965)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 778f98b65d9..da7301fa993 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_048_576
+context = 1_024_000

From 3b0487c2b809973b61c1562885b11782c0b3820c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 01:42:48 +0000
Subject: [PATCH 0359/2012] chore(sync): update OpenRouter model catalog
 (#5964)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576

From 7525dc310aab4f348b00080ecaab057c71031e03 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 03:38:08 +0000
Subject: [PATCH 0360/2012] chore(sync): update Kilo model catalog (#5968)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/anthracite-org/magnum-v4-72b.toml | 2 +-
 providers/kilo/models/mancer/weaver.toml                | 2 +-
 providers/kilo/models/nex-agi/nex-n2-mini.toml          | 2 +-
 providers/kilo/models/nex-agi/nex-n2-pro.toml           | 2 +-
 providers/kilo/models/undi95/remm-slerp-l2-13b.toml     | 2 +-
 5 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/anthracite-org/magnum-v4-72b.toml b/providers/kilo/models/anthracite-org/magnum-v4-72b.toml
index 181bbb3df41..0558f113c0a 100644
--- a/providers/kilo/models/anthracite-org/magnum-v4-72b.toml
+++ b/providers/kilo/models/anthracite-org/magnum-v4-72b.toml
@@ -10,7 +10,7 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 3
+input = 2.5
 output = 5
 
 [limit]
diff --git a/providers/kilo/models/mancer/weaver.toml b/providers/kilo/models/mancer/weaver.toml
index 9bd2ed94b99..498687047fe 100644
--- a/providers/kilo/models/mancer/weaver.toml
+++ b/providers/kilo/models/mancer/weaver.toml
@@ -11,7 +11,7 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0.5
+input = 0.4
 output = 0.75
 
 [limit]
diff --git a/providers/kilo/models/nex-agi/nex-n2-mini.toml b/providers/kilo/models/nex-agi/nex-n2-mini.toml
index 1cbab95cdaf..07eb0ca1d56 100644
--- a/providers/kilo/models/nex-agi/nex-n2-mini.toml
+++ b/providers/kilo/models/nex-agi/nex-n2-mini.toml
@@ -1,4 +1,4 @@
-name = "Nex AGI: Nex-N2-Mini"
+name = "Nex AGI: Nex-N2-Mini (retires Sep 4)"
 description = "Nex-N2-Mini is an open-source agentic mixture-of-experts model from Nex AGI, the smaller sibling in the Nex-N2 series. It accepts text and image input and is built for coding, tool use,..."
 family = "agi"
 release_date = "2026-06-24"
diff --git a/providers/kilo/models/nex-agi/nex-n2-pro.toml b/providers/kilo/models/nex-agi/nex-n2-pro.toml
index afe5b39162b..c0789522ba5 100644
--- a/providers/kilo/models/nex-agi/nex-n2-pro.toml
+++ b/providers/kilo/models/nex-agi/nex-n2-pro.toml
@@ -1,4 +1,4 @@
-name = "Nex AGI: Nex-N2-Pro"
+name = "Nex AGI: Nex-N2-Pro (retires Sep 4)"
 description = "Nex-N2-Pro is an agentic mixture-of-experts model from Nex AGI, with 17B active parameters out of 397B total. Built on the Qwen3.5 architecture, it accepts text and image input and produces..."
 family = "agi"
 release_date = "2026-06-08"
diff --git a/providers/kilo/models/undi95/remm-slerp-l2-13b.toml b/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
index a5a4bba82a0..96a93a9047f 100644
--- a/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
+++ b/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
@@ -10,7 +10,7 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.45
+input = 0.35
 output = 0.65
 
 [limit]

From 477a7ff526c1aa5da65dfb8b968ea7e6a97db520 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 03:38:28 +0000
Subject: [PATCH 0361/2012] chore(sync): update OpenRouter model catalog
 (#5966)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/anthracite-org/magnum-v4-72b.toml | 2 +-
 providers/openrouter/models/gryphe/mythomax-l2-13b.toml       | 2 +-
 providers/openrouter/models/mancer/weaver.toml                | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/anthracite-org/magnum-v4-72b.toml b/providers/openrouter/models/anthracite-org/magnum-v4-72b.toml
index e107403b036..8dbd3cf8d96 100644
--- a/providers/openrouter/models/anthracite-org/magnum-v4-72b.toml
+++ b/providers/openrouter/models/anthracite-org/magnum-v4-72b.toml
@@ -11,7 +11,7 @@ knowledge = "2024-06-30"
 open_weights = true
 
 [cost]
-input = 3
+input = 2.5
 output = 5
 
 [limit]
diff --git a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
index 8a0b69511e3..de962f15da2 100644
--- a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
@@ -11,7 +11,7 @@ knowledge = "2023-06-30"
 open_weights = true
 
 [cost]
-input = 0.4
+input = 0.35
 output = 0.6
 
 [limit]
diff --git a/providers/openrouter/models/mancer/weaver.toml b/providers/openrouter/models/mancer/weaver.toml
index 3bf39c850fa..1948144537d 100644
--- a/providers/openrouter/models/mancer/weaver.toml
+++ b/providers/openrouter/models/mancer/weaver.toml
@@ -12,7 +12,7 @@ knowledge = "2023-06-30"
 open_weights = false
 
 [cost]
-input = 0.5
+input = 0.4
 output = 0.75
 
 [limit]

From af6efa6d12373d99822132a24c5990146ecf69b0 Mon Sep 17 00:00:00 2001
From: MickeyXin <49945543+fx247562340@users.noreply.github.com>
Date: Tue, 1 Sep 2026 12:01:52 +0800
Subject: [PATCH 0362/2012] fix(provider): narrow Vancine model catalog (#5958)

Co-authored-by: fx247562340 <fx247562340@users.noreply.github.com>
---
 providers/vancine/models/LongCat-2.0.toml   | 19 -------------------
 providers/vancine/models/mimo-v2.5-pro.toml | 20 --------------------
 2 files changed, 39 deletions(-)
 delete mode 100644 providers/vancine/models/LongCat-2.0.toml
 delete mode 100644 providers/vancine/models/mimo-v2.5-pro.toml

diff --git a/providers/vancine/models/LongCat-2.0.toml b/providers/vancine/models/LongCat-2.0.toml
deleted file mode 100644
index 3958f79bf1a..00000000000
--- a/providers/vancine/models/LongCat-2.0.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-# Sources (accessed 2026-08-30):
-# models/meituan/longcat-2.0.toml
-# providers/longcat/models/LongCat-2.0.toml
-# Toggle: thinking.type = "enabled" | "disabled".
-# Interleaved field: reasoning_content.
-# Lab cost: input 0.75 / output 2.95 / cache_read 0.015.
-# Vancine public pricing (/api/pricing) is 0.24 / 0.96 / n/a / 0.0048 (USD/MTok).
-base_model = "meituan/longcat-2.0"
-
-[[reasoning_options]]
-type = "toggle"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.24
-output = 0.96
-cache_read = 0.0048
diff --git a/providers/vancine/models/mimo-v2.5-pro.toml b/providers/vancine/models/mimo-v2.5-pro.toml
deleted file mode 100644
index 29b4040f1af..00000000000
--- a/providers/vancine/models/mimo-v2.5-pro.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-# Sources (accessed 2026-08-30):
-# https://platform.xiaomimimo.com/static/docs/api/chat/openai-api.md
-# models/xiaomi/mimo-v2.5-pro.toml
-# providers/xiaomi/models/mimo-v2.5-pro.toml
-# Toggle: thinking.type = "enabled" | "disabled".
-# Interleaved field: reasoning_content.
-# Lab cost: input 0.435 / output 0.87 / cache_read 0.0036.
-# Vancine public pricing (/api/pricing) is 0.348 / 0.696 / n/a / 0.00288 (USD/MTok).
-base_model = "xiaomi/mimo-v2.5-pro"
-
-[[reasoning_options]]
-type = "toggle"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.348
-output = 0.696
-cache_read = 0.00288

From 623617ff2273a5e2d87ce1b1de6a0a10ecc03d7c Mon Sep 17 00:00:00 2001
From: Tao <76082878+K-3-LT@users.noreply.github.com>
Date: Tue, 1 Sep 2026 12:02:09 +0800
Subject: [PATCH 0363/2012] feat: update IteraCompute models (#5700)

---
 .../iteracompute/ornith-1.5-35b-a3b.toml      | 31 +++++++++++++++++++
 .../models/iteracompute/qwen3.8-27b.toml      | 22 ++++++++-----
 2 files changed, 46 insertions(+), 7 deletions(-)
 create mode 100644 providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml

diff --git a/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml b/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml
new file mode 100644
index 00000000000..59d4eb1b5ea
--- /dev/null
+++ b/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml
@@ -0,0 +1,31 @@
+# Sources (accessed 2026-08-28):
+# https://api.iteracompute.com/v1/models (model ID, availability, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B (upstream capabilities)
+#
+# IteraCompute serves the language-only NVFP4 build with a 327,680-token
+# combined window, up to 262,144 input tokens, and up to 65,536 output tokens.
+# Reasoning toggle wire path: reasoning_effort = "none" disables thinking;
+# any accepted non-none effort enables it. Responses stream reasoning through
+# the reasoning_content delta field.
+# Prices are in USD per million tokens.
+base_model = "deepreinforce/ornith-1.5-35b-a3b"
+attachment = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.30
+output = 3.00
+
+[limit]
+context = 327_680
+input = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
diff --git a/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml b/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
index 908d60482e5..8eb4170a5f0 100644
--- a/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
+++ b/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
@@ -1,16 +1,24 @@
-# Sources (accessed 2026-08-24):
+# Sources (accessed 2026-09-01):
 # https://api.iteracompute.com/v1/models (model ID, availability, and input/output pricing)
 # https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
-# Cache-read pricing is the rate supplied for the OpenCode/models.dev integration.
+# https://huggingface.co/gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 (served checkpoint and image input)
 #
 # IteraCompute serves Qwen3.8 27B through its OpenAI-compatible API.
-# Model capabilities, limits, and modalities are inherited from
-# models/alibaba/qwen3.8-27b.toml. Prices are in USD per million tokens.
+# The NVFP4 deployment accepts text and image input and extends the combined
+# serving window to 327,680 tokens with up to 262,144 input and 65,536 output tokens.
+# Prices are in USD per million tokens.
 # Effort: reasoning_effort = none|low|medium|xhigh (xhigh is the default).
 base_model = "alibaba/qwen3.8-27b"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
 
 [cost]
-input = 0.35
-output = 3.00
-cache_read = 0.10
+input = 0.30
+output = 2.50
+
+[limit]
+context = 327_680
+input = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image"]

From 3cdd9930ca2c8cf8b8907976745d2f995efffacb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 23:02:49 -0500
Subject: [PATCH 0364/2012] chore(sync): update Vercel AI Gateway model catalog
 (#5967)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add MiMo UltraSpeed reasoning toggle

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml   | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml

diff --git a/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml b/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml
new file mode 100644
index 00000000000..8311903d432
--- /dev/null
+++ b/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml
@@ -0,0 +1,9 @@
+# Toggle: reasoning.enabled = true|false
+base_model = "xiaomi/mimo-v2.5-pro-ultraspeed"
+name = "MiMo V2.5 Pro UltraSpeed"
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 1.305
+output = 2.61
+cache_read = 0.0108

From 3d96d856464dc88ac585d0fefd0a2663a463f7ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 31 Aug 2026 23:28:36 -0500
Subject: [PATCH 0365/2012] fix(sync): reuse Cloudflare sync credentials
 (#5969)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .github/workflows/sync-models.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/sync-models.yml b/.github/workflows/sync-models.yml
index f741985ae65..ac9f3690bcd 100644
--- a/.github/workflows/sync-models.yml
+++ b/.github/workflows/sync-models.yml
@@ -88,8 +88,8 @@ jobs:
           GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
           GOOGLE_GENERATIVE_AI_API_KEY: ${{ secrets.GOOGLE_GENERATIVE_AI_API_KEY }}
           XAI_API_KEY: ${{ secrets.XAI_API_KEY }}
-          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
-          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }}
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN }}
+          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID }}
           CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID }}
           CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN: ${{ secrets.CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN }}
 

From 7d2e8618b6e5a00d98623ef19d4a14d23b9149a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 04:30:30 +0000
Subject: [PATCH 0366/2012] chore(sync): update OpenRouter model catalog
 (#5972)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From 5ed395a28a5f69261d764df6f182282d04292ee4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 04:30:42 +0000
Subject: [PATCH 0367/2012] chore(sync): update NanoGPT model catalog (#5970)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 8a3ca0a93262a7ee8a85b91a2cbb6f97f20f7787 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 04:30:47 +0000
Subject: [PATCH 0368/2012] chore(sync): update Kilo model catalog (#5971)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index da7301fa993..778f98b65d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_024_000
+context = 1_048_576

From 4278c571d729ae2e4dd6ead67b2a8e33f3cc1a4f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 05:27:07 +0000
Subject: [PATCH 0369/2012] chore(sync): update Eden AI model catalog (#5974)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/flexai/Step-3.7-Flash.toml   | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/edenai/models/flexai/Step-3.7-Flash.toml

diff --git a/providers/edenai/models/flexai/Step-3.7-Flash.toml b/providers/edenai/models/flexai/Step-3.7-Flash.toml
new file mode 100644
index 00000000000..fb725b7c60f
--- /dev/null
+++ b/providers/edenai/models/flexai/Step-3.7-Flash.toml
@@ -0,0 +1,18 @@
+base_model = "stepfun/step-3.7-flash"
+base_model_omit = ["limit.input"]
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.2
+output = 1.15
+
+[limit]
+context = 262_144
+
+[modalities]
+input = ["text", "image"]

From 647f828fafc1c16287aae3c7c798fc0a00b3d0e4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 05:28:09 +0000
Subject: [PATCH 0370/2012] chore(sync): update OpenRouter model catalog
 (#5975)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/meta-llama/llama-4-scout.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/meta-llama/llama-4-scout.toml b/providers/openrouter/models/meta-llama/llama-4-scout.toml
index 9c41cc72a7f..58447cec7dd 100644
--- a/providers/openrouter/models/meta-llama/llama-4-scout.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-scout.toml
@@ -12,13 +12,12 @@ knowledge = "2024-08-31"
 open_weights = true
 
 [cost]
-input = 0.11
-output = 0.34
-cache_read = 0.055
+input = 0.1
+output = 0.3
 
 [limit]
 context = 1_310_720
-output = 8_192
+output = 16_384
 
 [modalities]
 input = ["text", "image"]

From b610c33ab84a8aec5855cdb46f516f309da21c05 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 05:28:17 +0000
Subject: [PATCH 0371/2012] chore(sync): update Kilo model catalog (#5973)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-4-scout.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-4-scout.toml b/providers/kilo/models/meta-llama/llama-4-scout.toml
index 94cda30825c..e70a3817365 100644
--- a/providers/kilo/models/meta-llama/llama-4-scout.toml
+++ b/providers/kilo/models/meta-llama/llama-4-scout.toml
@@ -15,8 +15,8 @@ input = 0.1
 output = 0.3
 
 [limit]
-context = 131_072
-output = 8_192
+context = 327_680
+output = 16_384
 
 [modalities]
 input = ["text", "image"]

From 0bd12aef64fb8bb69a3308ad4eb8b46fa2999191 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 06:42:25 +0000
Subject: [PATCH 0372/2012] chore(sync): update NanoGPT model catalog (#5977)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 7b1ab6ce1a212a1a29f5796865cc9fdf41bf6ac9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 06:43:00 +0000
Subject: [PATCH 0373/2012] chore(sync): update Kilo model catalog (#5976)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 778f98b65d9..da7301fa993 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_048_576
+context = 1_024_000

From 0498d30bdd0f93927c75a081f80baad35c901d57 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 06:43:07 +0000
Subject: [PATCH 0374/2012] chore(sync): update OpenRouter model catalog
 (#5978)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..77c62b9b593 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576

From 168c1302741e13848c12d6f9889f789c408d25f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 07:29:38 +0000
Subject: [PATCH 0375/2012] chore(sync): update Cortecs model catalog (#5979)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/devstral-2512.toml   |  6 ++---
 providers/cortecs/models/minimax-m2.5.toml    |  2 +-
 .../cortecs/models/mistral-medium-2508.toml   | 24 -------------------
 3 files changed, 4 insertions(+), 28 deletions(-)
 delete mode 100644 providers/cortecs/models/mistral-medium-2508.toml

diff --git a/providers/cortecs/models/devstral-2512.toml b/providers/cortecs/models/devstral-2512.toml
index 833006d927b..43fb308c802 100644
--- a/providers/cortecs/models/devstral-2512.toml
+++ b/providers/cortecs/models/devstral-2512.toml
@@ -3,10 +3,10 @@ description = "Mistral coding agent model for repository tasks and software engi
 structured_output = true
 
 [cost]
-input = 0.446
-output = 2.228
+input = 0.478
+output = 2.392
 cache_read = 0.045
 
 [limit]
-context = 262_000
+context = 256_000
 output = 262_000
diff --git a/providers/cortecs/models/minimax-m2.5.toml b/providers/cortecs/models/minimax-m2.5.toml
index 218e96ae663..abf22f19643 100644
--- a/providers/cortecs/models/minimax-m2.5.toml
+++ b/providers/cortecs/models/minimax-m2.5.toml
@@ -12,5 +12,5 @@ output = 1.186
 cache_read = 0.075
 
 [limit]
-context = 196_680
+context = 196_000
 output = 196_608
diff --git a/providers/cortecs/models/mistral-medium-2508.toml b/providers/cortecs/models/mistral-medium-2508.toml
deleted file mode 100644
index e81a6c42e6e..00000000000
--- a/providers/cortecs/models/mistral-medium-2508.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "mistral-medium-2508"
-description = "Mistral Medium 2508 is a frontier-class multimodal LLM with a 128,000 token context window, optimized for reasoning, coding, and multimodal tasks."
-release_date = "2024-08-07"
-last_updated = "2024-08-07"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.446
-output = 2.228
-cache_read = 0.045
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From fcf0187c61d96602671f881eef4923613a6b337d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 07:29:52 +0000
Subject: [PATCH 0376/2012] chore(sync): update OpenRouter model catalog
 (#5981)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index aa706cc488b..0d691163001 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,12 +12,13 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.55
+input = 0.0875
+output = 0.35
+cache_read = 0.0175
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]

From f1b825eec1b9125814b2cf5dea73464edefb96ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 07:30:17 +0000
Subject: [PATCH 0377/2012] chore(sync): update Kilo model catalog (#5980)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nex-agi/nex-n2-mini.toml       | 2 +-
 providers/kilo/models/nex-agi/nex-n2-pro.toml        | 2 +-
 providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/nex-agi/nex-n2-mini.toml b/providers/kilo/models/nex-agi/nex-n2-mini.toml
index 07eb0ca1d56..b4d27604099 100644
--- a/providers/kilo/models/nex-agi/nex-n2-mini.toml
+++ b/providers/kilo/models/nex-agi/nex-n2-mini.toml
@@ -1,4 +1,4 @@
-name = "Nex AGI: Nex-N2-Mini (retires Sep 4)"
+name = "Nex AGI: Nex-N2-Mini (retires Sep 8)"
 description = "Nex-N2-Mini is an open-source agentic mixture-of-experts model from Nex AGI, the smaller sibling in the Nex-N2 series. It accepts text and image input and is built for coding, tool use,..."
 family = "agi"
 release_date = "2026-06-24"
diff --git a/providers/kilo/models/nex-agi/nex-n2-pro.toml b/providers/kilo/models/nex-agi/nex-n2-pro.toml
index c0789522ba5..b15895650f1 100644
--- a/providers/kilo/models/nex-agi/nex-n2-pro.toml
+++ b/providers/kilo/models/nex-agi/nex-n2-pro.toml
@@ -1,4 +1,4 @@
-name = "Nex AGI: Nex-N2-Pro (retires Sep 4)"
+name = "Nex AGI: Nex-N2-Pro (retires Sep 8)"
 description = "Nex-N2-Pro is an agentic mixture-of-experts model from Nex AGI, with 17B active parameters out of 397B total. Built on the Qwen3.5 architecture, it accepts text and image input and produces..."
 family = "agi"
 release_date = "2026-06-08"
diff --git a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
index 3f37ded3629..b356342358d 100644
--- a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
@@ -16,7 +16,7 @@ output = 0.598
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]

From bc3946dcfd1d85674338cef1e9521f6a2a4828b1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 08:35:13 +0000
Subject: [PATCH 0378/2012] chore(sync): update OpenRouter model catalog
 (#5983)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/gryphe/mythomax-l2-13b.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
index de962f15da2..0ddba1dbce0 100644
--- a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
@@ -11,12 +11,12 @@ knowledge = "2023-06-30"
 open_weights = true
 
 [cost]
-input = 0.35
-output = 0.6
+input = 0.06
+output = 0.06
 
 [limit]
 context = 8_192
-output = 7_372
+output = 3_686
 
 [modalities]
 input = ["text"]

From 4f19782ea920b900986df7bf108362d191c75a41 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 08:35:42 +0000
Subject: [PATCH 0379/2012] chore(sync): update Kilo model catalog (#5982)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/gryphe/mythomax-l2-13b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/gryphe/mythomax-l2-13b.toml b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
index e402d2dc4a0..a7751171cf6 100644
--- a/providers/kilo/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
@@ -14,8 +14,8 @@ input = 0.06
 output = 0.06
 
 [limit]
-context = 8_192
-output = 7_372
+context = 4_096
+output = 3_686
 
 [modalities]
 input = ["text"]

From bbfaa87ec149cf850ded4db1eced8c2088c0afca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 10:29:52 +0000
Subject: [PATCH 0380/2012] chore(sync): update OpenRouter model catalog
 (#5990)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 77c62b9b593..fa97cb66a88 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From 298c78f4bba3210ee71b3de0689471281f46e6b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 10:30:10 +0000
Subject: [PATCH 0381/2012] chore(sync): update Chutes model catalog (#5988)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml b/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
index 989e3743ff1..ad2277c0b87 100644
--- a/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
+++ b/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
@@ -6,9 +6,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.35
-output = 2.75
-cache_read = 0.03499999999999999
+input = 0.32
+output = 2.5
+cache_read = 0.031999999999999994
 
 [limit]
 output = 65_536

From 8dc534d3cd9f1e0c1f1b2a409e1d3a6f3dd5e110 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 10:30:25 +0000
Subject: [PATCH 0382/2012] chore(sync): update Kilo model catalog (#5989)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 providers/kilo/models/minimax/minimax-m2.7:free.toml     | 4 +---
 providers/kilo/models/minimax/minimax-m3:free.toml       | 7 +------
 3 files changed, 3 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index da7301fa993..778f98b65d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_024_000
+context = 1_048_576
diff --git a/providers/kilo/models/minimax/minimax-m2.7:free.toml b/providers/kilo/models/minimax/minimax-m2.7:free.toml
index 54ac074faba..a0258eb4533 100644
--- a/providers/kilo/models/minimax/minimax-m2.7:free.toml
+++ b/providers/kilo/models/minimax/minimax-m2.7:free.toml
@@ -10,9 +10,7 @@ values = ["high"]
 [cost]
 input = 0
 output = 0
-reasoning = 0
-cache_read = 0
 
 [limit]
 context = 196_608
-output = 196_608
+output = 176_947
diff --git a/providers/kilo/models/minimax/minimax-m3:free.toml b/providers/kilo/models/minimax/minimax-m3:free.toml
index 1e4add53e23..d922c671141 100644
--- a/providers/kilo/models/minimax/minimax-m3:free.toml
+++ b/providers/kilo/models/minimax/minimax-m3:free.toml
@@ -10,11 +10,6 @@ values = ["none", "high"]
 [cost]
 input = 0
 output = 0
-reasoning = 0
-cache_read = 0
 
 [limit]
-output = 1_048_576
-
-[modalities]
-input = ["text", "image"]
+output = 943_718

From d6ecf46718cb111abdccf51323df3cae63257395 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 10:32:09 +0000
Subject: [PATCH 0383/2012] chore(sync): update NanoGPT model catalog (#5991)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 002f312d1457bfe198d08bb9bff27f946a077539 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 12:36:41 +0000
Subject: [PATCH 0384/2012] chore(sync): update Kilo model catalog (#5995)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/anthropic/claude-opus-4.7-fast.toml | 13 -------------
 1 file changed, 13 deletions(-)
 delete mode 100644 providers/kilo/models/anthropic/claude-opus-4.7-fast.toml

diff --git a/providers/kilo/models/anthropic/claude-opus-4.7-fast.toml b/providers/kilo/models/anthropic/claude-opus-4.7-fast.toml
deleted file mode 100644
index d9e36623fa9..00000000000
--- a/providers/kilo/models/anthropic/claude-opus-4.7-fast.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-opus-4-7"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 30
-output = 150
-cache_read = 3
-cache_write = 37.5

From 9113d8445319de81590b588e89cef78cb2a5ce84 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 12:37:00 +0000
Subject: [PATCH 0385/2012] chore(sync): update OpenRouter model catalog
 (#5996)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../anthropic/claude-opus-4.7-fast.toml       | 19 -------------------
 1 file changed, 19 deletions(-)
 delete mode 100644 providers/openrouter/models/anthropic/claude-opus-4.7-fast.toml

diff --git a/providers/openrouter/models/anthropic/claude-opus-4.7-fast.toml b/providers/openrouter/models/anthropic/claude-opus-4.7-fast.toml
deleted file mode 100644
index 78269e8a76d..00000000000
--- a/providers/openrouter/models/anthropic/claude-opus-4.7-fast.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-# This route maps effort through top-level `verbosity`, not `reasoning.effort`.
-# https://openrouter.ai/docs/api/reference/parameters#verbosity (accessed 2026-06-25)
-base_model = "anthropic/claude-opus-4-7"
-name = "Claude Opus 4.7 (Fast)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 30
-output = 150
-cache_read = 3
-cache_write = 37.5

From ca22f5d0f94b63a414cb304d064f9eaf8bcde238 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 13:27:00 +0000
Subject: [PATCH 0386/2012] chore(sync): update Charm Hyper model catalog
 (#5999)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gpt-oss-120b.toml | 6 +++---
 providers/hyper/models/kimi-k2.5.toml    | 6 +++---
 providers/hyper/models/minimax-m2.7.toml | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index ad08deed2d7..4995ffb98a5 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.188
-output = 0.7
-cache_write = 0.094
+input = 0.19
+output = 0.63
+cache_write = 0.095
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index debc53b807a..ea59490a67b 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5444
-output = 2.855
-cache_write = 0.2722
+input = 0.5504
+output = 2.885
+cache_write = 0.2752
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index de482c56069..de3e51335c2 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.424
-output = 1.612
-cache_write = 0.212
+input = 0.404
+output = 1.496
+cache_write = 0.202
 
 [limit]
 context = 262_100

From 78f8513184cb0f3a2b251f1b9ced9a1f7c920a23 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 13:27:12 +0000
Subject: [PATCH 0387/2012] chore(sync): update Kilo model catalog (#5998)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/anthropic/claude-opus-4.8-fast.toml | 13 -------------
 .../kilo/models/anthropic/claude-opus-5-fast.toml   | 13 -------------
 2 files changed, 26 deletions(-)
 delete mode 100644 providers/kilo/models/anthropic/claude-opus-4.8-fast.toml
 delete mode 100644 providers/kilo/models/anthropic/claude-opus-5-fast.toml

diff --git a/providers/kilo/models/anthropic/claude-opus-4.8-fast.toml b/providers/kilo/models/anthropic/claude-opus-4.8-fast.toml
deleted file mode 100644
index 693ca9822b2..00000000000
--- a/providers/kilo/models/anthropic/claude-opus-4.8-fast.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-opus-4-8"
-description = "Fast-mode variant of [Opus 4.8](/anthropic/claude-opus-4.8) - identical capabilities with higher output speed at 2x pricing relative to regular Opus 4.8. Learn more in Anthropic's docs: https://platform.claude.com/docs/en/build-with-claude/fast-mode"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
diff --git a/providers/kilo/models/anthropic/claude-opus-5-fast.toml b/providers/kilo/models/anthropic/claude-opus-5-fast.toml
deleted file mode 100644
index d6d2efe26b9..00000000000
--- a/providers/kilo/models/anthropic/claude-opus-5-fast.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-opus-5"
-description = "Fast-mode variant of [Opus 5](/anthropic/claude-opus-5) - identical capabilities with higher output speed at 2x pricing relative to regular Opus 5. Learn more in Anthropic's docs: https://platform.claude.com/docs/en/build-with-claude/fast-mode"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5

From 88de45743b3f8fa06f68ac7dc5eda7eccccb2c91 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 13:27:26 +0000
Subject: [PATCH 0388/2012] chore(sync): update OpenRouter model catalog
 (#5997)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../anthropic/claude-opus-4.8-fast.toml       | 19 -------------------
 .../models/anthropic/claude-opus-5-fast.toml  | 17 -----------------
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +++---
 3 files changed, 3 insertions(+), 39 deletions(-)
 delete mode 100644 providers/openrouter/models/anthropic/claude-opus-4.8-fast.toml
 delete mode 100644 providers/openrouter/models/anthropic/claude-opus-5-fast.toml

diff --git a/providers/openrouter/models/anthropic/claude-opus-4.8-fast.toml b/providers/openrouter/models/anthropic/claude-opus-4.8-fast.toml
deleted file mode 100644
index 07ca464df48..00000000000
--- a/providers/openrouter/models/anthropic/claude-opus-4.8-fast.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-# This route maps effort through top-level `verbosity`, not `reasoning.effort`.
-# https://openrouter.ai/docs/api/reference/parameters#verbosity (accessed 2026-06-25)
-base_model = "anthropic/claude-opus-4-8"
-name = "Claude Opus 4.8 (Fast)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
diff --git a/providers/openrouter/models/anthropic/claude-opus-5-fast.toml b/providers/openrouter/models/anthropic/claude-opus-5-fast.toml
deleted file mode 100644
index 6d282fa6b32..00000000000
--- a/providers/openrouter/models/anthropic/claude-opus-5-fast.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "anthropic/claude-opus-5"
-name = "Claude Opus 5 (Fast)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-structured_output = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576

From 27bb91486582ae7bc5c165f1b990bdba416d778e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 14:29:21 +0000
Subject: [PATCH 0389/2012] chore(sync): update Kilo model catalog (#6002)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 80dc12b4ce8..5b7d7be3b0d 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04998
+output = 0.09996
+cache_read = 0.009996
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 131_072
 
 [modalities]
 input = ["text"]

From 5a5fff485204832d21b9e8a62ae7523fde8a79a5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 14:30:12 +0000
Subject: [PATCH 0390/2012] chore(sync): update OpenRouter model catalog
 (#6003)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/inception/mercury-2.5-preview.toml | 28 +++++++++++++++++++
 .../~deepseek/deepseek-v4-flash-latest.toml   |  8 +++---
 3 files changed, 35 insertions(+), 7 deletions(-)
 create mode 100644 providers/openrouter/models/inception/mercury-2.5-preview.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inception/mercury-2.5-preview.toml b/providers/openrouter/models/inception/mercury-2.5-preview.toml
new file mode 100644
index 00000000000..df3e6f4de7c
--- /dev/null
+++ b/providers/openrouter/models/inception/mercury-2.5-preview.toml
@@ -0,0 +1,28 @@
+name = "Mercury 2.5 Preview"
+description = "Preview model for early access evaluation, prototyping, and compatibility testing"
+family = "mercury"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.04
+output = 0.15
+cache_read = 0.004
+
+[limit]
+context = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 27d5c88cb9e..4e3d7984114 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04998
+output = 0.09996
+cache_read = 0.009996
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 131_072
 
 [modalities]
 input = ["text"]

From d3b24adb0a1fa7bef54d727b969201e7fa4f2eba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 15:43:08 +0000
Subject: [PATCH 0391/2012] chore(sync): update NanoGPT model catalog (#6008)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/inception/mercury-2.5-preview.toml | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/nano-gpt/models/inception/mercury-2.5-preview.toml

diff --git a/providers/nano-gpt/models/inception/mercury-2.5-preview.toml b/providers/nano-gpt/models/inception/mercury-2.5-preview.toml
new file mode 100644
index 00000000000..c1b7e926993
--- /dev/null
+++ b/providers/nano-gpt/models/inception/mercury-2.5-preview.toml
@@ -0,0 +1,28 @@
+name = "Mercury 2.5 Preview"
+description = "Mercury 2.5 Preview is Inception's latest and most intelligent diffusion language model. Instead of generating tokens strictly one at a time, it produces and refines multiple tokens in parallel, reaching up to 1,107 tokens per second on standard GPUs. It delivers a 10+ point intelligence gain over Mercury 2, with tunable reasoning, parallel tool calls, schema-aligned JSON output, and a 260K context window. It is built for latency-sensitive production work such as search agents, voice pipelines, customer support, rapid coding iteration, and coding subagents."
+family = "mercury"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.04
+output = 0.15
+cache_read = 0.004
+
+[limit]
+context = 260_000
+input = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 63b3612b76552408915c2057f0e9cd7572438b7d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 15:43:10 +0000
Subject: [PATCH 0392/2012] chore(sync): update Eden AI model catalog (#6006)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 5 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 48c5c557a03..3fc44c890c7 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75374
-output = 0.75374
+input = 0.75335
+output = 0.75335
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 8b46b24c3fd..dd5e7202e80 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17394
-output = 0.75374
+input = 0.17385
+output = 0.75335
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 5359e6b91b0..a787a43195b 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46384
-output = 0.92768
+input = 0.4636
+output = 0.9272
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 315161c8eea..ffbb2b1b144 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17394
-output = 0.69576
+input = 0.17385
+output = 0.6954
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 9f6e0059657..8ef573d4ac8 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04364
-output = 1.04364
+input = 1.0431
+output = 1.0431

From 3165c72faaad16e0cc956fa72039afd3cd0c2205 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 15:43:16 +0000
Subject: [PATCH 0393/2012] chore(sync): update Kilo model catalog (#6005)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/inception/mercury-2.5-preview.toml | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/kilo/models/inception/mercury-2.5-preview.toml

diff --git a/providers/kilo/models/inception/mercury-2.5-preview.toml b/providers/kilo/models/inception/mercury-2.5-preview.toml
new file mode 100644
index 00000000000..3ac644d23b9
--- /dev/null
+++ b/providers/kilo/models/inception/mercury-2.5-preview.toml
@@ -0,0 +1,28 @@
+name = "Inception: Mercury 2.5 Preview"
+description = "Mercury 2.5 is the fastest reasoning LLM, and the latest diffusion LLM (dLLM) from Inception. Instead of generating tokens sequentially, Mercury 2.5 produces and refines multiple tokens in parallel, achieving..."
+family = "mercury"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.2
+output = 0.75
+cache_read = 0.02
+
+[limit]
+context = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 29807acdacb831bc9e16ff5a4928061712b1ce76 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 15:43:45 +0000
Subject: [PATCH 0394/2012] chore(sync): update OpenRouter model catalog
 (#6007)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..f7dc249d032 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.0763
+output = 0.1526
+cache_read = 0.01526
 
 [limit]
 context = 1_048_576

From 3cf10ed4b8558a3d225d2d978dc2c0076c8a33a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 16:45:37 +0000
Subject: [PATCH 0395/2012] chore(sync): update LLM Gateway model catalog
 (#6011)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml  | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
index 49ae721b432..3be5d36f155 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM-5.3 Flash (SCX.ai)"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,6 +13,3 @@ cache_read = 0.024
 
 [limit]
 context = 1_048_576
-
-[modalities]
-input = ["text"]

From 73a6943d1e5ff6f6dc86f938746c5c121d905361 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 16:45:49 +0000
Subject: [PATCH 0396/2012] chore(sync): update OpenRouter model catalog
 (#6010)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f7dc249d032..d6a823fa222 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0763
-output = 0.1526
-cache_read = 0.01526
+input = 0.07476
+output = 0.14952
+cache_read = 0.014952
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 6a2c6c70fb2372557d436c048f94232c9fc81abd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 16:46:16 +0000
Subject: [PATCH 0397/2012] chore(sync): update Kilo model catalog (#6012)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From ddbba206283100e25a3644c1c8ea12bbe70ea637 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 17:24:23 +0000
Subject: [PATCH 0398/2012] chore(sync): update Kilo model catalog (#6016)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 4626ba1427a..f39407dc694 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -1,6 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-vision-exp"
 description = "DeepSeek V4 Flash Vision Exp is an experimental vision-enabled version of [DeepSeek V4 Flash 0731](https://openrouter.ai/deepseek/deepseek-v4-flash-0731) from DeepSeek, adding image understanding while matching the base model on text capabilities including agents,..."
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"

From 8b31d72df248e318ce257d16e8c1effda7ff8604 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 17:25:41 +0000
Subject: [PATCH 0399/2012] chore(sync): update OpenRouter model catalog
 (#6015)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index fa97cb66a88..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -2,7 +2,6 @@
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4-flash-vision-exp"
 description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"

From bf06cd2a3ae2922acd435ddc441ca5874f93ee8a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 13:04:32 -0500
Subject: [PATCH 0400/2012] feat: add Claude Fable 5.1 (#6017)

* feat: add Claude Fable 5.1

* fix: mark Fable 5.1 structured output support

* docs: cite Fable 5.1 Bedrock regional pricing

---------

Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>
---
 models/anthropic/claude-fable-5-1.toml        | 19 +++++++++++++++++++
 .../models/anthropic.claude-fable-5-1.toml    |  8 ++++++++
 .../global.anthropic.claude-fable-5-1.toml    |  9 +++++++++
 .../models/us.anthropic.claude-fable-5-1.toml | 11 +++++++++++
 .../anthropic/models/claude-fable-5-1.toml    |  9 +++++++++
 .../models/claude-fable-5-1.toml              | 13 +++++++++++++
 providers/azure/models/claude-fable-5-1.toml  | 13 +++++++++++++
 .../models/claude-fable-5-1@default.toml      |  9 +++++++++
 .../models/claude-fable-5-1@default.toml      | 12 ++++++++++++
 9 files changed, 103 insertions(+)
 create mode 100644 models/anthropic/claude-fable-5-1.toml
 create mode 100644 providers/amazon-bedrock/models/anthropic.claude-fable-5-1.toml
 create mode 100644 providers/amazon-bedrock/models/global.anthropic.claude-fable-5-1.toml
 create mode 100644 providers/amazon-bedrock/models/us.anthropic.claude-fable-5-1.toml
 create mode 100644 providers/anthropic/models/claude-fable-5-1.toml
 create mode 100644 providers/azure-cognitive-services/models/claude-fable-5-1.toml
 create mode 100644 providers/azure/models/claude-fable-5-1.toml
 create mode 100644 providers/google-vertex-anthropic/models/claude-fable-5-1@default.toml
 create mode 100644 providers/google-vertex/models/claude-fable-5-1@default.toml

diff --git a/models/anthropic/claude-fable-5-1.toml b/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..7bc4479a33f
--- /dev/null
+++ b/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,19 @@
+name = "Claude Fable 5.1"
+description = "Claude model for demanding reasoning and long-horizon agentic work"
+family = "claude-fable"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+open_weights = false
+knowledge = "2026-06"
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-fable-5-1.toml b/providers/amazon-bedrock/models/anthropic.claude-fable-5-1.toml
new file mode 100644
index 00000000000..51bc102b192
--- /dev/null
+++ b/providers/amazon-bedrock/models/anthropic.claude-fable-5-1.toml
@@ -0,0 +1,8 @@
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-fable-5-1.toml b/providers/amazon-bedrock/models/global.anthropic.claude-fable-5-1.toml
new file mode 100644
index 00000000000..597b9728ebc
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-fable-5-1.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Fable 5.1 (Global)"
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-fable-5-1.toml b/providers/amazon-bedrock/models/us.anthropic.claude-fable-5-1.toml
new file mode 100644
index 00000000000..7aef6c516b1
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-fable-5-1.toml
@@ -0,0 +1,11 @@
+# Source: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock#regions
+# Fable 5.1 regional availability is US-only at launch; regional endpoints carry a 10% premium over global.
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Fable 5.1 (US)"
+
+[cost]
+input = 11
+output = 55
+cache_read = 0.275
+cache_write = 13.75
diff --git a/providers/anthropic/models/claude-fable-5-1.toml b/providers/anthropic/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..65d3cdf102d
--- /dev/null
+++ b/providers/anthropic/models/claude-fable-5-1.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/azure-cognitive-services/models/claude-fable-5-1.toml b/providers/azure-cognitive-services/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..90a46203a46
--- /dev/null
+++ b/providers/azure-cognitive-services/models/claude-fable-5-1.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.00
+output = 50.00
+cache_read = 0.25
+cache_write = 12.50
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"
diff --git a/providers/azure/models/claude-fable-5-1.toml b/providers/azure/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..e79229c0802
--- /dev/null
+++ b/providers/azure/models/claude-fable-5-1.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+structured_output = true
+
+[cost]
+input = 10.00
+output = 50.00
+cache_read = 0.25
+cache_write = 12.50
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"
diff --git a/providers/google-vertex-anthropic/models/claude-fable-5-1@default.toml b/providers/google-vertex-anthropic/models/claude-fable-5-1@default.toml
new file mode 100644
index 00000000000..65d3cdf102d
--- /dev/null
+++ b/providers/google-vertex-anthropic/models/claude-fable-5-1@default.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/google-vertex/models/claude-fable-5-1@default.toml b/providers/google-vertex/models/claude-fable-5-1@default.toml
new file mode 100644
index 00000000000..83568ba4aeb
--- /dev/null
+++ b/providers/google-vertex/models/claude-fable-5-1@default.toml
@@ -0,0 +1,12 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/google-vertex/anthropic"

From f20fe9bb94b70a44cbc17e3a5ab790af4be845a5 Mon Sep 17 00:00:00 2001
From: R44VC0RP <89211796+R44VC0RP@users.noreply.github.com>
Date: Tue, 1 Sep 2026 18:03:35 +0000
Subject: [PATCH 0401/2012] feat: add Claude Fable 5.1 to OpenCode

---
 providers/opencode/models/claude-fable-5-1.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/opencode/models/claude-fable-5-1.toml

diff --git a/providers/opencode/models/claude-fable-5-1.toml b/providers/opencode/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..cf4e7afeb8b
--- /dev/null
+++ b/providers/opencode/models/claude-fable-5-1.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-fable-5-1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/anthropic"

From 4b6f6e8f1ca1eedc399b7587e9db7786fd97c5ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 18:32:00 +0000
Subject: [PATCH 0402/2012] chore(sync): update Merge Gateway model catalog
 (#6022)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5-1.toml       | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/merge-gateway/models/anthropic/claude-fable-5-1.toml

diff --git a/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml b/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..d0243bebbee
--- /dev/null
+++ b/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,16 @@
+base_model = "anthropic/claude-fable-5-1"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[modalities]
+input = ["text", "image"]

From c3b969c2358241e9484d7e3463bebd757754c3ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 18:32:20 +0000
Subject: [PATCH 0403/2012] chore(sync): update OpenRouter model catalog
 (#6019)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml    | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/openrouter/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/openrouter/models/anthropic/claude-fable-5.1.toml b/providers/openrouter/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..2138e5ed4ca
--- /dev/null
+++ b/providers/openrouter/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,29 @@
+name = "Claude Fable 5.1"
+description = "Claude model for creative writing, analysis, and controlled agent workflows"
+family = "claude-fable"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From c3dbd955dc2c4b648d9284f41b39a6c288210e44 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 18:32:39 +0000
Subject: [PATCH 0404/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6020)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-vision-exp.toml       | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/providers/vercel/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/vercel/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 39732a870e7..c3974ea5f6a 100644
--- a/providers/vercel/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -1,7 +1,17 @@
 base_model = "deepseek/deepseek-v4-flash-vision-exp"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh"]
 
 [cost]
 input = 0.22
 output = 0.66
 cache_read = 0.007
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 228b5e41a6f68d6660ca200838c72d9e66424564 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 18:34:15 +0000
Subject: [PATCH 0405/2012] chore(sync): update Kilo model catalog (#6021)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml    | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/kilo/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/kilo/models/anthropic/claude-fable-5.1.toml b/providers/kilo/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..34a347b41c2
--- /dev/null
+++ b/providers/kilo/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,29 @@
+name = "Anthropic: Claude Fable 5.1 ($$$$)"
+description = "Claude Fable 5.1 improves on Claude Fable 5 across the board, with the biggest gains in agentic coding, long-running agentic workflows, and knowledge work: long code refactors, front-end and visual..."
+family = "claude-fable"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From 128df6b4d72f926e74efa56fe5d71aacd81effba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:22:55 +0000
Subject: [PATCH 0406/2012] chore(sync): update Eden AI model catalog (#6026)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/google/gemini-3.7-flash.toml | 12 ++++++------
 .../edenai/models/google/gemini-flash-latest.toml    | 12 ++++++------
 providers/edenai/models/vertex/gemini-3.7-flash.toml | 12 ++++++------
 .../edenai/models/vertex/gemini-3.7-flash@eu.toml    | 12 ++++++------
 .../edenai/models/vertex/gemini-3.7-flash@us.toml    | 12 ++++++------
 .../edenai/models/vertex/gemini-flash-latest.toml    | 12 ++++++------
 6 files changed, 36 insertions(+), 36 deletions(-)

diff --git a/providers/edenai/models/google/gemini-3.7-flash.toml b/providers/edenai/models/google/gemini-3.7-flash.toml
index 501847ffb65..1a64f9681ee 100644
--- a/providers/edenai/models/google/gemini-3.7-flash.toml
+++ b/providers/edenai/models/google/gemini-3.7-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/google/gemini-flash-latest.toml b/providers/edenai/models/google/gemini-flash-latest.toml
index 501847ffb65..1a64f9681ee 100644
--- a/providers/edenai/models/google/gemini-flash-latest.toml
+++ b/providers/edenai/models/google/gemini-flash-latest.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash.toml b/providers/edenai/models/vertex/gemini-3.7-flash.toml
index 501847ffb65..1a64f9681ee 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
index b9a37e88446..f582538d697 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash@us.toml b/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
index a9baa56c393..c422d1e56f7 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-flash-latest.toml b/providers/edenai/models/vertex/gemini-flash-latest.toml
index 501847ffb65..1a64f9681ee 100644
--- a/providers/edenai/models/vertex/gemini-flash-latest.toml
+++ b/providers/edenai/models/vertex/gemini-flash-latest.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75

From 2918d42b2c6273d553847f6679e5de1d0b7c792b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:23:05 +0000
Subject: [PATCH 0407/2012] chore(sync): update OpenRouter model catalog
 (#6028)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 .../openrouter/models/~anthropic/claude-fable-latest.toml  | 7 +++++--
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d6a823fa222..0a2a3d73d05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07476
-output = 0.14952
-cache_read = 0.014952
+input = 0.07322
+output = 0.14644
+cache_read = 0.014644
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~anthropic/claude-fable-latest.toml b/providers/openrouter/models/~anthropic/claude-fable-latest.toml
index 5eb3002ec4c..384dddac050 100644
--- a/providers/openrouter/models/~anthropic/claude-fable-latest.toml
+++ b/providers/openrouter/models/~anthropic/claude-fable-latest.toml
@@ -1,6 +1,5 @@
 # This route maps effort through top-level `verbosity`, not `reasoning.effort`.
 # https://openrouter.ai/docs/api/reference/parameters#verbosity (accessed 2026-06-25)
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Fable Latest"
 description = "Claude model for creative writing, analysis, and controlled agent workflows"
 family = "claude-fable"
@@ -13,10 +12,14 @@ tool_call = true
 structured_output = true
 open_weights = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
 [cost]
 input = 10
 output = 50
-cache_read = 1
+cache_read = 0.25
 cache_write = 12.5
 
 [limit]

From 76e8a8f26df10fb5b65d9e79ce7081ba9d27aec0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:23:18 +0000
Subject: [PATCH 0408/2012] chore(sync): update Kilo model catalog (#6025)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~anthropic/claude-fable-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~anthropic/claude-fable-latest.toml b/providers/kilo/models/~anthropic/claude-fable-latest.toml
index 112340248c3..7dcc4eb9bf5 100644
--- a/providers/kilo/models/~anthropic/claude-fable-latest.toml
+++ b/providers/kilo/models/~anthropic/claude-fable-latest.toml
@@ -17,7 +17,7 @@ values = ["low", "medium", "high", "xhigh", "max"]
 [cost]
 input = 10
 output = 50
-cache_read = 1
+cache_read = 0.25
 cache_write = 12.5
 
 [limit]

From 91162712171a23de5ee7bf99986070a12846a9ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:23:51 +0000
Subject: [PATCH 0409/2012] chore(sync): update NanoGPT model catalog (#6029)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml    | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/nano-gpt/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml b/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..d9568b44267
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,29 @@
+name = "Claude Fable 5.1"
+description = "Claude Fable 5.1 improves on Fable 5 across agentic coding, long-running workflows, front-end and visual code generation, finance, analysis, and knowledge work, with more concise plans and summaries. Anthropic retains prompts and outputs for 30 days; Zero Data Retention is not available."
+family = "claude-fable"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From 2eade5bad21199840bd75c905ea583dcd421733d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:24:05 +0000
Subject: [PATCH 0410/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6031)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/claude-fable-5-1.toml | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/llmgateway/models/claude-fable-5-1.toml

diff --git a/providers/llmgateway/models/claude-fable-5-1.toml b/providers/llmgateway/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..dd248407b50
--- /dev/null
+++ b/providers/llmgateway/models/claude-fable-5-1.toml
@@ -0,0 +1,11 @@
+base_model = "anthropic/claude-fable-5-1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5

From a048484b2db81b6d67074283131d000d56dcfc23 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:24:07 +0000
Subject: [PATCH 0411/2012] chore(sync): update LLM Gateway model catalog
 (#6024)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5-1.toml          | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/anthropic/claude-fable-5-1.toml

diff --git a/providers/llmgateway-providers/models/anthropic/claude-fable-5-1.toml b/providers/llmgateway-providers/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..6fe53a1c861
--- /dev/null
+++ b/providers/llmgateway-providers/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-fable-5-1"
+name = "Claude Fable 5.1 (Anthropic)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5

From f0215f3cb336a275bf03912f015702d20b6bf871 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 19:24:46 +0000
Subject: [PATCH 0412/2012] chore(sync): update Venice model catalog (#6030)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/claude-fable-5-1.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/venice/models/claude-fable-5-1.toml

diff --git a/providers/venice/models/claude-fable-5-1.toml b/providers/venice/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..948441cc485
--- /dev/null
+++ b/providers/venice/models/claude-fable-5-1.toml
@@ -0,0 +1,17 @@
+base_model = "anthropic/claude-fable-5-1"
+description = "Claude model for creative writing, analysis, and controlled agent workflows"
+release_date = "2026-08-29"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[modalities]
+input = ["text", "image"]

From 22393f9468839a22214620a254fb2a0036ce1afa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 14:34:40 -0500
Subject: [PATCH 0413/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6023)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add Claude Fable 5.1 reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden <rekram1-node@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml    | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/vercel/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/vercel/models/anthropic/claude-fable-5.1.toml b/providers/vercel/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..4fec091705b
--- /dev/null
+++ b/providers/vercel/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,24 @@
+name = "Claude Fable 5.1"
+description = "Claude model for creative writing, analysis, and controlled agent workflows"
+family = "claude-fable"
+release_date = "2026-08-31"
+last_updated = "2026-08-31"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From ee3b1e5e0440561c30cd2bbbff3b4f6ee400ab36 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 14:34:53 -0500
Subject: [PATCH 0414/2012] chore(sync): update Requesty model catalog (#6027)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gemini-2.5-flash-lite@eu.toml      | 18 +++++++++++++
 .../requesty/models/gemini-2.5-pro@eu.toml    | 25 +++++++++++++++++++
 2 files changed, 43 insertions(+)
 create mode 100644 providers/requesty/models/gemini-2.5-flash-lite@eu.toml
 create mode 100644 providers/requesty/models/gemini-2.5-pro@eu.toml

diff --git a/providers/requesty/models/gemini-2.5-flash-lite@eu.toml b/providers/requesty/models/gemini-2.5-flash-lite@eu.toml
new file mode 100644
index 00000000000..08edc6123b1
--- /dev/null
+++ b/providers/requesty/models/gemini-2.5-flash-lite@eu.toml
@@ -0,0 +1,18 @@
+base_model = "google/gemini-2.5-flash-lite"
+name = "Gemini 2.5 Flash-Lite (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.1
+output = 0.4
+cache_read = 0.01
+cache_write = 0.18333
+
+[limit]
+output = 65_535
diff --git a/providers/requesty/models/gemini-2.5-pro@eu.toml b/providers/requesty/models/gemini-2.5-pro@eu.toml
new file mode 100644
index 00000000000..652b94e934d
--- /dev/null
+++ b/providers/requesty/models/gemini-2.5-pro@eu.toml
@@ -0,0 +1,25 @@
+base_model = "google/gemini-2.5-pro"
+name = "Gemini 2.5 Pro (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 1.25
+output = 10
+cache_read = 0.31
+cache_write = 2.375
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.5
+output = 15
+cache_read = 0.62
+cache_write = 4.75
+
+[limit]
+output = 65_535

From 23bb7d23af1891d541d4b9a8a6002f941b4b77fb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 20:26:10 +0000
Subject: [PATCH 0415/2012] chore(sync): update OpenRouter model catalog
 (#6034)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 7 +++----
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0a2a3d73d05..5e3d330bd1b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07322
-output = 0.14644
-cache_read = 0.014644
+input = 0.07168
+output = 0.14336
+cache_read = 0.014336
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 95460b6f5f3..92d05b69dd5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.6
-output = 3.2
-cache_read = 0.135
+input = 1.02747
+output = 2.05494
+cache_read = 0.085623
 
 [limit]
 context = 1_048_576
-output = 393_216

From 0a4de73725e25672aedea55617303df584a516a0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 20:26:41 +0000
Subject: [PATCH 0416/2012] chore(sync): update Kilo model catalog (#6035)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index e58e24281d4..88500bd7dd8 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,5 +11,4 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000

From 0947af91ee76d99a3d92ed9944f713610aafbb9c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 20:27:12 +0000
Subject: [PATCH 0417/2012] chore(sync): update Merge Gateway model catalog
 (#6033)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../merge-gateway/models/anthropic/claude-fable-5-1.toml    | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml b/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml
index d0243bebbee..6f1a3fda3d9 100644
--- a/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml
+++ b/providers/merge-gateway/models/anthropic/claude-fable-5-1.toml
@@ -1,6 +1,5 @@
 base_model = "anthropic/claude-fable-5-1"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"
@@ -11,6 +10,3 @@ input = 10
 output = 50
 cache_read = 0.25
 cache_write = 12.5
-
-[modalities]
-input = ["text", "image"]

From d23abd0c2cca29d3c0a7eb48320aca1059d14a47 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 20:28:02 +0000
Subject: [PATCH 0418/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6037)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/deepseek/deepseek-v3.toml | 12 ------------
 1 file changed, 12 deletions(-)
 delete mode 100644 providers/vercel/models/deepseek/deepseek-v3.toml

diff --git a/providers/vercel/models/deepseek/deepseek-v3.toml b/providers/vercel/models/deepseek/deepseek-v3.toml
deleted file mode 100644
index 9e57b84026b..00000000000
--- a/providers/vercel/models/deepseek/deepseek-v3.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "deepseek/deepseek-v3"
-name = "DeepSeek V3 0324"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-
-[cost]
-input = 0.27
-output = 1.12
-cache_read = 0.135
-
-[limit]
-context = 163_840
-output = 163_840

From f540c9820fa1930a38a05cb0bed1e8afc9ea6ca4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 21:24:11 +0000
Subject: [PATCH 0419/2012] chore(sync): update Venice model catalog (#6040)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/deepseek-v4-flash.toml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/providers/venice/models/deepseek-v4-flash.toml b/providers/venice/models/deepseek-v4-flash.toml
index f78deda6358..58460c4910d 100644
--- a/providers/venice/models/deepseek-v4-flash.toml
+++ b/providers/venice/models/deepseek-v4-flash.toml
@@ -2,11 +2,14 @@ base_model = "deepseek/deepseek-v4-flash"
 name = "DeepSeek V4 Flash 0423"
 family = "deepseek"
 last_updated = "2026-06-11"
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
 [cost]
 input = 0.138
 output = 0.275

From 59ecd6723f57c2d7b7c8ce6976d56f3c27cc48d4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 21:24:18 +0000
Subject: [PATCH 0420/2012] chore(sync): update OpenRouter model catalog
 (#6041)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92d05b69dd5..e9c8db3462a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02747
-output = 2.05494
-cache_read = 0.085623
+input = 1.024164
+output = 2.048328
+cache_read = 0.085347
 
 [limit]
 context = 1_048_576

From fab929caa2d5b01d9ec3b06122f41647550c7627 Mon Sep 17 00:00:00 2001
From: Mathias Stearn <redbeard0531@gmail.com>
Date: Tue, 1 Sep 2026 23:25:10 +0200
Subject: [PATCH 0421/2012] Update cost for deepseek-v4-flash-0731 on fireworks
 (#6032)

---
 .../accounts/fireworks/models/deepseek-v4-flash-0731.toml   | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
index e9f174ee395..51306765904 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
@@ -11,6 +11,6 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.22
+output = 0.66
+cache_read = 0.007

From cab7e608196a258351a5c73fa4966c274975c312 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 16:35:16 -0500
Subject: [PATCH 0422/2012] chore(sync): update DigitalOcean model catalog
 (#6036)

* chore(sync): update DigitalOcean model catalog

* fix(digitalocean): inherit Fable 5.1 base model

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/anthropic-claude-fable-5.1.toml        | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/digitalocean/models/anthropic-claude-fable-5.1.toml

diff --git a/providers/digitalocean/models/anthropic-claude-fable-5.1.toml b/providers/digitalocean/models/anthropic-claude-fable-5.1.toml
new file mode 100644
index 00000000000..1c259df3835
--- /dev/null
+++ b/providers/digitalocean/models/anthropic-claude-fable-5.1.toml
@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-fable-5-1"
+name = "Anthropic Claude Fable 5.1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[modalities]
+input = ["text", "image"]

From a39fca7ee2e88e536fbd7808c304c2ebb37a310b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 16:35:37 -0500
Subject: [PATCH 0423/2012] fix(sync): resolve dotted Claude versions across
 families (#6042)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/digitalocean.ts   |  4 +-
 .../core/src/sync/providers/openrouter.ts     |  2 +-
 packages/core/test/sync.test.ts               | 37 +++++++++++++++++++
 3 files changed, 40 insertions(+), 3 deletions(-)

diff --git a/packages/core/src/sync/providers/digitalocean.ts b/packages/core/src/sync/providers/digitalocean.ts
index e200fb6af2b..f3942fcddf9 100644
--- a/packages/core/src/sync/providers/digitalocean.ts
+++ b/packages/core/src/sync/providers/digitalocean.ts
@@ -575,13 +575,13 @@ export function resolveDigitalOceanBaseModel(id: string) {
   if (id.startsWith("gemma")) candidates.push(`google/${id}`);
 
   // anthropic-claude-5-sonnet → anthropic/claude-sonnet-5
-  const anthropicSwapped = id.match(/^anthropic-claude-(\d+(?:\.\d+)?)-(opus|sonnet|haiku)$/);
+  const anthropicSwapped = id.match(/^anthropic-claude-(\d+(?:\.\d+)?)-([a-z]+)$/);
   if (anthropicSwapped !== null) {
     candidates.push(`anthropic/claude-${anthropicSwapped[2]}-${anthropicSwapped[1]}`);
   }
   // anthropic-claude-opus-5 → anthropic/claude-opus-5
   // also normalize dotted versions: anthropic-claude-opus-4.6 → anthropic/claude-opus-4-6
-  const anthropicFamily = id.match(/^anthropic-claude-(opus|sonnet|haiku)-(\d+(?:\.\d+)?)$/);
+  const anthropicFamily = id.match(/^anthropic-claude-([a-z]+)-(\d+(?:\.\d+)?)$/);
   if (anthropicFamily !== null) {
     const version = anthropicFamily[2].replaceAll(".", "-");
     candidates.push(`anthropic/claude-${anthropicFamily[1]}-${anthropicFamily[2]}`);
diff --git a/packages/core/src/sync/providers/openrouter.ts b/packages/core/src/sync/providers/openrouter.ts
index 82a9cfc76b5..77164f7caa6 100644
--- a/packages/core/src/sync/providers/openrouter.ts
+++ b/packages/core/src/sync/providers/openrouter.ts
@@ -569,7 +569,7 @@ function canonicalCandidates(provider: string, modelID: string) {
 
   if (provider === "anthropic") {
     for (const candidate of [...candidates]) {
-      candidates.push(candidate.replace(/(claude-(?:opus|sonnet|haiku)-\d+)\.(\d+)/, "$1-$2"));
+      candidates.push(candidate.replace(/(claude-[a-z]+-\d+)\.(\d+)/, "$1-$2"));
       candidates.push(candidate.replace(/^claude-3\.5-/, "claude-3-5-"));
     }
   }
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 987dd262877..d04d7e75817 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -1830,9 +1830,38 @@ test("resolves DigitalOcean IDs to canonical model metadata", () => {
   expect(resolveDigitalOceanBaseModel("mimo-v2.5-pro")).toBe("xiaomi/mimo-v2.5-pro");
   expect(resolveDigitalOceanBaseModel("anthropic-claude-5-sonnet")).toBe("anthropic/claude-sonnet-5");
   expect(resolveDigitalOceanBaseModel("anthropic-claude-opus-5")).toBe("anthropic/claude-opus-5");
+  expect(resolveDigitalOceanBaseModel("anthropic-claude-fable-5.1")).toBe("anthropic/claude-fable-5-1");
+  expect(resolveDigitalOceanBaseModel("anthropic-claude-5.1-fable")).toBe("anthropic/claude-fable-5-1");
+  expect(resolveDigitalOceanBaseModel("anthropic-claude-unknown-99.1")).toBeUndefined();
   expect(resolveDigitalOceanBaseModel("openai-gpt-5.6-luna")).toBe("openai/gpt-5.6-luna");
 });
 
+test("new DigitalOcean Fable models emit only base metadata overrides", () => {
+  const translated = digitalocean.translateModel(
+    digitalOceanModel({
+      id: "anthropic-claude-fable-5.1",
+      name: "Anthropic Claude Fable 5.1",
+      reasoning_efforts: ["low", "medium", "high", "xhigh", "max"],
+      modalities: { input: ["text", "image"], output: ["text"] },
+      max_output_tokens: 128_000,
+      created_at: "2026-09-01T00:00:00Z",
+      pricing: { input: 10, output: 50, cacheRead: 0.25, cacheWrite: 12.5 },
+    }),
+    { existing: () => undefined, authored: () => undefined },
+  );
+
+  expect(translated).toEqual({
+    id: "anthropic-claude-fable-5.1",
+    model: {
+      base_model: "anthropic/claude-fable-5-1",
+      name: "Anthropic Claude Fable 5.1",
+      reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
+      cost: { input: 10, output: 50, cache_read: 0.25, cache_write: 12.5 },
+      modalities: { input: ["text", "image"] },
+    },
+  });
+});
+
 test("new DigitalOcean base models inherit intrinsic capabilities", () => {
   const model = buildDigitalOceanModel(
     digitalOceanModel({
@@ -3005,6 +3034,14 @@ test("factors OpenRouter Pro routes against canonical OpenAI metadata", () => {
   expect("release_date" in model).toBe(false);
 });
 
+test("resolves dotted Claude versions without a family allowlist", () => {
+  expect(resolveCanonicalBaseModel("anthropic/claude-fable-5.1")).toBe("anthropic/claude-fable-5-1");
+  expect(resolveCanonicalBaseModel("anthropic/claude-fable-5.1-fast")).toBe("anthropic/claude-fable-5-1");
+  expect(resolveCanonicalBaseModel("anthropic/claude-opus-4.6")).toBe("anthropic/claude-opus-4-6");
+  expect(resolveCanonicalBaseModel("anthropic/claude-3.5-sonnet-20241022")).toBe("anthropic/claude-3-5-sonnet-20241022");
+  expect(resolveCanonicalBaseModel("anthropic/claude-unknown-99.1")).toBeUndefined();
+});
+
 test("resolves SpaceXAI provider IDs to canonical xAI metadata", () => {
   expect(resolveCanonicalBaseModel("spacexai/grok-4.5")).toBe("xai/grok-4.5");
 });

From 2e18d0a5b8b476284b61e1df72d09fa958713007 Mon Sep 17 00:00:00 2001
From: Ryan Vogel <ryan@mandarin3d.com>
Date: Tue, 1 Sep 2026 17:42:18 -0400
Subject: [PATCH 0424/2012] Revert "feat: add Claude Fable 5.1 to OpenCode"

---
 providers/opencode/models/claude-fable-5-1.toml | 14 --------------
 1 file changed, 14 deletions(-)
 delete mode 100644 providers/opencode/models/claude-fable-5-1.toml

diff --git a/providers/opencode/models/claude-fable-5-1.toml b/providers/opencode/models/claude-fable-5-1.toml
deleted file mode 100644
index cf4e7afeb8b..00000000000
--- a/providers/opencode/models/claude-fable-5-1.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "anthropic/claude-fable-5-1"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 10
-output = 50
-cache_read = 0.25
-cache_write = 12.5
-
-[provider]
-npm = "@ai-sdk/anthropic"

From 21318f6f6810da1d9f57a3f6e7c8927558a8f50f Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Tue, 1 Sep 2026 14:57:00 -0700
Subject: [PATCH 0425/2012] chore(sync): factor vercel claude-fable-5.1 into
 base_model form

Regenerated by bun models:sync vercel now that #6042 resolves dotted
Claude IDs across families. The entry was written standalone before the
canonicalization fix, which left vercel unlinked from the
anthropic/claude-fable-5-1 model page.
---
 .../models/anthropic/claude-fable-5.1.toml    | 22 +++++--------------
 1 file changed, 5 insertions(+), 17 deletions(-)

diff --git a/providers/vercel/models/anthropic/claude-fable-5.1.toml b/providers/vercel/models/anthropic/claude-fable-5.1.toml
index 4fec091705b..b1bc2dcf02f 100644
--- a/providers/vercel/models/anthropic/claude-fable-5.1.toml
+++ b/providers/vercel/models/anthropic/claude-fable-5.1.toml
@@ -1,24 +1,12 @@
-name = "Claude Fable 5.1"
+base_model = "anthropic/claude-fable-5-1"
 description = "Claude model for creative writing, analysis, and controlled agent workflows"
-family = "claude-fable"
-release_date = "2026-08-31"
-last_updated = "2026-08-31"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 10
 output = 50
 cache_read = 0.25
 cache_write = 12.5
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]

From fde0f4fbaf0166f75c0f2fec52a419ccf190aedc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 22:23:51 +0000
Subject: [PATCH 0426/2012] chore(sync): update Venice model catalog (#6048)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/claude-fable-5-1.toml     | 8 ++++----
 providers/venice/models/claude-opus-4-6.toml      | 6 ++++--
 providers/venice/models/claude-opus-4-7.toml      | 6 ++++--
 providers/venice/models/claude-opus-4-8-fast.toml | 6 ++++--
 providers/venice/models/claude-opus-4-8.toml      | 6 ++++--
 providers/venice/models/claude-opus-5-fast.toml   | 5 ++++-
 providers/venice/models/claude-opus-5.toml        | 5 ++++-
 providers/venice/models/claude-sonnet-4-6.toml    | 6 ++++--
 providers/venice/models/claude-sonnet-5.toml      | 5 ++++-
 9 files changed, 36 insertions(+), 17 deletions(-)

diff --git a/providers/venice/models/claude-fable-5-1.toml b/providers/venice/models/claude-fable-5-1.toml
index 948441cc485..b4b1e7008ec 100644
--- a/providers/venice/models/claude-fable-5-1.toml
+++ b/providers/venice/models/claude-fable-5-1.toml
@@ -8,10 +8,10 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 10
-output = 50
-cache_read = 0.25
-cache_write = 12.5
+input = 12
+output = 60
+cache_read = 0.3
+cache_write = 15
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/venice/models/claude-opus-4-6.toml b/providers/venice/models/claude-opus-4-6.toml
index 40d473d4409..fb04a060e4a 100644
--- a/providers/venice/models/claude-opus-4-6.toml
+++ b/providers/venice/models/claude-opus-4-6.toml
@@ -1,10 +1,12 @@
 # Live metadata: reasoning only; no effort control or dedicated budget (2026-06-25).
 # https://api.venice.ai/api/v1/models?type=text
-
 base_model = "anthropic/claude-opus-4-6"
 last_updated = "2026-06-11"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
 
 [cost]
 input = 6
diff --git a/providers/venice/models/claude-opus-4-7.toml b/providers/venice/models/claude-opus-4-7.toml
index d27c7c1f918..1c42f4a0bff 100644
--- a/providers/venice/models/claude-opus-4-7.toml
+++ b/providers/venice/models/claude-opus-4-7.toml
@@ -1,10 +1,12 @@
 # Live metadata: reasoning only; no effort control or dedicated budget (2026-06-25).
 # https://api.venice.ai/api/v1/models?type=text
-
 base_model = "anthropic/claude-opus-4-7"
 last_updated = "2026-06-11"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 6
diff --git a/providers/venice/models/claude-opus-4-8-fast.toml b/providers/venice/models/claude-opus-4-8-fast.toml
index 14dd26ef9e9..c60496926f6 100644
--- a/providers/venice/models/claude-opus-4-8-fast.toml
+++ b/providers/venice/models/claude-opus-4-8-fast.toml
@@ -1,11 +1,13 @@
 # Live metadata: reasoning only; no effort control or dedicated budget (2026-06-25).
 # https://api.venice.ai/api/v1/models?type=text
-
 base_model = "anthropic/claude-opus-4-8"
 name = "Claude Opus 4.8 Fast"
 last_updated = "2026-06-11"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 12
diff --git a/providers/venice/models/claude-opus-4-8.toml b/providers/venice/models/claude-opus-4-8.toml
index 031f143208e..873e3d37aec 100644
--- a/providers/venice/models/claude-opus-4-8.toml
+++ b/providers/venice/models/claude-opus-4-8.toml
@@ -1,10 +1,12 @@
 # Live metadata: reasoning only; no effort control or dedicated budget (2026-06-25).
 # https://api.venice.ai/api/v1/models?type=text
-
 base_model = "anthropic/claude-opus-4-8"
 last_updated = "2026-06-11"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 6
diff --git a/providers/venice/models/claude-opus-5-fast.toml b/providers/venice/models/claude-opus-5-fast.toml
index 242cbb2d53c..43c582febc4 100644
--- a/providers/venice/models/claude-opus-5-fast.toml
+++ b/providers/venice/models/claude-opus-5-fast.toml
@@ -3,7 +3,10 @@ name = "Claude Opus 5 Fast"
 description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
 release_date = "2026-07-23"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 12
diff --git a/providers/venice/models/claude-opus-5.toml b/providers/venice/models/claude-opus-5.toml
index 28bbe41fa76..a231678fb86 100644
--- a/providers/venice/models/claude-opus-5.toml
+++ b/providers/venice/models/claude-opus-5.toml
@@ -2,7 +2,10 @@ base_model = "anthropic/claude-opus-5"
 description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
 release_date = "2026-07-23"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 6
diff --git a/providers/venice/models/claude-sonnet-4-6.toml b/providers/venice/models/claude-sonnet-4-6.toml
index 5aed1c8be87..f4fb860ec66 100644
--- a/providers/venice/models/claude-sonnet-4-6.toml
+++ b/providers/venice/models/claude-sonnet-4-6.toml
@@ -1,10 +1,12 @@
 # Live metadata: reasoning only; no effort control or dedicated budget (2026-06-25).
 # https://api.venice.ai/api/v1/models?type=text
-
 base_model = "anthropic/claude-sonnet-4-6"
 last_updated = "2026-06-11"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
 
 [cost]
 input = 3.6
diff --git a/providers/venice/models/claude-sonnet-5.toml b/providers/venice/models/claude-sonnet-5.toml
index 04e34721ec0..9a367547106 100644
--- a/providers/venice/models/claude-sonnet-5.toml
+++ b/providers/venice/models/claude-sonnet-5.toml
@@ -5,7 +5,10 @@ description = "Balanced Claude model for coding, analysis, agent workflows, and
 release_date = "2026-06-29"
 last_updated = "2026-07-01"
 structured_output = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 3

From 4d6fdb2b0a45bceb102d351e3db83a968d437ebd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 22:23:55 +0000
Subject: [PATCH 0427/2012] chore(sync): update NanoGPT model catalog (#6050)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml       | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)

diff --git a/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml b/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml
index d9568b44267..a9045c8b330 100644
--- a/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml
+++ b/providers/nano-gpt/models/anthropic/claude-fable-5.1.toml
@@ -1,13 +1,5 @@
-name = "Claude Fable 5.1"
-description = "Claude Fable 5.1 improves on Fable 5 across agentic coding, long-running workflows, front-end and visual code generation, finance, analysis, and knowledge work, with more concise plans and summaries. Anthropic retains prompts and outputs for 30 days; Zero Data Retention is not available."
-family = "claude-fable"
-release_date = "2026-09-01"
-last_updated = "2026-09-01"
-attachment = true
-reasoning = true
-tool_call = true
+base_model = "anthropic/claude-fable-5-1"
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,10 +12,4 @@ cache_read = 0.25
 cache_write = 12.5
 
 [limit]
-context = 1_000_000
 input = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]

From 2b09cf90125f9943a1610c13b6555b591fa8495f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 22:24:27 +0000
Subject: [PATCH 0428/2012] chore(sync): update OpenRouter model catalog
 (#6049)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml     | 18 +-----------------
 .../models/deepseek/deepseek-v4-pro.toml       |  6 +++---
 2 files changed, 4 insertions(+), 20 deletions(-)

diff --git a/providers/openrouter/models/anthropic/claude-fable-5.1.toml b/providers/openrouter/models/anthropic/claude-fable-5.1.toml
index 2138e5ed4ca..9f6a8f9989c 100644
--- a/providers/openrouter/models/anthropic/claude-fable-5.1.toml
+++ b/providers/openrouter/models/anthropic/claude-fable-5.1.toml
@@ -1,14 +1,6 @@
-name = "Claude Fable 5.1"
+base_model = "anthropic/claude-fable-5-1"
 description = "Claude model for creative writing, analysis, and controlled agent workflows"
-family = "claude-fable"
-release_date = "2026-09-01"
-last_updated = "2026-09-01"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,11 +11,3 @@ input = 10
 output = 50
 cache_read = 0.25
 cache_write = 12.5
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e9c8db3462a..92d05b69dd5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.024164
-output = 2.048328
-cache_read = 0.085347
+input = 1.02747
+output = 2.05494
+cache_read = 0.085623
 
 [limit]
 context = 1_048_576

From 664c900357c6ee0d9d19462bd79c291215114fa8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 22:25:21 +0000
Subject: [PATCH 0429/2012] chore(sync): update Kilo model catalog (#6047)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml     | 18 +-----------------
 1 file changed, 1 insertion(+), 17 deletions(-)

diff --git a/providers/kilo/models/anthropic/claude-fable-5.1.toml b/providers/kilo/models/anthropic/claude-fable-5.1.toml
index 34a347b41c2..225f81f488f 100644
--- a/providers/kilo/models/anthropic/claude-fable-5.1.toml
+++ b/providers/kilo/models/anthropic/claude-fable-5.1.toml
@@ -1,14 +1,6 @@
-name = "Anthropic: Claude Fable 5.1 ($$$$)"
+base_model = "anthropic/claude-fable-5-1"
 description = "Claude Fable 5.1 improves on Claude Fable 5 across the board, with the biggest gains in agentic coding, long-running agentic workflows, and knowledge work: long code refactors, front-end and visual..."
-family = "claude-fable"
-release_date = "2026-09-01"
-last_updated = "2026-09-01"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,11 +11,3 @@ input = 10
 output = 50
 cache_read = 0.25
 cache_write = 12.5
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]

From 07b82b8ce25e971ded366bc02f0c580d2a5de013 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 23:21:53 +0000
Subject: [PATCH 0430/2012] chore(sync): update Kilo model catalog (#6057)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index c79bdfce589..ad993c8f794 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,7 +13,3 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
-
-[limit]
-context = 1_000_000
-output = 262_144

From c827ca0c21c1dc3620535dfa197598b1447cc1f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 23:21:58 +0000
Subject: [PATCH 0431/2012] chore(sync): update Merge Gateway model catalog
 (#6058)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..faa70b7c1e0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -11,8 +12,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 145ebefe807f2bf20b90cebda95338d250fd351c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 23:22:11 +0000
Subject: [PATCH 0432/2012] chore(sync): update OpenRouter model catalog
 (#6055)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml     | 3 +--
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5e3d330bd1b..27d552efed8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07168
-output = 0.14336
-cache_read = 0.014336
+input = 0.07014
+output = 0.14028
+cache_read = 0.014028
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92d05b69dd5..e9c8db3462a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02747
-output = 2.05494
-cache_read = 0.085623
+input = 1.024164
+output = 2.048328
+cache_read = 0.085347
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index abacdd68c88..51225871db8 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,8 +14,7 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.25
+cache_read = 0.2
 
 [limit]
 context = 1_048_576
-output = 262_144

From f7af17dfaf28e3e2fb7dbf6ca72ba73e8436d59e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 23:22:17 +0000
Subject: [PATCH 0433/2012] chore(sync): update Eden AI model catalog (#6056)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/anthropic/claude-fable-5-1.toml    | 12 ++++++++++++
 .../edenai/models/anthropic/claude-fable-latest.toml |  4 ++--
 .../databricks-deepseek-v4-flash-0731.toml           | 12 ++++++++++++
 .../databricks/databricks-deepseek-v4-pro-0813.toml  | 12 ++++++++++++
 4 files changed, 38 insertions(+), 2 deletions(-)
 create mode 100644 providers/edenai/models/anthropic/claude-fable-5-1.toml
 create mode 100644 providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
 create mode 100644 providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml

diff --git a/providers/edenai/models/anthropic/claude-fable-5-1.toml b/providers/edenai/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..6f1a3fda3d9
--- /dev/null
+++ b/providers/edenai/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,12 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/edenai/models/anthropic/claude-fable-latest.toml b/providers/edenai/models/anthropic/claude-fable-latest.toml
index 60a0268f1dc..6f1a3fda3d9 100644
--- a/providers/edenai/models/anthropic/claude-fable-latest.toml
+++ b/providers/edenai/models/anthropic/claude-fable-latest.toml
@@ -1,4 +1,4 @@
-base_model = "anthropic/claude-fable-5"
+base_model = "anthropic/claude-fable-5-1"
 structured_output = true
 
 [[reasoning_options]]
@@ -8,5 +8,5 @@ values = ["low", "medium", "high", "xhigh", "max"]
 [cost]
 input = 10
 output = 50
-cache_read = 1
+cache_read = 0.25
 cache_write = 12.5
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..baee0759016
--- /dev/null
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
@@ -0,0 +1,12 @@
+base_model = "deepseek/deepseek-v4-flash-0731"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.028
+cache_write = 0.14
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..b5283d22349
--- /dev/null
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
@@ -0,0 +1,12 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 1.31999
+output = 3.95997
+cache_read = 0.13202
+cache_write = 1.31999

From d463cb0b700e47ea99629b54e9cf8caca15845ff Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 01:00:06 +0000
Subject: [PATCH 0434/2012] chore(sync): update Kilo model catalog (#6061)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat-v3.1.toml    | 4 ++--
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 2 +-
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml    | 4 ++--
 providers/kilo/models/tencent/hy3.toml                    | 6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                   | 1 -
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 6 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
index 7717353dce8..4269d8ff1fa 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
@@ -20,8 +20,8 @@ output = 1
 cache_read = 0.135
 
 [limit]
-context = 161_000
-output = 144_900
+context = 163_840
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 778f98b65d9..da7301fa993 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.132
 
 [limit]
-context = 1_048_576
+context = 1_024_000
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index c977cf138f9..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 262_144
-output = 16_384
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..18843e6fee9 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bfdca8a62a6..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -12,4 +12,3 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 5b7d7be3b0d..80dc12b4ce8 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04998
-output = 0.09996
-cache_read = 0.009996
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 393_216
 
 [modalities]
 input = ["text"]

From aacba0778d5c1fd07bdee4eff5b37083dd1d8d3b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 01:00:21 +0000
Subject: [PATCH 0435/2012] chore(sync): update OpenRouter model catalog
 (#6060)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-chat-v3.1.toml | 10 ++++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml  |  6 +++---
 .../models/deepseek/deepseek-v4-pro-0813.toml          |  6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml    |  6 +++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml      |  8 ++++----
 providers/openrouter/models/tencent/hy3.toml           |  6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml          |  9 +++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  8 ++++----
 8 files changed, 31 insertions(+), 28 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
index 139b9db9f6e..24302ed5e58 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "DeepSeek V3.1"
 description = "DeepSeek chat model for instruction following, coding, and analysis"
 family = "deepseek"
@@ -15,13 +17,13 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 1.65
-cache_read = 0.55
+input = 0.25
+output = 0.95
+cache_read = 0.13
 
 [limit]
 context = 163_840
-output = 144_900
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 27d552efed8..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07014
-output = 0.14028
-cache_read = 0.014028
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e9c8db3462a..66443ade4c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.024164
-output = 2.048328
-cache_read = 0.085347
+input = 1.04226
+output = 2.08452
+cache_read = 0.086855
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index de4e34c6528..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.5
-output = 2.2
-cache_read = 0.1
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 16_384
+output = 32_768
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..4febc56bf89 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 6991b6090c0..64652242a0e 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "zhipuai/glm-5.2"
 
 [interleaved]
@@ -11,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.19
-output = 3.74
-cache_read = 0.221
+input = 0.966
+output = 3.036
+cache_read = 0.1932
 
 [limit]
 context = 1_048_576
-output = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 4e3d7984114..27d5c88cb9e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04998
-output = 0.09996
-cache_read = 0.009996
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 393_216
 
 [modalities]
 input = ["text"]

From 480df9ca32c3412d427e904c9f400adee45a525f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 01:35:04 +0000
Subject: [PATCH 0436/2012] chore(sync): update OpenRouter model catalog
 (#6063)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 88db6797e7e6234c3d32b45dc4dfd88f9d7bfd00 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 01:35:09 +0000
Subject: [PATCH 0437/2012] chore(sync): update NanoGPT model catalog (#6064)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 92c4ef57fcb7477e292dcff3e38649703840358d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 02:50:16 +0000
Subject: [PATCH 0438/2012] chore(sync): update Kilo model catalog (#6065)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~z-ai/glm-flash-latest.toml   | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/kilo/models/~z-ai/glm-flash-latest.toml

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
new file mode 100644
index 00000000000..e4ce4e5dfe6
--- /dev/null
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -0,0 +1,28 @@
+name = "Z.ai: GLM Flash Latest"
+description = "This model always redirects to the latest model in the GLM Flash family."
+family = "glm-flash"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

From ebac698493beab910896f1934b04dc531c09d821 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 02:50:29 +0000
Subject: [PATCH 0439/2012] chore(sync): update OpenRouter model catalog
 (#6066)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~z-ai/glm-flash-latest.toml        | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/openrouter/models/~z-ai/glm-flash-latest.toml

diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
new file mode 100644
index 00000000000..c32fca67d4b
--- /dev/null
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -0,0 +1,28 @@
+name = "GLM Flash Latest"
+description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+family = "glm-flash"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+
+[limit]
+context = 1_310_720
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

From 50f3a4e210fd293b8ac419343e7a9f006fdfdc39 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 03:31:08 +0000
Subject: [PATCH 0440/2012] chore(sync): update CrossModel model catalog
 (#6067)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5-1.toml         | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/crossmodel/models/anthropic/claude-fable-5-1.toml

diff --git a/providers/crossmodel/models/anthropic/claude-fable-5-1.toml b/providers/crossmodel/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..ca327804184
--- /dev/null
+++ b/providers/crossmodel/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-fable-5-1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[modalities]
+input = ["text", "image"]

From 5f52e6de18d254dbd5405ff5f3a825f157d7dcba Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Wed, 2 Sep 2026 12:05:49 +0800
Subject: [PATCH 0441/2012] feat(opencode): add Claude Fable 5.1 to Zen

---
 providers/opencode/models/claude-fable-5-1.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/opencode/models/claude-fable-5-1.toml

diff --git a/providers/opencode/models/claude-fable-5-1.toml b/providers/opencode/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..393e9120010
--- /dev/null
+++ b/providers/opencode/models/claude-fable-5-1.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-fable-5-1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file

From f954bdc602359dce121ddf1293d582374ebb7a57 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 04:32:14 +0000
Subject: [PATCH 0442/2012] chore(sync): update NanoGPT model catalog (#6071)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 3640ee80179e301173ad1aba271ef9a7c7a111c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 04:32:30 +0000
Subject: [PATCH 0443/2012] chore(sync): update LLM Gateway model catalog
 (#6068)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/gpt-oss-20b.toml     | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml b/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml
new file mode 100644
index 00000000000..09259fddd1f
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Consensus Protocol)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.04
+output = 0.19
+cache_read = 0.01
+
+[limit]
+context = 65_536

From 819f921e0f6921a98ac4e88abeb2242ee7a916a9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 04:32:44 +0000
Subject: [PATCH 0444/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6069)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gpt-oss-20b.toml | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway/models/gpt-oss-20b.toml b/providers/llmgateway/models/gpt-oss-20b.toml
index 19f47947bea..b7834e1b4d8 100644
--- a/providers/llmgateway/models/gpt-oss-20b.toml
+++ b/providers/llmgateway/models/gpt-oss-20b.toml
@@ -15,8 +15,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.05
-output = 0.2
+input = 0.04
+output = 0.19
+cache_read = 0.01
 
 [limit]
 context = 131_072

From 20dedb1758f8f383549f5f279ebc84dbb914d7d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 04:32:51 +0000
Subject: [PATCH 0445/2012] chore(sync): update OpenRouter model catalog
 (#6072)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From a2d1435882c6f9233ad532c05b2b3143d79158e2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 06:45:00 +0000
Subject: [PATCH 0446/2012] chore(sync): update Cortecs model catalog (#6074)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/glm-5.3.toml | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/providers/cortecs/models/glm-5.3.toml b/providers/cortecs/models/glm-5.3.toml
index 39cd4d9b019..d2aacc59483 100644
--- a/providers/cortecs/models/glm-5.3.toml
+++ b/providers/cortecs/models/glm-5.3.toml
@@ -3,12 +3,15 @@
 # https://docs.z.ai/guides/llm/glm-5.3
 # https://docs.cortecs.ai/examples/reasoning
 base_model = "zhipuai/glm-5.3"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
-input = 1.75
-output = 4.499
-cache_read = 0.438
+input = 1.4
+output = 4.399
+cache_read = 0.26
 
 [limit]
 context = 1_048_576

From edbe5e6b2516d1f516f6bf2d0eb8b238a9d6bffb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 06:45:09 +0000
Subject: [PATCH 0447/2012] chore(sync): update NanoGPT model catalog (#6076)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 203a1fd705e5d94c082df83173f634ac4e81f419 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 06:45:12 +0000
Subject: [PATCH 0448/2012] chore(sync): update OpenRouter model catalog
 (#6075)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From effc8f9c5a8e56aadf0d2de1e0a2556960e0af7c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 07:30:33 +0000
Subject: [PATCH 0449/2012] chore(sync): update NanoGPT model catalog (#6078)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-max-0902.toml      | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml

diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..d1e3bd8866a
--- /dev/null
+++ b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
@@ -0,0 +1,29 @@
+name = "Qwen3.8 Max 0902"
+description = "Qwen3.8 Max 0902 is Alibaba's September 2 checkpoint of its flagship Qwen3.8 Max model for coding, knowledge work, data analysis, and long-running agent workflows. It supports text, image, video, PDF input, selectable thinking, tool calling, structured output, and a near-million-token context window."
+family = "qwen3.8-max"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.17
+cache_write = 2.5
+
+[limit]
+context = 991_808
+input = 991_808
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]

From fb19b8fa5ae88008dd373b933c8b3252ef2c36ae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 08:34:38 +0000
Subject: [PATCH 0450/2012] chore(sync): update Requesty model catalog (#6079)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/glm-5.3-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
index f918675673a..12d3bbf1031 100644
--- a/providers/requesty/models/glm-5.3-flash.toml
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -13,4 +13,4 @@ output = 0.25
 cache_read = 0.015
 
 [limit]
-context = 131_072
+context = 1_048_576

From a5fb4745af44cd30a1dfc141b9dda24db1126d5f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 09:29:44 +0000
Subject: [PATCH 0451/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6083)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway/models/gemini-3.8-flash.toml   | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/llmgateway/models/gemini-3.8-flash.toml

diff --git a/providers/llmgateway/models/gemini-3.8-flash.toml b/providers/llmgateway/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..8eee5addfa4
--- /dev/null
+++ b/providers/llmgateway/models/gemini-3.8-flash.toml
@@ -0,0 +1,29 @@
+name = "Gemini 3.8 Flash"
+description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
+family = "gemini"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.08333
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image", "audio"]
+output = ["text"]

From 081b53c8ad6262ab1b7b695b9da14cd80aaa0697 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 09:30:05 +0000
Subject: [PATCH 0452/2012] chore(sync): update LLM Gateway model catalog
 (#6082)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../google-vertex/gemini-3.8-flash.toml       | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml

diff --git a/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml b/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..85d6071eb17
--- /dev/null
+++ b/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml
@@ -0,0 +1,29 @@
+name = "Gemini 3.8 Flash (Google Vertex AI)"
+description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
+family = "gemini"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.08333
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio"]
+output = ["text"]

From b71c16b42e2bf4ed36f2e1ce4e33cc7fcd3adf8e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 09:30:23 +0000
Subject: [PATCH 0453/2012] chore(sync): update NanoGPT model catalog (#6080)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ornith-ai/ornith-1.5-9b.toml       | 25 -------------------
 .../ornith-ai/ornith-1.5-9b:thinking.toml     | 25 -------------------
 .../qwen/qwen3.6-35b-a3b-uncensored.toml      | 25 -------------------
 .../qwen3.6-35b-a3b-uncensored:thinking.toml  | 25 -------------------
 .../models/z-ai/glm-5.3-flash-uncensored.toml |  4 +--
 5 files changed, 2 insertions(+), 102 deletions(-)
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml

diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
deleted file mode 100644
index 43cf8a042d3..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 9B"
-description = "Ornith 1.5 9B is an FP8 dense open-weight reasoning model built for agentic coding, tool use, visual understanding, and efficient long-context work."
-family = "ornith"
-release_date = "2026-07-29"
-last_updated = "2026-08-23"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
deleted file mode 100644
index 64843cc0f5d..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 9B Thinking"
-description = "Ornith 1.5 9B with thinking enabled for more deliberate agentic coding, tool use, visual understanding, and long-context work."
-family = "ornith"
-release_date = "2026-07-29"
-last_updated = "2026-08-24"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
deleted file mode 100644
index b9e509e8e09..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen 3.6 35B A3B Uncensored"
-description = "Qwen 3.6 35B A3B Uncensored is an FP8 open-weight mixture-of-experts model tuned for fewer refusals across chat, coding, tool use, and multimodal tasks."
-family = "qwen3.6"
-release_date = "2026-07-29"
-last_updated = "2026-08-21"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.95
-cache_read = 0.075
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
deleted file mode 100644
index de3e48de51a..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen 3.6 35B A3B Uncensored Thinking"
-description = "Qwen 3.6 35B A3B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and complex chat tasks."
-family = "qwen3.6"
-release_date = "2026-07-29"
-last_updated = "2026-08-24"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.95
-cache_read = 0.075
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 753a724d11b..3b7212170b1 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -19,8 +19,8 @@ output = 1.4
 cache_read = 0.175
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]

From 42303715630a4a6fc9d454966f8ab1871b714dc3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 10:28:22 +0000
Subject: [PATCH 0454/2012] chore(sync): update Kilo model catalog (#6085)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/meituan/longcat-2.0-free.toml | 29 -------------------
 1 file changed, 29 deletions(-)
 delete mode 100644 providers/kilo/models/meituan/longcat-2.0-free.toml

diff --git a/providers/kilo/models/meituan/longcat-2.0-free.toml b/providers/kilo/models/meituan/longcat-2.0-free.toml
deleted file mode 100644
index 582e354a9eb..00000000000
--- a/providers/kilo/models/meituan/longcat-2.0-free.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-name = "Meituan: LongCat 2.0 (free)"
-description = "LongCat 2.0 is a sparse mixture-of-experts language model from Meituan, with 48B active parameters out of 1.6T total. It is suited for coding, repository-level changes, long-horizon problem solving, and agentic workflows. Available free in Kilo for a limited time."
-family = "longcat"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0
-output = 0
-reasoning = 0
-cache_read = 0
-
-[limit]
-context = 1_048_756
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 882e5e20dece3ccd6f2a90057f4793676dc1cc59 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 10:28:35 +0000
Subject: [PATCH 0455/2012] chore(sync): update OpenRouter model catalog
 (#6086)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From a0db99d4c7b1232cb818dc2768dfbff9aaf80875 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 10:28:46 +0000
Subject: [PATCH 0456/2012] chore(sync): update Eden AI model catalog (#6084)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/databricks/databricks-gpt-oss-120b.toml | 2 +-
 .../edenai/models/databricks/databricks-gpt-oss-120b@eu.toml    | 2 +-
 providers/edenai/models/databricks/databricks-gpt-oss-20b.toml  | 2 +-
 .../edenai/models/databricks/databricks-gpt-oss-20b@eu.toml     | 2 +-
 4 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml b/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
index 2757b223875..ba07e71a376 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
@@ -8,5 +8,5 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.15001
 output = 0.59997
-cache_read = 0.15001
+cache_read = 0.015001
 cache_write = 0.15001
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
index 731a49d8387..37a595e6865 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
@@ -9,5 +9,5 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.15001
 output = 0.59997
-cache_read = 0.15001
+cache_read = 0.015001
 cache_write = 0.15001
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml b/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
index e2b022fe2c3..4e3ee6c0c1b 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
@@ -8,5 +8,5 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.07
 output = 0.30002
-cache_read = 0.07
+cache_read = 0.007
 cache_write = 0.07
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
index a12ca346593..6cee3a77f8b 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
@@ -9,5 +9,5 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.07
 output = 0.30002
-cache_read = 0.07
+cache_read = 0.007
 cache_write = 0.07

From e816bd96babb7f71076acc2efadcf15306a7415f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 10:28:48 +0000
Subject: [PATCH 0457/2012] chore(sync): update NanoGPT model catalog (#6087)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml      | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 3b7212170b1..2d0019c9056 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -19,8 +19,8 @@ output = 1.4
 cache_read = 0.175
 
 [limit]
-context = 524_288
-input = 524_288
+context = 1_048_576
+input = 1_048_576
 output = 32_768
 
 [modalities]

From 2b19bb01d98b2eda4113f48e975a3e454c213c17 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 11:26:01 +0000
Subject: [PATCH 0458/2012] chore(sync): update NanoGPT model catalog (#6092)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/Baichuan4-Air.toml  | 24 -------------------
 .../nano-gpt/models/Baichuan4-Turbo.toml      | 24 -------------------
 .../models/Tongyi-Zhiwen/QwenLong-L1-32B.toml | 24 -------------------
 .../deepseek/deepseek-prover-v2-671b.toml     | 24 -------------------
 .../models/kwaipilot/kat-coder-air-v2.5.toml  | 24 -------------------
 .../models/tencent/Hunyuan-MT-7B.toml         | 24 -------------------
 .../models/z-ai/glm-4.6v-flash-original.toml  | 24 -------------------
 7 files changed, 168 deletions(-)
 delete mode 100644 providers/nano-gpt/models/Baichuan4-Air.toml
 delete mode 100644 providers/nano-gpt/models/Baichuan4-Turbo.toml
 delete mode 100644 providers/nano-gpt/models/Tongyi-Zhiwen/QwenLong-L1-32B.toml
 delete mode 100644 providers/nano-gpt/models/deepseek/deepseek-prover-v2-671b.toml
 delete mode 100644 providers/nano-gpt/models/kwaipilot/kat-coder-air-v2.5.toml
 delete mode 100644 providers/nano-gpt/models/tencent/Hunyuan-MT-7B.toml
 delete mode 100644 providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml

diff --git a/providers/nano-gpt/models/Baichuan4-Air.toml b/providers/nano-gpt/models/Baichuan4-Air.toml
deleted file mode 100644
index 265ce6975db..00000000000
--- a/providers/nano-gpt/models/Baichuan4-Air.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Baichuan 4 Air"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "baichuan"
-release_date = "2025-08-19"
-last_updated = "2025-08-19"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.157
-output = 0.157
-cache_read = 0.0785
-
-[limit]
-context = 32_768
-input = 32_768
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/Baichuan4-Turbo.toml b/providers/nano-gpt/models/Baichuan4-Turbo.toml
deleted file mode 100644
index e07eb628d42..00000000000
--- a/providers/nano-gpt/models/Baichuan4-Turbo.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Baichuan 4 Turbo"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "baichuan"
-release_date = "2025-08-19"
-last_updated = "2025-08-19"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.42
-output = 2.42
-cache_read = 1.21
-
-[limit]
-context = 128_000
-input = 128_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/Tongyi-Zhiwen/QwenLong-L1-32B.toml b/providers/nano-gpt/models/Tongyi-Zhiwen/QwenLong-L1-32B.toml
deleted file mode 100644
index 04cc80f3aee..00000000000
--- a/providers/nano-gpt/models/Tongyi-Zhiwen/QwenLong-L1-32B.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "QwenLong L1 32B"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2024-01-01"
-last_updated = "2025-01-25"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.14
-output = 0.6
-cache_read = 0.07
-
-[limit]
-context = 128_000
-input = 128_000
-output = 40_960
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-prover-v2-671b.toml b/providers/nano-gpt/models/deepseek/deepseek-prover-v2-671b.toml
deleted file mode 100644
index 2d005be6c62..00000000000
--- a/providers/nano-gpt/models/deepseek/deepseek-prover-v2-671b.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "DeepSeek Prover v2 671B"
-description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
-family = "deepseek"
-release_date = "2025-11-15"
-last_updated = "2025-04-30"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 1
-output = 2.5
-cache_read = 0.5
-
-[limit]
-context = 160_000
-input = 160_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/kwaipilot/kat-coder-air-v2.5.toml b/providers/nano-gpt/models/kwaipilot/kat-coder-air-v2.5.toml
deleted file mode 100644
index 2bbd5fcc96d..00000000000
--- a/providers/nano-gpt/models/kwaipilot/kat-coder-air-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "KAT Coder Air V2.5"
-description = "Fast, cost-efficient KAT Coder model for code generation, editing, debugging, and agentic software-development workflows."
-family = "kat-coder"
-release_date = "2026-07-14"
-last_updated = "2026-07-14"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.03
-
-[limit]
-context = 256_000
-input = 256_000
-output = 80_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/tencent/Hunyuan-MT-7B.toml b/providers/nano-gpt/models/tencent/Hunyuan-MT-7B.toml
deleted file mode 100644
index 611660049bb..00000000000
--- a/providers/nano-gpt/models/tencent/Hunyuan-MT-7B.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Hunyuan MT 7B"
-description = "Translation model for multilingual conversion, localization, and cross-language workflows"
-family = "hunyuan"
-release_date = "2025-08-15"
-last_updated = "2025-09-18"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 10
-output = 20
-cache_read = 5
-
-[limit]
-context = 8_192
-input = 8_192
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml b/providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml
deleted file mode 100644
index cbe86397742..00000000000
--- a/providers/nano-gpt/models/z-ai/glm-4.6v-flash-original.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM 4.6V Flash"
-description = "GLM-4.6V-Flash (9B), a lightweight model optimized for local deployment and low-latency applications. Scales context window to 128k tokens and achieves SoTA performance in visual understanding among similar-scale models."
-family = "glm"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.1
-output = 0.4
-cache_read = 0.05
-
-[limit]
-context = 128_000
-input = 128_000
-output = 24_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 4618e533e2caca4edda1c644a59f4ff966e3cc9f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 11:26:14 +0000
Subject: [PATCH 0459/2012] chore(sync): update OpenRouter model catalog
 (#6091)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index d6026834e83..51793fd9815 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.17
-output = 3.96
-cache_read = 0.234
+input = 1.15
+output = 3.5
+cache_read = 0.1
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From 03c2631946bb7ce2735e8e37d04197c4b910ff66 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 11:26:26 +0000
Subject: [PATCH 0460/2012] chore(sync): update Kilo model catalog (#6090)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index a2f41897231..82fdc44362b 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.17
-output = 3.96
-cache_read = 0.234
+input = 1.15
+output = 3.5
+cache_read = 0.1
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 8d99452f988109d44e7f75d1702f483a8f5181fe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 13:27:48 +0000
Subject: [PATCH 0461/2012] chore(sync): update Charm Hyper model catalog
 (#6095)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml | 6 +++---
 providers/hyper/models/glm-5.1.toml            | 6 +++---
 providers/hyper/models/glm-5.toml              | 6 +++---
 providers/hyper/models/gpt-oss-120b.toml       | 6 +++---
 providers/hyper/models/kimi-k2.5.toml          | 6 +++---
 providers/hyper/models/minimax-m2.7.toml       | 6 +++---
 6 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 85218385ac6..7c758447185 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.12
-output = 0.42
-cache_write = 0.06
+input = 0.116
+output = 0.38
+cache_write = 0.058
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index a1b201fde20..4c86ffebd03 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.332
-output = 4.312
-cache_write = 0.666
+input = 1.29
+output = 4.22
+cache_write = 0.645
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 01e0d21e4fb..aaa2e1d6169 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.91
-output = 2.934
-cache_write = 0.455
+input = 0.85
+output = 2.774
+cache_write = 0.425
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index 4995ffb98a5..ad08deed2d7 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.19
-output = 0.63
-cache_write = 0.095
+input = 0.188
+output = 0.7
+cache_write = 0.094
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index ea59490a67b..3d00be969d8 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5504
-output = 2.885
-cache_write = 0.2752
+input = 0.5284
+output = 2.785
+cache_write = 0.2642
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index de3e51335c2..350561786a4 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.404
-output = 1.496
-cache_write = 0.202
+input = 0.426
+output = 1.62
+cache_write = 0.213
 
 [limit]
 context = 262_100

From 6cdd48e31dd964af8cbd1fe9aab9b07567b7ad25 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 13:28:39 +0000
Subject: [PATCH 0462/2012] chore(sync): update OpenRouter model catalog
 (#6094)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 66443ade4c4..4bf670bbd00 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.04226
-output = 2.08452
-cache_read = 0.086855
+input = 1.039302
+output = 2.078604
+cache_read = 0.086609
 
 [limit]
 context = 1_048_576

From a1ae8090f58e0f7ec5661b672cea03b644fb637e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 14:27:55 +0000
Subject: [PATCH 0463/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6097)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.3.toml b/providers/llmgateway/models/glm-5.3.toml
index 9bc1eda393c..66cc365baaa 100644
--- a/providers/llmgateway/models/glm-5.3.toml
+++ b/providers/llmgateway/models/glm-5.3.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.3
+input = 1.2
 output = 4
-cache_read = 0.25
+cache_read = 0.2
 
 [limit]
 context = 1_048_576

From e6709263677b0e5116b04309a35c1a40ace8a05d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 14:28:04 +0000
Subject: [PATCH 0464/2012] chore(sync): update OpenRouter model catalog
 (#6098)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4bf670bbd00..97904bfc0b8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.039302
-output = 2.078604
-cache_read = 0.086609
+input = 1.035996
+output = 2.071992
+cache_read = 0.086333
 
 [limit]
 context = 1_048_576

From 10518f4f9f57e4eef88df801bd6185cf8acdb13d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 14:28:11 +0000
Subject: [PATCH 0465/2012] chore(sync): update LLM Gateway model catalog
 (#6096)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/runware/glm-5.3.toml  | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/runware/glm-5.3.toml

diff --git a/providers/llmgateway-providers/models/runware/glm-5.3.toml b/providers/llmgateway-providers/models/runware/glm-5.3.toml
new file mode 100644
index 00000000000..e73a60cb89e
--- /dev/null
+++ b/providers/llmgateway-providers/models/runware/glm-5.3.toml
@@ -0,0 +1,11 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (Runware)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.2

From a52dadd2785616716ff13c4cf03ffee641a3abc7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 15:30:17 +0000
Subject: [PATCH 0466/2012] chore(sync): update Eden AI model catalog (#6104)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 5 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 3fc44c890c7..465b5f9eb20 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75335
-output = 0.75335
+input = 0.75257
+output = 0.75257
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index dd5e7202e80..46d0db0b3d5 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17385
-output = 0.75335
+input = 0.17367
+output = 0.75257
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index a787a43195b..34338102b03 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4636
-output = 0.9272
+input = 0.46312
+output = 0.92624
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index ffbb2b1b144..e17608fe294 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17385
-output = 0.6954
+input = 0.17367
+output = 0.69468
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 8ef573d4ac8..818df2948ac 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.0431
-output = 1.0431
+input = 1.04202
+output = 1.04202

From bb444886c5d5cac3625b190430364b4261cbb91d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 15:30:44 +0000
Subject: [PATCH 0467/2012] chore(sync): update OpenRouter model catalog
 (#6103)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/google/gemini-3.8-flash.toml       | 30 +++++++++++++++++++
 3 files changed, 36 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/google/gemini-3.8-flash.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 97904bfc0b8..b239680e290 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.035996
-output = 2.071992
-cache_read = 0.086333
+input = 1.034256
+output = 2.068512
+cache_read = 0.086188
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemini-3.8-flash.toml b/providers/openrouter/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..20d17832d30
--- /dev/null
+++ b/providers/openrouter/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,30 @@
+name = "Gemini 3.8 Flash"
+description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
+family = "gemini"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]

From e750f0d19f3cde3322834fa6e4303d726b5d68f2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 15:31:21 +0000
Subject: [PATCH 0468/2012] chore(sync): update Kilo model catalog (#6105)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/google/gemini-3.8-flash.toml  | 30 +++++++++++++++++++
 1 file changed, 30 insertions(+)
 create mode 100644 providers/kilo/models/google/gemini-3.8-flash.toml

diff --git a/providers/kilo/models/google/gemini-3.8-flash.toml b/providers/kilo/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..27385be8262
--- /dev/null
+++ b/providers/kilo/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,30 @@
+name = "Google: Gemini 3.8 Flash"
+description = "Gemini 3.8 Flash is Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows."
+family = "gemini"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]

From 0934c9278d6eecd2ab22540929bb6e8dad5f8d75 Mon Sep 17 00:00:00 2001
From: Asmae_ELAZRAK <113793164+ASMAE20@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:15:39 +0100
Subject: [PATCH 0469/2012] Feat: update cortecs field and models file (#5986)

* fix: add cortecs field

* fix: update models file

* fix: restore reasoning field
---
 packages/core/src/sync/providers/cortecs.ts               | 3 ++-
 providers/cortecs/models/apertus-70b.toml                 | 2 +-
 providers/cortecs/models/claude-4-5-sonnet.toml           | 3 ---
 providers/cortecs/models/claude-4-6-sonnet.toml           | 2 +-
 providers/cortecs/models/claude-haiku-4-5.toml            | 3 ---
 providers/cortecs/models/claude-opus-5.toml               | 3 ---
 providers/cortecs/models/claude-opus4-5.toml              | 3 ---
 providers/cortecs/models/claude-opus4-6.toml              | 3 ---
 providers/cortecs/models/claude-opus4-7.toml              | 3 ---
 providers/cortecs/models/claude-opus4-8.toml              | 3 ---
 providers/cortecs/models/claude-sonnet-4.toml             | 2 +-
 providers/cortecs/models/claude-sonnet-5.toml             | 3 ---
 providers/cortecs/models/gemini-2.5-flash.toml            | 2 +-
 providers/cortecs/models/gemini-3.1-flash-lite.toml       | 2 +-
 providers/cortecs/models/gemini-3.5-flash-lite.toml       | 2 +-
 providers/cortecs/models/gemini-3.5-flash.toml            | 7 +++++--
 providers/cortecs/models/gemini-3.6-flash.toml            | 7 +++++--
 providers/cortecs/models/gemini-3.7-flash.toml            | 2 +-
 providers/cortecs/models/gemma-3-27b-it.toml              | 2 +-
 providers/cortecs/models/gemma-4-26b-a4b-it.toml          | 2 +-
 providers/cortecs/models/glm-4.7.toml                     | 2 +-
 providers/cortecs/models/glm-5.2.toml                     | 2 +-
 providers/cortecs/models/glm-5.3-flash.toml               | 2 +-
 providers/cortecs/models/glm-5.3.toml                     | 2 +-
 providers/cortecs/models/gpt-4.1-mini.toml                | 3 ---
 providers/cortecs/models/gpt-4.1-nano.toml                | 3 ---
 providers/cortecs/models/gpt-4.1.toml                     | 3 ---
 providers/cortecs/models/gpt-4o-mini.toml                 | 2 +-
 providers/cortecs/models/gpt-4o.toml                      | 2 +-
 providers/cortecs/models/gpt-5-mini.toml                  | 3 ---
 providers/cortecs/models/gpt-5-nano.toml                  | 3 ---
 providers/cortecs/models/gpt-5.1.toml                     | 3 ---
 providers/cortecs/models/gpt-5.4.toml                     | 3 ---
 providers/cortecs/models/gpt-5.6-luna.toml                | 3 ---
 providers/cortecs/models/gpt-5.6-sol.toml                 | 3 ---
 providers/cortecs/models/gpt-5.6-terra.toml               | 3 ---
 providers/cortecs/models/gpt-5.toml                       | 3 ---
 providers/cortecs/models/gpt-oss-120b.toml                | 2 +-
 providers/cortecs/models/kimi-k2.5.toml                   | 3 ---
 providers/cortecs/models/kimi-k2.6.toml                   | 3 ---
 providers/cortecs/models/kimi-k3.toml                     | 8 ++++----
 providers/cortecs/models/minimax-m2.7.toml                | 2 +-
 providers/cortecs/models/minimax-m2.toml                  | 2 +-
 providers/cortecs/models/mistral-7b-instruct-v0.2.toml    | 2 +-
 providers/cortecs/models/mistral-large-2402.toml          | 2 +-
 providers/cortecs/models/mistral-nemo-instruct-2407.toml  | 2 +-
 providers/cortecs/models/mistral-small-2603.toml          | 1 -
 providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml  | 2 +-
 providers/cortecs/models/nova-2-lite.toml                 | 2 +-
 providers/cortecs/models/nova-lite-v1.toml                | 2 +-
 providers/cortecs/models/nova-micro-v1.toml               | 2 +-
 providers/cortecs/models/nova-pro-v1.toml                 | 2 +-
 providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml | 2 +-
 providers/cortecs/models/pixtral-12b-2409.toml            | 2 +-
 .../cortecs/models/qwen3-235b-a22b-instruct-2507.toml     | 2 +-
 .../cortecs/models/qwen3-coder-30b-a3b-instruct.toml      | 2 +-
 providers/cortecs/models/qwen3.5-397b-a17b.toml           | 2 +-
 providers/cortecs/models/qwen3.6-35b-a3b.toml             | 2 +-
 58 files changed, 48 insertions(+), 105 deletions(-)

diff --git a/packages/core/src/sync/providers/cortecs.ts b/packages/core/src/sync/providers/cortecs.ts
index 548bcf10124..78403b3c085 100644
--- a/packages/core/src/sync/providers/cortecs.ts
+++ b/packages/core/src/sync/providers/cortecs.ts
@@ -36,6 +36,7 @@ export const CortecsModel = z.object({
     cache_write_cost: z.number().nonnegative().optional(),
   }).passthrough(),
   context_size: z.number().int().positive(),
+  max_output_tokens: z.number().int().positive().optional(),
   input_modalities: z.array(z.string()).transform(modalities).default(["text"]),
   output_modalities: z.array(z.string()).transform(modalities).default(["text"]),
   supported_features: z.array(z.string()).default([]),
@@ -97,7 +98,7 @@ export function buildCortecsModel(
   const limit = {
     context: model.context_size,
     input: existing?.limit?.input,
-    output: authored?.limit?.output ?? model.context_size,
+    output: model.max_output_tokens ?? authored?.limit?.output ?? model.context_size,
   };
   const cost = {
     input: usd(model.pricing.input_token),
diff --git a/providers/cortecs/models/apertus-70b.toml b/providers/cortecs/models/apertus-70b.toml
index 783b1649332..8426b58bca8 100644
--- a/providers/cortecs/models/apertus-70b.toml
+++ b/providers/cortecs/models/apertus-70b.toml
@@ -10,4 +10,4 @@ input = 1.393
 output = 2.228
 
 [limit]
-output = 65_536
+output = 16_384
diff --git a/providers/cortecs/models/claude-4-5-sonnet.toml b/providers/cortecs/models/claude-4-5-sonnet.toml
index 5475848656a..69fc1faeab7 100644
--- a/providers/cortecs/models/claude-4-5-sonnet.toml
+++ b/providers/cortecs/models/claude-4-5-sonnet.toml
@@ -15,6 +15,3 @@ input = 2.989
 output = 14.945
 cache_read = 0.326
 cache_write = 4.078
-
-[limit]
-output = 200_000
diff --git a/providers/cortecs/models/claude-4-6-sonnet.toml b/providers/cortecs/models/claude-4-6-sonnet.toml
index e2d65163b1a..a5f3b417307 100644
--- a/providers/cortecs/models/claude-4-6-sonnet.toml
+++ b/providers/cortecs/models/claude-4-6-sonnet.toml
@@ -17,4 +17,4 @@ cache_read = 0.32
 cache_write = 3.999
 
 [limit]
-output = 1_000_000
+output = 128_000
diff --git a/providers/cortecs/models/claude-haiku-4-5.toml b/providers/cortecs/models/claude-haiku-4-5.toml
index 942a91e1fdc..c485e9f43de 100644
--- a/providers/cortecs/models/claude-haiku-4-5.toml
+++ b/providers/cortecs/models/claude-haiku-4-5.toml
@@ -15,6 +15,3 @@ input = 0.996
 output = 4.982
 cache_read = 0.099
 cache_write = 1.186
-
-[limit]
-output = 200_000
diff --git a/providers/cortecs/models/claude-opus-5.toml b/providers/cortecs/models/claude-opus-5.toml
index 8ae10b28450..426c2b22cd4 100644
--- a/providers/cortecs/models/claude-opus-5.toml
+++ b/providers/cortecs/models/claude-opus-5.toml
@@ -14,6 +14,3 @@ input = 5.5
 output = 27.498
 cache_read = 0.55
 cache_write = 6.874
-
-[limit]
-output = 1_000_000
diff --git a/providers/cortecs/models/claude-opus4-5.toml b/providers/cortecs/models/claude-opus4-5.toml
index 527ff4a8047..fde13af0f52 100644
--- a/providers/cortecs/models/claude-opus4-5.toml
+++ b/providers/cortecs/models/claude-opus4-5.toml
@@ -15,6 +15,3 @@ input = 5.313
 output = 26.568
 cache_read = 0.531
 cache_write = 6.645
-
-[limit]
-output = 200_000
diff --git a/providers/cortecs/models/claude-opus4-6.toml b/providers/cortecs/models/claude-opus4-6.toml
index f2ca3eb90e8..4b81fbe2029 100644
--- a/providers/cortecs/models/claude-opus4-6.toml
+++ b/providers/cortecs/models/claude-opus4-6.toml
@@ -15,6 +15,3 @@ input = 5.313
 output = 26.561
 cache_read = 0.531
 cache_write = 6.645
-
-[limit]
-output = 1_000_000
diff --git a/providers/cortecs/models/claude-opus4-7.toml b/providers/cortecs/models/claude-opus4-7.toml
index 60d9dca5c81..3e358b82ab6 100644
--- a/providers/cortecs/models/claude-opus4-7.toml
+++ b/providers/cortecs/models/claude-opus4-7.toml
@@ -15,6 +15,3 @@ input = 5.437
 output = 27.186
 cache_read = 0.544
 cache_write = 6.797
-
-[limit]
-output = 1_000_000
diff --git a/providers/cortecs/models/claude-opus4-8.toml b/providers/cortecs/models/claude-opus4-8.toml
index 532ea5fb652..02700fd1d57 100644
--- a/providers/cortecs/models/claude-opus4-8.toml
+++ b/providers/cortecs/models/claude-opus4-8.toml
@@ -14,6 +14,3 @@ input = 5.437
 output = 27.186
 cache_read = 0.544
 cache_write = 6.797
-
-[limit]
-output = 1_000_000
diff --git a/providers/cortecs/models/claude-sonnet-4.toml b/providers/cortecs/models/claude-sonnet-4.toml
index 1d96a667821..5661a430c8e 100644
--- a/providers/cortecs/models/claude-sonnet-4.toml
+++ b/providers/cortecs/models/claude-sonnet-4.toml
@@ -17,4 +17,4 @@ cache_read = 0.29
 cache_write = 3.624
 
 [limit]
-output = 200_000
+output = 65_000
diff --git a/providers/cortecs/models/claude-sonnet-5.toml b/providers/cortecs/models/claude-sonnet-5.toml
index a682c995740..9e2f3de9663 100644
--- a/providers/cortecs/models/claude-sonnet-5.toml
+++ b/providers/cortecs/models/claude-sonnet-5.toml
@@ -14,6 +14,3 @@ input = 2.2
 output = 11
 cache_read = 0.219
 cache_write = 2.749
-
-[limit]
-output = 1_000_000
diff --git a/providers/cortecs/models/gemini-2.5-flash.toml b/providers/cortecs/models/gemini-2.5-flash.toml
index 4d841841ee5..f00fc4a0d8e 100644
--- a/providers/cortecs/models/gemini-2.5-flash.toml
+++ b/providers/cortecs/models/gemini-2.5-flash.toml
@@ -8,7 +8,7 @@ cache_read = 0.029
 cache_write = 0.097
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio", "pdf"]
diff --git a/providers/cortecs/models/gemini-3.1-flash-lite.toml b/providers/cortecs/models/gemini-3.1-flash-lite.toml
index cad8b92430e..db91e44dc3f 100644
--- a/providers/cortecs/models/gemini-3.1-flash-lite.toml
+++ b/providers/cortecs/models/gemini-3.1-flash-lite.toml
@@ -8,7 +8,7 @@ cache_read = 0.025
 cache_write = 0.082
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio", "pdf"]
diff --git a/providers/cortecs/models/gemini-3.5-flash-lite.toml b/providers/cortecs/models/gemini-3.5-flash-lite.toml
index b10a0dd42aa..a4f70fb7fb1 100644
--- a/providers/cortecs/models/gemini-3.5-flash-lite.toml
+++ b/providers/cortecs/models/gemini-3.5-flash-lite.toml
@@ -10,7 +10,7 @@ output = 2.749
 cache_read = 0.033
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio", "pdf"]
diff --git a/providers/cortecs/models/gemini-3.5-flash.toml b/providers/cortecs/models/gemini-3.5-flash.toml
index a9041ead902..d56b015c984 100644
--- a/providers/cortecs/models/gemini-3.5-flash.toml
+++ b/providers/cortecs/models/gemini-3.5-flash.toml
@@ -1,5 +1,8 @@
 base_model = "google/gemini-3.5-flash"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 1.649
@@ -8,7 +11,7 @@ cache_read = 0.165
 cache_write = 1
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/cortecs/models/gemini-3.6-flash.toml b/providers/cortecs/models/gemini-3.6-flash.toml
index d575827d8c3..23c47e0f96c 100644
--- a/providers/cortecs/models/gemini-3.6-flash.toml
+++ b/providers/cortecs/models/gemini-3.6-flash.toml
@@ -1,5 +1,8 @@
 base_model = "google/gemini-3.6-flash"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 0.75
@@ -8,7 +11,7 @@ cache_read = 0.075
 cache_write = 0.038
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/cortecs/models/gemini-3.7-flash.toml b/providers/cortecs/models/gemini-3.7-flash.toml
index fac48db17e5..e323f03e232 100644
--- a/providers/cortecs/models/gemini-3.7-flash.toml
+++ b/providers/cortecs/models/gemini-3.7-flash.toml
@@ -11,7 +11,7 @@ cache_read = 0.075
 cache_write = 0.038
 
 [limit]
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image", "audio", "pdf"]
diff --git a/providers/cortecs/models/gemma-3-27b-it.toml b/providers/cortecs/models/gemma-3-27b-it.toml
index 29c974f0a53..ee11f1ec4c3 100644
--- a/providers/cortecs/models/gemma-3-27b-it.toml
+++ b/providers/cortecs/models/gemma-3-27b-it.toml
@@ -16,7 +16,7 @@ output = 0.299
 
 [limit]
 context = 131_000
-output = 131_000
+output = 110_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/cortecs/models/gemma-4-26b-a4b-it.toml b/providers/cortecs/models/gemma-4-26b-a4b-it.toml
index dd9940d78d3..8b2a7a279a8 100644
--- a/providers/cortecs/models/gemma-4-26b-a4b-it.toml
+++ b/providers/cortecs/models/gemma-4-26b-a4b-it.toml
@@ -7,4 +7,4 @@ output = 0.557
 
 [limit]
 context = 262_000
-output = 262_000
+output = 81_920
diff --git a/providers/cortecs/models/glm-4.7.toml b/providers/cortecs/models/glm-4.7.toml
index 18cd487e492..e3b731338d6 100644
--- a/providers/cortecs/models/glm-4.7.toml
+++ b/providers/cortecs/models/glm-4.7.toml
@@ -12,4 +12,4 @@ output = 2.785
 
 [limit]
 context = 202_752
-output = 198_000
+output = 202_752
diff --git a/providers/cortecs/models/glm-5.2.toml b/providers/cortecs/models/glm-5.2.toml
index 48706497a1b..984c3e96d79 100644
--- a/providers/cortecs/models/glm-5.2.toml
+++ b/providers/cortecs/models/glm-5.2.toml
@@ -14,4 +14,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 1_000_000
diff --git a/providers/cortecs/models/glm-5.3-flash.toml b/providers/cortecs/models/glm-5.3-flash.toml
index 6836aac9380..6f2582b9c88 100644
--- a/providers/cortecs/models/glm-5.3-flash.toml
+++ b/providers/cortecs/models/glm-5.3-flash.toml
@@ -17,7 +17,7 @@ cache_read = 0.05
 
 [limit]
 context = 1_048_576
-output = 64_000
+output = 1_048_576
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/cortecs/models/glm-5.3.toml b/providers/cortecs/models/glm-5.3.toml
index d2aacc59483..95ee62cddee 100644
--- a/providers/cortecs/models/glm-5.3.toml
+++ b/providers/cortecs/models/glm-5.3.toml
@@ -15,4 +15,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 64_000
diff --git a/providers/cortecs/models/gpt-4.1-mini.toml b/providers/cortecs/models/gpt-4.1-mini.toml
index 3d2310d557d..5268c76faa2 100644
--- a/providers/cortecs/models/gpt-4.1-mini.toml
+++ b/providers/cortecs/models/gpt-4.1-mini.toml
@@ -5,6 +5,3 @@ structured_output = false
 input = 0.434
 output = 1.704
 cache_read = 0.134
-
-[limit]
-output = 1_047_576
diff --git a/providers/cortecs/models/gpt-4.1-nano.toml b/providers/cortecs/models/gpt-4.1-nano.toml
index 0c6a4e99e56..b6cb72a1908 100644
--- a/providers/cortecs/models/gpt-4.1-nano.toml
+++ b/providers/cortecs/models/gpt-4.1-nano.toml
@@ -6,8 +6,5 @@ input = 0.111
 output = 0.434
 cache_read = 0.056
 
-[limit]
-output = 1_047_576
-
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/gpt-4.1.toml b/providers/cortecs/models/gpt-4.1.toml
index d8c2e921d2d..31f3538c837 100644
--- a/providers/cortecs/models/gpt-4.1.toml
+++ b/providers/cortecs/models/gpt-4.1.toml
@@ -6,6 +6,3 @@ structured_output = false
 input = 2.192
 output = 8.769
 cache_read = 0.546
-
-[limit]
-output = 1_047_576
diff --git a/providers/cortecs/models/gpt-4o-mini.toml b/providers/cortecs/models/gpt-4o-mini.toml
index d8ac084b883..99352a76f62 100644
--- a/providers/cortecs/models/gpt-4o-mini.toml
+++ b/providers/cortecs/models/gpt-4o-mini.toml
@@ -7,4 +7,4 @@ output = 0.638
 cache_read = 0.081
 
 [limit]
-output = 128_000
+output = 16_000
diff --git a/providers/cortecs/models/gpt-4o.toml b/providers/cortecs/models/gpt-4o.toml
index d75037a4936..83c6e434966 100644
--- a/providers/cortecs/models/gpt-4o.toml
+++ b/providers/cortecs/models/gpt-4o.toml
@@ -7,4 +7,4 @@ output = 10.635
 cache_read = 1.33
 
 [limit]
-output = 128_000
+output = 16_000
diff --git a/providers/cortecs/models/gpt-5-mini.toml b/providers/cortecs/models/gpt-5-mini.toml
index e23c43c9146..70417f62649 100644
--- a/providers/cortecs/models/gpt-5-mini.toml
+++ b/providers/cortecs/models/gpt-5-mini.toml
@@ -9,6 +9,3 @@ values = ["low", "medium", "high"]
 input = 0.279
 output = 2.192
 cache_read = 0.056
-
-[limit]
-output = 400_000
diff --git a/providers/cortecs/models/gpt-5-nano.toml b/providers/cortecs/models/gpt-5-nano.toml
index b2d796c6046..ab55c508d23 100644
--- a/providers/cortecs/models/gpt-5-nano.toml
+++ b/providers/cortecs/models/gpt-5-nano.toml
@@ -9,6 +9,3 @@ values = ["low", "medium", "high"]
 input = 0.06
 output = 0.439
 cache_read = 0.019
-
-[limit]
-output = 400_000
diff --git a/providers/cortecs/models/gpt-5.1.toml b/providers/cortecs/models/gpt-5.1.toml
index be114ef4418..9f429a64d6b 100644
--- a/providers/cortecs/models/gpt-5.1.toml
+++ b/providers/cortecs/models/gpt-5.1.toml
@@ -10,8 +10,5 @@ input = 1.375
 output = 10.96
 cache_read = 0.156
 
-[limit]
-output = 400_000
-
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/gpt-5.4.toml b/providers/cortecs/models/gpt-5.4.toml
index f5bac51cadd..5c151e3cf46 100644
--- a/providers/cortecs/models/gpt-5.4.toml
+++ b/providers/cortecs/models/gpt-5.4.toml
@@ -9,6 +9,3 @@ values = ["low", "medium", "high"]
 input = 2.898
 output = 15.453
 cache_read = 0.242
-
-[limit]
-output = 1_050_000
diff --git a/providers/cortecs/models/gpt-5.6-luna.toml b/providers/cortecs/models/gpt-5.6-luna.toml
index dacd7a3259d..622fb066769 100644
--- a/providers/cortecs/models/gpt-5.6-luna.toml
+++ b/providers/cortecs/models/gpt-5.6-luna.toml
@@ -9,6 +9,3 @@ input = 0.219
 output = 1.32
 cache_read = 0.022
 cache_write = 0.275
-
-[limit]
-output = 1_050_000
diff --git a/providers/cortecs/models/gpt-5.6-sol.toml b/providers/cortecs/models/gpt-5.6-sol.toml
index 9fc74af0fe1..753a0ca5ff5 100644
--- a/providers/cortecs/models/gpt-5.6-sol.toml
+++ b/providers/cortecs/models/gpt-5.6-sol.toml
@@ -9,6 +9,3 @@ input = 5.5
 output = 32.998
 cache_read = 0.55
 cache_write = 6.879
-
-[limit]
-output = 1_050_000
diff --git a/providers/cortecs/models/gpt-5.6-terra.toml b/providers/cortecs/models/gpt-5.6-terra.toml
index d3006e80082..4839da3c8eb 100644
--- a/providers/cortecs/models/gpt-5.6-terra.toml
+++ b/providers/cortecs/models/gpt-5.6-terra.toml
@@ -9,6 +9,3 @@ input = 2.2
 output = 13.199
 cache_read = 0.219
 cache_write = 2.749
-
-[limit]
-output = 1_050_000
diff --git a/providers/cortecs/models/gpt-5.toml b/providers/cortecs/models/gpt-5.toml
index 32fd3f83daa..5073da5f49a 100644
--- a/providers/cortecs/models/gpt-5.toml
+++ b/providers/cortecs/models/gpt-5.toml
@@ -9,6 +9,3 @@ values = ["low", "medium", "high"]
 input = 1.375
 output = 10.96
 cache_read = 0.156
-
-[limit]
-output = 400_000
diff --git a/providers/cortecs/models/gpt-oss-120b.toml b/providers/cortecs/models/gpt-oss-120b.toml
index 7391e05e7e9..9abc1327a5e 100644
--- a/providers/cortecs/models/gpt-oss-120b.toml
+++ b/providers/cortecs/models/gpt-oss-120b.toml
@@ -12,4 +12,4 @@ cache_read = 0.01
 
 [limit]
 context = 131_000
-output = 128_000
+output = 131_000
diff --git a/providers/cortecs/models/kimi-k2.5.toml b/providers/cortecs/models/kimi-k2.5.toml
index 6e9672dcb6f..1440935f803 100644
--- a/providers/cortecs/models/kimi-k2.5.toml
+++ b/providers/cortecs/models/kimi-k2.5.toml
@@ -13,8 +13,5 @@ input = 0.495
 output = 2.768
 cache_read = 0.124
 
-[limit]
-output = 256_000
-
 [modalities]
 input = ["text", "image"]
diff --git a/providers/cortecs/models/kimi-k2.6.toml b/providers/cortecs/models/kimi-k2.6.toml
index f8d00fc7213..6d220564edf 100644
--- a/providers/cortecs/models/kimi-k2.6.toml
+++ b/providers/cortecs/models/kimi-k2.6.toml
@@ -12,8 +12,5 @@ input = 0.773
 output = 3.38
 cache_read = 0.193
 
-[limit]
-output = 256_000
-
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/kimi-k3.toml b/providers/cortecs/models/kimi-k3.toml
index 002c2338efb..c8d1f2dc939 100644
--- a/providers/cortecs/models/kimi-k3.toml
+++ b/providers/cortecs/models/kimi-k3.toml
@@ -1,8 +1,8 @@
-# Cortecs lists EUR 2.693 input and EUR 13.464 output per million tokens.
-# Converted at 1 EUR = 1.114 USD and rounded to the provider's displayed USD rates.
-# https://api.cortecs.ai/v1/models (accessed 2026-07-29)
 base_model = "moonshotai/kimi-k3"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/cortecs/models/minimax-m2.7.toml b/providers/cortecs/models/minimax-m2.7.toml
index de169793535..ed8a05934b6 100644
--- a/providers/cortecs/models/minimax-m2.7.toml
+++ b/providers/cortecs/models/minimax-m2.7.toml
@@ -9,4 +9,4 @@ output = 2.674
 
 [limit]
 context = 196_608
-output = 196_072
+output = 196_608
diff --git a/providers/cortecs/models/minimax-m2.toml b/providers/cortecs/models/minimax-m2.toml
index 6e428e287e3..eecd36f3753 100644
--- a/providers/cortecs/models/minimax-m2.toml
+++ b/providers/cortecs/models/minimax-m2.toml
@@ -12,4 +12,4 @@ output = 1.405
 
 [limit]
 context = 400_000
-output = 400_000
+output = 196_000
diff --git a/providers/cortecs/models/mistral-7b-instruct-v0.2.toml b/providers/cortecs/models/mistral-7b-instruct-v0.2.toml
index ce87e65b9bf..a601fdfa2c7 100644
--- a/providers/cortecs/models/mistral-7b-instruct-v0.2.toml
+++ b/providers/cortecs/models/mistral-7b-instruct-v0.2.toml
@@ -15,7 +15,7 @@ output = 0.219
 
 [limit]
 context = 32_000
-output = 32_000
+output = 8_192
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/cortecs/models/mistral-large-2402.toml b/providers/cortecs/models/mistral-large-2402.toml
index 0c7134b9313..af6a48d7bbd 100644
--- a/providers/cortecs/models/mistral-large-2402.toml
+++ b/providers/cortecs/models/mistral-large-2402.toml
@@ -16,7 +16,7 @@ output = 12.952
 
 [limit]
 context = 32_000
-output = 32_000
+output = 8_192
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/cortecs/models/mistral-nemo-instruct-2407.toml b/providers/cortecs/models/mistral-nemo-instruct-2407.toml
index 6b38bfabf38..fc3b22b762a 100644
--- a/providers/cortecs/models/mistral-nemo-instruct-2407.toml
+++ b/providers/cortecs/models/mistral-nemo-instruct-2407.toml
@@ -16,7 +16,7 @@ cache_read = 0.014
 
 [limit]
 context = 128_000
-output = 131_072
+output = 128_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/cortecs/models/mistral-small-2603.toml b/providers/cortecs/models/mistral-small-2603.toml
index 4ec49734866..89e30734258 100644
--- a/providers/cortecs/models/mistral-small-2603.toml
+++ b/providers/cortecs/models/mistral-small-2603.toml
@@ -9,4 +9,3 @@ cache_read = 0.014
 
 [limit]
 context = 262_144
-output = 262_144
diff --git a/providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml b/providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml
index 2551797c97f..08b5b8de12b 100644
--- a/providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml
+++ b/providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml
@@ -17,7 +17,7 @@ output = 0.758
 
 [limit]
 context = 32_000
-output = 32_000
+output = 4_096
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/cortecs/models/nova-2-lite.toml b/providers/cortecs/models/nova-2-lite.toml
index 2b4e1653e2d..dea37eb50ba 100644
--- a/providers/cortecs/models/nova-2-lite.toml
+++ b/providers/cortecs/models/nova-2-lite.toml
@@ -16,7 +16,7 @@ output = 3.144
 
 [limit]
 context = 1_000_000
-output = 1_000_000
+output = 65_535
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/nova-lite-v1.toml b/providers/cortecs/models/nova-lite-v1.toml
index c3f62b107d4..54c1e1a2dbd 100644
--- a/providers/cortecs/models/nova-lite-v1.toml
+++ b/providers/cortecs/models/nova-lite-v1.toml
@@ -16,7 +16,7 @@ output = 0.275
 
 [limit]
 context = 300_000
-output = 300_000
+output = 10_000
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/nova-micro-v1.toml b/providers/cortecs/models/nova-micro-v1.toml
index ce8595fb24c..da65b2b4568 100644
--- a/providers/cortecs/models/nova-micro-v1.toml
+++ b/providers/cortecs/models/nova-micro-v1.toml
@@ -16,7 +16,7 @@ output = 0.159
 
 [limit]
 context = 128_000
-output = 128_000
+output = 10_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/cortecs/models/nova-pro-v1.toml b/providers/cortecs/models/nova-pro-v1.toml
index 262b70c0844..bc55aace80b 100644
--- a/providers/cortecs/models/nova-pro-v1.toml
+++ b/providers/cortecs/models/nova-pro-v1.toml
@@ -18,7 +18,7 @@ output = 3.671
 
 [limit]
 context = 300_000
-output = 5_000
+output = 10_000
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml b/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
index 4b8a70d55a5..8dbfc72233f 100644
--- a/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
+++ b/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
@@ -16,7 +16,7 @@ output = 0.237
 
 [limit]
 context = 300_000
-output = 300_000
+output = 65_536
 
 [modalities]
 input = ["text"]
diff --git a/providers/cortecs/models/pixtral-12b-2409.toml b/providers/cortecs/models/pixtral-12b-2409.toml
index 62a978e522d..cbcac29c708 100644
--- a/providers/cortecs/models/pixtral-12b-2409.toml
+++ b/providers/cortecs/models/pixtral-12b-2409.toml
@@ -16,7 +16,7 @@ output = 0.223
 
 [limit]
 context = 128_000
-output = 128_000
+output = 4_096
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/cortecs/models/qwen3-235b-a22b-instruct-2507.toml b/providers/cortecs/models/qwen3-235b-a22b-instruct-2507.toml
index c2159b2636b..1d3bf693a61 100644
--- a/providers/cortecs/models/qwen3-235b-a22b-instruct-2507.toml
+++ b/providers/cortecs/models/qwen3-235b-a22b-instruct-2507.toml
@@ -9,4 +9,4 @@ cache_read = 0.018
 
 [limit]
 context = 262_000
-output = 131_000
+output = 262_000
diff --git a/providers/cortecs/models/qwen3-coder-30b-a3b-instruct.toml b/providers/cortecs/models/qwen3-coder-30b-a3b-instruct.toml
index 09bb3833a2a..12e76ebfac8 100644
--- a/providers/cortecs/models/qwen3-coder-30b-a3b-instruct.toml
+++ b/providers/cortecs/models/qwen3-coder-30b-a3b-instruct.toml
@@ -7,4 +7,4 @@ output = 0.245
 cache_read = 0.014
 
 [limit]
-output = 262_000
+output = 262_144
diff --git a/providers/cortecs/models/qwen3.5-397b-a17b.toml b/providers/cortecs/models/qwen3.5-397b-a17b.toml
index a2ce4e256da..9a484c4ea40 100644
--- a/providers/cortecs/models/qwen3.5-397b-a17b.toml
+++ b/providers/cortecs/models/qwen3.5-397b-a17b.toml
@@ -9,7 +9,7 @@ output = 4.01
 
 [limit]
 context = 262_000
-output = 250_000
+output = 262_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/cortecs/models/qwen3.6-35b-a3b.toml b/providers/cortecs/models/qwen3.6-35b-a3b.toml
index 2c3a0bd4707..247086bb784 100644
--- a/providers/cortecs/models/qwen3.6-35b-a3b.toml
+++ b/providers/cortecs/models/qwen3.6-35b-a3b.toml
@@ -7,7 +7,7 @@ output = 0.557
 
 [limit]
 context = 262_000
-output = 262_000
+output = 32_768
 
 [modalities]
 input = ["text", "image"]

From 552d456df7a35929cff2c1a9595ed7b48fe0f0e5 Mon Sep 17 00:00:00 2001
From: Nabs <nabil@instafork.com>
Date: Wed, 2 Sep 2026 17:15:59 +0100
Subject: [PATCH 0470/2012] feat(google): add Gemini 3.8 Flash (#6106)

---
 models/google/gemini-3.8-flash.toml           | 22 +++++++++++++++++++
 .../models/gemini-3.8-flash.toml              | 11 ++++++++++
 providers/google/models/gemini-3.8-flash.toml | 14 ++++++++++++
 3 files changed, 47 insertions(+)
 create mode 100644 models/google/gemini-3.8-flash.toml
 create mode 100644 providers/google-vertex/models/gemini-3.8-flash.toml
 create mode 100644 providers/google/models/gemini-3.8-flash.toml

diff --git a/models/google/gemini-3.8-flash.toml b/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..c568d5402c1
--- /dev/null
+++ b/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,22 @@
+# Sources:
+# - https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
+# - https://ai.google.dev/gemini-api/docs/latest-model
+name = "Gemini 3.8 Flash"
+description = "Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows"
+family = "gemini-flash"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]
+output = ["text"]
diff --git a/providers/google-vertex/models/gemini-3.8-flash.toml b/providers/google-vertex/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..b134b307f96
--- /dev/null
+++ b/providers/google-vertex/models/gemini-3.8-flash.toml
@@ -0,0 +1,11 @@
+# Sources:
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+# - https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
+base_model = "google/gemini-3.8-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75
diff --git a/providers/google/models/gemini-3.8-flash.toml b/providers/google/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..071c149245f
--- /dev/null
+++ b/providers/google/models/gemini-3.8-flash.toml
@@ -0,0 +1,14 @@
+# Sources:
+# - https://ai.google.dev/gemini-api/docs/pricing
+# - https://ai.google.dev/gemini-api/docs/latest-model
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75

From db3d5e679c0a5e9d5cb9b4624e7df92acb9ed536 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 11:16:14 -0500
Subject: [PATCH 0471/2012] chore(sync): update NanoGPT model catalog (#6100)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/glm-4.7.toml    | 15 ----------
 .../models/cohere/north-mini-code.toml        | 13 ---------
 .../models/command-a-plus-05-2026.toml        | 16 ----------
 .../models/google/gemini-3.8-flash.toml       | 29 +++++++++++++++++++
 .../models/kwaipilot/kat-coder-pro-v2.5.toml  | 24 ---------------
 .../models/kwaipilot/kat-coder-pro-v2.toml    | 25 ----------------
 .../models/openai/gpt-4-turbo-preview.toml    | 23 ---------------
 .../openai/gpt-4o-mini-search-preview.toml    | 24 ---------------
 .../models/openai/gpt-4o-search-preview.toml  | 24 ---------------
 .../nano-gpt/models/openai/gpt-5-codex.toml   | 16 ----------
 .../models/openai/o3-deep-research.toml       | 20 -------------
 .../models/openai/o4-mini-deep-research.toml  | 20 -------------
 providers/nano-gpt/models/sarvam-30b.toml     | 17 -----------
 13 files changed, 29 insertions(+), 237 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/glm-4.7.toml
 delete mode 100644 providers/nano-gpt/models/cohere/north-mini-code.toml
 delete mode 100644 providers/nano-gpt/models/command-a-plus-05-2026.toml
 create mode 100644 providers/nano-gpt/models/google/gemini-3.8-flash.toml
 delete mode 100644 providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.5.toml
 delete mode 100644 providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.toml
 delete mode 100644 providers/nano-gpt/models/openai/gpt-4-turbo-preview.toml
 delete mode 100644 providers/nano-gpt/models/openai/gpt-4o-mini-search-preview.toml
 delete mode 100644 providers/nano-gpt/models/openai/gpt-4o-search-preview.toml
 delete mode 100644 providers/nano-gpt/models/openai/gpt-5-codex.toml
 delete mode 100644 providers/nano-gpt/models/openai/o3-deep-research.toml
 delete mode 100644 providers/nano-gpt/models/openai/o4-mini-deep-research.toml
 delete mode 100644 providers/nano-gpt/models/sarvam-30b.toml

diff --git a/providers/nano-gpt/models/TEE/glm-4.7.toml b/providers/nano-gpt/models/TEE/glm-4.7.toml
deleted file mode 100644
index 30b32ec9a5d..00000000000
--- a/providers/nano-gpt/models/TEE/glm-4.7.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "zhipuai/glm-4.7"
-name = "GLM 4.7 TEE"
-reasoning = false
-tool_call = false
-structured_output = false
-
-[cost]
-input = 0.85
-output = 3.3
-cache_read = 0.425
-
-[limit]
-context = 131_000
-input = 131_000
-output = 65_535
diff --git a/providers/nano-gpt/models/cohere/north-mini-code.toml b/providers/nano-gpt/models/cohere/north-mini-code.toml
deleted file mode 100644
index 5add4237c89..00000000000
--- a/providers/nano-gpt/models/cohere/north-mini-code.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "cohere/north-mini-code-1-0"
-name = "Cohere North Mini Code 1.0"
-tool_call = false
-structured_output = false
-reasoning_options = []
-
-[cost]
-input = 0.2
-output = 0.8
-cache_read = 0.1
-
-[limit]
-input = 256_000
diff --git a/providers/nano-gpt/models/command-a-plus-05-2026.toml b/providers/nano-gpt/models/command-a-plus-05-2026.toml
deleted file mode 100644
index c53f6278954..00000000000
--- a/providers/nano-gpt/models/command-a-plus-05-2026.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-# Not included in subscription
-base_model = "cohere/command-a-plus-05-2026"
-name = "Cohere Command A+ (05/2026)"
-tool_call = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 2.5
-output = 10
-cache_read = 1.25
-
-[limit]
-input = 128_000
diff --git a/providers/nano-gpt/models/google/gemini-3.8-flash.toml b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..27ce1105456
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,29 @@
+name = "Gemini 3.8 Flash"
+description = "Google's fast multimodal model for agentic workloads, including coding, tool use, image understanding, PDF and document extraction, audio, and video. Its capabilities, limits, reasoning behavior, and provisional pricing currently mirror Gemini 3.7 Flash."
+family = "gemini"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.375
+output = 1.875
+cache_read = 0.0375
+cache_write = 0.020833
+
+[limit]
+context = 1_048_576
+input = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.5.toml b/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.5.toml
deleted file mode 100644
index 21d84bb8053..00000000000
--- a/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "KAT Coder Pro V2.5"
-description = "Higher-capability KAT Coder model for complex code generation, repository-scale editing, debugging, and agentic software-development workflows."
-family = "kat-coder"
-release_date = "2026-07-14"
-last_updated = "2026-07-14"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.74
-output = 2.96
-cache_read = 0.15
-
-[limit]
-context = 256_000
-input = 256_000
-output = 80_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.toml b/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.toml
deleted file mode 100644
index d60c7fb0bc1..00000000000
--- a/providers/nano-gpt/models/kwaipilot/kat-coder-pro-v2.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Not included in subscription
-name = "KAT Coder Pro V2"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-03-28"
-last_updated = "2026-03-28"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.15
-
-[limit]
-context = 256_000
-input = 256_000
-output = 80_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-4-turbo-preview.toml b/providers/nano-gpt/models/openai/gpt-4-turbo-preview.toml
deleted file mode 100644
index d875ca7149b..00000000000
--- a/providers/nano-gpt/models/openai/gpt-4-turbo-preview.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "GPT-4 Turbo Preview"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-01"
-last_updated = "2024-01-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 10
-output = 30
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-4o-mini-search-preview.toml b/providers/nano-gpt/models/openai/gpt-4o-mini-search-preview.toml
deleted file mode 100644
index c800c0963e9..00000000000
--- a/providers/nano-gpt/models/openai/gpt-4o-mini-search-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GPT-4o mini Search Preview"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt-mini"
-release_date = "2026-03-10"
-last_updated = "2024-07-18"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.075
-
-[limit]
-context = 128_000
-input = 128_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-4o-search-preview.toml b/providers/nano-gpt/models/openai/gpt-4o-search-preview.toml
deleted file mode 100644
index d62866deb54..00000000000
--- a/providers/nano-gpt/models/openai/gpt-4o-search-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GPT-4o Search Preview"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-03-10"
-last_updated = "2024-05-13"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.5
-output = 10
-cache_read = 1.25
-
-[limit]
-context = 128_000
-input = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-5-codex.toml b/providers/nano-gpt/models/openai/gpt-5-codex.toml
deleted file mode 100644
index 764a0f7b309..00000000000
--- a/providers/nano-gpt/models/openai/gpt-5-codex.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "openai/gpt-5-codex"
-name = "GPT-5 Codex"
-reasoning_options = []
-
-[cost]
-input = 1.25
-output = 10
-cache_read = 0.125
-
-[limit]
-context = 256_000
-input = 256_000
-output = 32_768
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/openai/o3-deep-research.toml b/providers/nano-gpt/models/openai/o3-deep-research.toml
deleted file mode 100644
index b6822f52d20..00000000000
--- a/providers/nano-gpt/models/openai/o3-deep-research.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-base_model = "openai/o3-deep-research"
-name = "OpenAI o3 Deep Research"
-attachment = false
-tool_call = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["medium"]
-
-[cost]
-input = 11
-output = 44
-cache_read = 5.5
-
-[limit]
-input = 200_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/openai/o4-mini-deep-research.toml b/providers/nano-gpt/models/openai/o4-mini-deep-research.toml
deleted file mode 100644
index 0c56b868c8e..00000000000
--- a/providers/nano-gpt/models/openai/o4-mini-deep-research.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-base_model = "openai/o4-mini-deep-research"
-name = "OpenAI o4-mini Deep Research"
-attachment = false
-tool_call = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["medium"]
-
-[cost]
-input = 2.2
-output = 8.8
-cache_read = 1.1
-
-[limit]
-input = 200_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/sarvam-30b.toml b/providers/nano-gpt/models/sarvam-30b.toml
deleted file mode 100644
index fdceb106862..00000000000
--- a/providers/nano-gpt/models/sarvam-30b.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-# Not included in subscription
-base_model = "sarvam/sarvam-30b"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.028
-output = 0.111
-cache_read = 0.017
-
-[limit]
-context = 65_536
-input = 65_536
-output = 4_096

From e0160ce821ae3da9ecea8c80c65c965d280e0336 Mon Sep 17 00:00:00 2001
From: Ryan Scott <rscott@zendesk.com>
Date: Thu, 3 Sep 2026 02:16:54 +1000
Subject: [PATCH 0472/2012] fix(fireworks-ai): update GLM 5.3 Flash cache
 pricing (#6093)

---
 .../models/accounts/fireworks/models/glm-5p3-flash.toml       | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
index 9c92c6284f7..7e8d208090a 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
@@ -1,4 +1,4 @@
-# Pricing: https://fireworks.ai/models/fireworks/glm-5p3-flash (accessed 2026-08-29)
+# Pricing: https://fireworks.ai/models/fireworks/glm-5p3-flash (accessed 2026-09-02)
 # Released on Fireworks Serverless API 2026-08-26.
 # Thinking-only model: reasoning cannot be disabled. Use the same two effective
 # tiers as Fireworks GLM 5.3: low/medium collapse to high; max/xhigh select max.
@@ -16,4 +16,4 @@ field = "reasoning_content"
 [cost]
 input = 0.15
 output = 0.50
-cache_read = 0.029
+cache_read = 0.03

From 05116e518c3cf8caff5a604224c2e2b1b131a47a Mon Sep 17 00:00:00 2001
From: Philip Olson <philip.olson@databricks.com>
Date: Wed, 2 Sep 2026 09:17:29 -0700
Subject: [PATCH 0473/2012] Neon: correct [cost] on six models (#6062)

Bring six neon [cost] blocks in line with Neon's published catalog at
https://neon.com/models.json.
---
 providers/neon/models/gpt-5-6-luna.toml  | 14 ++++++++------
 providers/neon/models/gpt-5-6-sol.toml   |  2 ++
 providers/neon/models/gpt-5-6-terra.toml | 14 ++++++++------
 providers/neon/models/gpt-oss-120b.toml  |  4 ++--
 providers/neon/models/gpt-oss-20b.toml   |  4 ++--
 providers/neon/models/inkling.toml       |  5 +++++
 6 files changed, 27 insertions(+), 16 deletions(-)

diff --git a/providers/neon/models/gpt-5-6-luna.toml b/providers/neon/models/gpt-5-6-luna.toml
index 6018ac810d4..55c129cc5c3 100644
--- a/providers/neon/models/gpt-5-6-luna.toml
+++ b/providers/neon/models/gpt-5-6-luna.toml
@@ -2,15 +2,17 @@ base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
-input = 1
-output = 6
-cache_read = 0.1
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 2
-output = 9
-cache_read = 0.2
+input = 0.4
+output = 1.8
+cache_read = 0.04
+cache_write = 0.5
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/neon/models/gpt-5-6-sol.toml b/providers/neon/models/gpt-5-6-sol.toml
index ee6667402b5..f7c9d9fbae4 100644
--- a/providers/neon/models/gpt-5-6-sol.toml
+++ b/providers/neon/models/gpt-5-6-sol.toml
@@ -5,12 +5,14 @@ reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high
 input = 5
 output = 30
 cache_read = 0.5
+cache_write = 6.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
 input = 10
 output = 45
 cache_read = 1
+cache_write = 12.5
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/neon/models/gpt-5-6-terra.toml b/providers/neon/models/gpt-5-6-terra.toml
index 4ea2eb9233f..0881c058469 100644
--- a/providers/neon/models/gpt-5-6-terra.toml
+++ b/providers/neon/models/gpt-5-6-terra.toml
@@ -2,15 +2,17 @@ base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
-input = 2.5
-output = 15
-cache_read = 0.25
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 5
-output = 22.5
-cache_read = 0.5
+input = 4
+output = 18
+cache_read = 0.4
+cache_write = 5
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/neon/models/gpt-oss-120b.toml b/providers/neon/models/gpt-oss-120b.toml
index 869b4931685..2b2aa08fa90 100644
--- a/providers/neon/models/gpt-oss-120b.toml
+++ b/providers/neon/models/gpt-oss-120b.toml
@@ -2,8 +2,8 @@ base_model = "openai/gpt-oss-120b"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 0.072
-output = 0.28
+input = 0.15
+output = 0.6
 
 # The Neon gateway caps output below the upstream entry: `max_tokens` one above
 # 25_000 returns 400. Measured at the boundary 2026-08-08. `context` is inherited.
diff --git a/providers/neon/models/gpt-oss-20b.toml b/providers/neon/models/gpt-oss-20b.toml
index 05d93cb5b2c..b7c9aff260d 100644
--- a/providers/neon/models/gpt-oss-20b.toml
+++ b/providers/neon/models/gpt-oss-20b.toml
@@ -12,8 +12,8 @@ open_weights = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 0.05
-output = 0.20
+input = 0.07
+output = 0.3
 
 [limit]
 context = 131_072
diff --git a/providers/neon/models/inkling.toml b/providers/neon/models/inkling.toml
index d86e80c646b..01e473cd300 100644
--- a/providers/neon/models/inkling.toml
+++ b/providers/neon/models/inkling.toml
@@ -8,6 +8,11 @@ reasoning_options = [
   { type = "effort", values = ["none", "minimal", "low", "medium", "high", "max"] }, # API: {"reasoning_effort": <v>}
 ]
 
+[cost]
+input = 1
+output = 4.05
+cache_read = 0.17
+
 # The Neon gateway caps output below the upstream entry: `max_tokens` one above
 # 65_536 returns 400. Measured at the boundary 2026-08-08. `context` is inherited.
 [limit]

From 4b992de68fbe7d92e6a1e4879acf8c2b62a08197 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 3 Sep 2026 00:26:17 +0800
Subject: [PATCH 0474/2012] feat(opencode): add Muse Spark 1.3 and Gemini 3.8
 Flash

---
 models/google/gemini-3.8-flash.toml           | 52 ++++++++++++++++++-
 models/meta/muse-spark-1.3.toml               | 33 ++++++++++++
 .../models/muse-spark-1.3-contributor.toml    | 12 +++++
 .../opencode/models/gemini-3.8-flash.toml     | 14 +++++
 .../muse-spark-1.3-contributor-free.toml      | 13 +++++
 5 files changed, 123 insertions(+), 1 deletion(-)
 create mode 100644 models/meta/muse-spark-1.3.toml
 create mode 100644 providers/opencode-go/models/muse-spark-1.3-contributor.toml
 create mode 100644 providers/opencode/models/gemini-3.8-flash.toml
 create mode 100644 providers/opencode/models/muse-spark-1.3-contributor-free.toml

diff --git a/models/google/gemini-3.8-flash.toml b/models/google/gemini-3.8-flash.toml
index c568d5402c1..480ef0f8d3e 100644
--- a/models/google/gemini-3.8-flash.toml
+++ b/models/google/gemini-3.8-flash.toml
@@ -1,6 +1,8 @@
 # Sources:
 # - https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
 # - https://ai.google.dev/gemini-api/docs/latest-model
+# - https://deepmind.google/models/model-cards/gemini-3-8-flash/
+# - https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash
 name = "Gemini 3.8 Flash"
 description = "Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows"
 family = "gemini-flash"
@@ -8,9 +10,10 @@ release_date = "2026-09-02"
 last_updated = "2026-09-02"
 attachment = true
 reasoning = true
-temperature = true
+temperature = false
 tool_call = true
 structured_output = true
+knowledge = "2026-03"
 open_weights = false
 
 [limit]
@@ -20,3 +23,50 @@ output = 65_536
 [modalities]
 input = ["text", "image", "video", "audio", "pdf"]
 output = ["text"]
+
+[[links]]
+label = "Gemini 3.8 Flash model card"
+url = "https://deepmind.google/models/model-cards/gemini-3-8-flash/"
+type = "model_card"
+
+[[links]]
+label = "Gemini 3.8 Flash developer guide"
+url = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+type = "docs"
+
+[[links]]
+label = "Gemini 3.8 Flash specifications"
+url = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-8-flash"
+type = "docs"
+
+[[benchmarks]]
+name = "Terminal-Bench"
+score = 90.8
+version = "2.1"
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+
+[[benchmarks]]
+name = "SWE-Bench Pro"
+score = 61.6
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+
+[[benchmarks]]
+name = "SWE-Atlas"
+score = 51.9
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+
+[[benchmarks]]
+name = "CharXiv"
+score = 86.2
+variant = "Multimodal"
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+
+[[benchmarks]]
+name = "GDP.pdf"
+score = 35.0
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
+
+[[benchmarks]]
+name = "Humanity's Last Exam"
+score = 45.4
+source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
diff --git a/models/meta/muse-spark-1.3.toml b/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..dade6a3814e
--- /dev/null
+++ b/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,33 @@
+# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
+# Dates reflect first confirmed availability, not a public Meta announcement.
+# References for the preceding model and shared API:
+# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
+# https://dev.meta.ai/docs/models
+
+name = "Muse Spark 1.3"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
+
+[[links]]
+label = "Muse Spark 1.2 reference"
+url = "https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2"
+type = "blog"
+
+[[links]]
+label = "Meta model documentation"
+url = "https://dev.meta.ai/docs/models"
+type = "docs"
diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..4c14157ed9b
--- /dev/null
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -0,0 +1,12 @@
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 Contributor"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[provider]
+npm = "@ai-sdk/openai"
diff --git a/providers/opencode/models/gemini-3.8-flash.toml b/providers/opencode/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..2f995cd4138
--- /dev/null
+++ b/providers/opencode/models/gemini-3.8-flash.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.50
+output = 7.50
+cache_read = 0.15
+input_audio = 1.50
+
+[provider]
+npm = "@ai-sdk/google"
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
new file mode 100644
index 00000000000..d9cf351e4f3
--- /dev/null
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -0,0 +1,13 @@
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 Free"
+family = "muse-free"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[provider]
+npm = "@ai-sdk/openai"

From c062fb37eebad42c92a1dd7b7bec8f8bb1f08ff6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:29:56 +0000
Subject: [PATCH 0475/2012] chore(sync): update Cortecs model catalog (#6112)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/deepseek-v4-pro-0813.toml | 7 +++++--
 providers/cortecs/models/devstral-2512.toml        | 2 +-
 providers/cortecs/models/glm-5.3.toml              | 2 +-
 providers/cortecs/models/minimax-m2.5.toml         | 2 +-
 4 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/providers/cortecs/models/deepseek-v4-pro-0813.toml b/providers/cortecs/models/deepseek-v4-pro-0813.toml
index 181ae5d1396..70e5bf74363 100644
--- a/providers/cortecs/models/deepseek-v4-pro-0813.toml
+++ b/providers/cortecs/models/deepseek-v4-pro-0813.toml
@@ -1,5 +1,8 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 2
@@ -8,4 +11,4 @@ cache_read = 0.5
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 64_000
diff --git a/providers/cortecs/models/devstral-2512.toml b/providers/cortecs/models/devstral-2512.toml
index 43fb308c802..4154947ed95 100644
--- a/providers/cortecs/models/devstral-2512.toml
+++ b/providers/cortecs/models/devstral-2512.toml
@@ -9,4 +9,4 @@ cache_read = 0.045
 
 [limit]
 context = 256_000
-output = 262_000
+output = 256_000
diff --git a/providers/cortecs/models/glm-5.3.toml b/providers/cortecs/models/glm-5.3.toml
index 95ee62cddee..d2aacc59483 100644
--- a/providers/cortecs/models/glm-5.3.toml
+++ b/providers/cortecs/models/glm-5.3.toml
@@ -15,4 +15,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 64_000
+output = 1_048_576
diff --git a/providers/cortecs/models/minimax-m2.5.toml b/providers/cortecs/models/minimax-m2.5.toml
index abf22f19643..af074e5dfe9 100644
--- a/providers/cortecs/models/minimax-m2.5.toml
+++ b/providers/cortecs/models/minimax-m2.5.toml
@@ -13,4 +13,4 @@ cache_read = 0.075
 
 [limit]
 context = 196_000
-output = 196_608
+output = 196_000

From fdd407adb7ee46b5c1fed285c74df6a5e190934c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:30:04 +0000
Subject: [PATCH 0476/2012] chore(sync): update NanoGPT model catalog (#6109)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/gemini-3.7-flash.toml       |  8 +++---
 .../models/google/gemini-3.8-flash.toml       | 25 ++++---------------
 .../models/google/gemini-flash-latest.toml    |  8 +++---
 3 files changed, 13 insertions(+), 28 deletions(-)

diff --git a/providers/nano-gpt/models/google/gemini-3.7-flash.toml b/providers/nano-gpt/models/google/gemini-3.7-flash.toml
index e6ceafdd5d1..4ccc6179c15 100644
--- a/providers/nano-gpt/models/google/gemini-3.7-flash.toml
+++ b/providers/nano-gpt/models/google/gemini-3.7-flash.toml
@@ -5,10 +5,10 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.075
 
 [limit]
 input = 1_048_576
diff --git a/providers/nano-gpt/models/google/gemini-3.8-flash.toml b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
index 27ce1105456..2ff854266fd 100644
--- a/providers/nano-gpt/models/google/gemini-3.8-flash.toml
+++ b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
@@ -1,29 +1,14 @@
-name = "Gemini 3.8 Flash"
-description = "Google's fast multimodal model for agentic workloads, including coding, tool use, image understanding, PDF and document extraction, audio, and video. Its capabilities, limits, reasoning behavior, and provisional pricing currently mirror Gemini 3.7 Flash."
-family = "gemini"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
+base_model = "google/gemini-3.8-flash"
 
 [[reasoning_options]]
 type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.075
 
 [limit]
-context = 1_048_576
 input = 1_048_576
-output = 65_536
-
-[modalities]
-input = ["text", "image", "video", "audio", "pdf"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemini-flash-latest.toml b/providers/nano-gpt/models/google/gemini-flash-latest.toml
index 0cc3c5e9717..1060d3dd657 100644
--- a/providers/nano-gpt/models/google/gemini-flash-latest.toml
+++ b/providers/nano-gpt/models/google/gemini-flash-latest.toml
@@ -7,10 +7,10 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.375
-output = 1.875
-cache_read = 0.0375
-cache_write = 0.020833
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.075
 
 [limit]
 input = 1_048_576

From 9a67702f33b26032db635183b33ae4ec7c7b94b2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:30:25 +0000
Subject: [PATCH 0477/2012] chore(sync): update Kilo model catalog (#6108)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/google/gemini-3.8-flash.toml  | 29 ++++---------------
 providers/kilo/models/tencent/hy3.toml        |  6 ++--
 2 files changed, 9 insertions(+), 26 deletions(-)

diff --git a/providers/kilo/models/google/gemini-3.8-flash.toml b/providers/kilo/models/google/gemini-3.8-flash.toml
index 27385be8262..35a21fd5441 100644
--- a/providers/kilo/models/google/gemini-3.8-flash.toml
+++ b/providers/kilo/models/google/gemini-3.8-flash.toml
@@ -1,30 +1,13 @@
-name = "Google: Gemini 3.8 Flash"
+base_model = "google/gemini-3.8-flash"
 description = "Gemini 3.8 Flash is Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows."
-family = "gemini"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-
-[limit]
-context = 1_048_576
-output = 65_536
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 18843e6fee9..d43287f7c3c 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 420877e3ebe3fc91f3666e804bd26f4df821b9c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:30:29 +0000
Subject: [PATCH 0478/2012] chore(sync): update OpenRouter model catalog
 (#6111)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/google/gemini-3.8-flash.toml       | 29 ++++---------------
 providers/openrouter/models/tencent/hy3.toml  |  6 ++--
 4 files changed, 15 insertions(+), 32 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b239680e290..07235ccdb08 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.034256
-output = 2.068512
-cache_read = 0.086188
+input = 1.030776
+output = 2.061552
+cache_read = 0.085898
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemini-3.8-flash.toml b/providers/openrouter/models/google/gemini-3.8-flash.toml
index 20d17832d30..ead931aabb0 100644
--- a/providers/openrouter/models/google/gemini-3.8-flash.toml
+++ b/providers/openrouter/models/google/gemini-3.8-flash.toml
@@ -1,30 +1,13 @@
-name = "Gemini 3.8 Flash"
+base_model = "google/gemini-3.8-flash"
 description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
-family = "gemini"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-
-[limit]
-context = 1_048_576
-output = 65_536
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 4febc56bf89..3b9b7ca7964 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 3d19dd8c64c349add7c63e15c74dcbe1894077c8 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 3 Sep 2026 00:31:33 +0800
Subject: [PATCH 0479/2012] refactor(opencode): keep launch metadata in
 provider entries

---
 models/google/gemini-3.8-flash.toml           | 52 +------------------
 models/meta/muse-spark-1.3.toml               | 33 ------------
 .../models/muse-spark-1.3-contributor.toml    | 21 +++++++-
 .../muse-spark-1.3-contributor-free.toml      | 20 ++++++-
 4 files changed, 40 insertions(+), 86 deletions(-)
 delete mode 100644 models/meta/muse-spark-1.3.toml

diff --git a/models/google/gemini-3.8-flash.toml b/models/google/gemini-3.8-flash.toml
index 480ef0f8d3e..c568d5402c1 100644
--- a/models/google/gemini-3.8-flash.toml
+++ b/models/google/gemini-3.8-flash.toml
@@ -1,8 +1,6 @@
 # Sources:
 # - https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
 # - https://ai.google.dev/gemini-api/docs/latest-model
-# - https://deepmind.google/models/model-cards/gemini-3-8-flash/
-# - https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash
 name = "Gemini 3.8 Flash"
 description = "Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows"
 family = "gemini-flash"
@@ -10,10 +8,9 @@ release_date = "2026-09-02"
 last_updated = "2026-09-02"
 attachment = true
 reasoning = true
-temperature = false
+temperature = true
 tool_call = true
 structured_output = true
-knowledge = "2026-03"
 open_weights = false
 
 [limit]
@@ -23,50 +20,3 @@ output = 65_536
 [modalities]
 input = ["text", "image", "video", "audio", "pdf"]
 output = ["text"]
-
-[[links]]
-label = "Gemini 3.8 Flash model card"
-url = "https://deepmind.google/models/model-cards/gemini-3-8-flash/"
-type = "model_card"
-
-[[links]]
-label = "Gemini 3.8 Flash developer guide"
-url = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-type = "docs"
-
-[[links]]
-label = "Gemini 3.8 Flash specifications"
-url = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-8-flash"
-type = "docs"
-
-[[benchmarks]]
-name = "Terminal-Bench"
-score = 90.8
-version = "2.1"
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-
-[[benchmarks]]
-name = "SWE-Bench Pro"
-score = 61.6
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-
-[[benchmarks]]
-name = "SWE-Atlas"
-score = 51.9
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-
-[[benchmarks]]
-name = "CharXiv"
-score = 86.2
-variant = "Multimodal"
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-
-[[benchmarks]]
-name = "GDP.pdf"
-score = 35.0
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
-
-[[benchmarks]]
-name = "Humanity's Last Exam"
-score = 45.4
-source = "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-3-8-flash"
diff --git a/models/meta/muse-spark-1.3.toml b/models/meta/muse-spark-1.3.toml
deleted file mode 100644
index dade6a3814e..00000000000
--- a/models/meta/muse-spark-1.3.toml
+++ /dev/null
@@ -1,33 +0,0 @@
-# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
-# Dates reflect first confirmed availability, not a public Meta announcement.
-# References for the preceding model and shared API:
-# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
-# https://dev.meta.ai/docs/models
-
-name = "Muse Spark 1.3"
-description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
-
-[[links]]
-label = "Muse Spark 1.2 reference"
-url = "https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2"
-type = "blog"
-
-[[links]]
-label = "Meta model documentation"
-url = "https://dev.meta.ai/docs/models"
-type = "docs"
diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
index 4c14157ed9b..4350c4496ee 100644
--- a/providers/opencode-go/models/muse-spark-1.3-contributor.toml
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -1,8 +1,27 @@
-base_model = "meta/muse-spark-1.3"
+# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
+# Dates reflect first confirmed availability, not a public Meta announcement.
+# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
+# https://dev.meta.ai/docs/models
 name = "Muse Spark 1.3 Contributor"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
+
 [cost]
 input = 0.1
 output = 0.2
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
index d9cf351e4f3..9db748eb3da 100644
--- a/providers/opencode/models/muse-spark-1.3-contributor-free.toml
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -1,9 +1,27 @@
-base_model = "meta/muse-spark-1.3"
+# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
+# Dates reflect first confirmed availability, not a public Meta announcement.
+# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
+# https://dev.meta.ai/docs/models
 name = "Muse Spark 1.3 Free"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
 family = "muse-free"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
+
 [cost]
 input = 0
 output = 0

From d27e8f571dc80f772c72ce336d6785f485c37825 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:31:43 +0000
Subject: [PATCH 0480/2012] chore(sync): update Merge Gateway model catalog
 (#6110)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../merge-gateway/models/google/gemini-3.8-flash.toml      | 7 +++++++
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 4 files changed, 17 insertions(+), 11 deletions(-)
 create mode 100644 providers/merge-gateway/models/google/gemini-3.8-flash.toml

diff --git a/providers/merge-gateway/models/google/gemini-3.8-flash.toml b/providers/merge-gateway/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..c151d3854a2
--- /dev/null
+++ b/providers/merge-gateway/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemini-3.8-flash"
+reasoning_options = []
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index c680bd0f847..16d15cbb411 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index faa70b7c1e0..3ef83d21d9d 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index b4bdd294444..0a7a2254ee8 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From e831b6b8d344be28f59c6eb3215939132016dfeb Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 3 Sep 2026 00:40:39 +0800
Subject: [PATCH 0481/2012] fix(opencode): align Muse Spark 1.3 capabilities
 with 1.2

---
 providers/opencode-go/models/muse-spark-1.3-contributor.toml   | 2 ++
 providers/opencode/models/muse-spark-1.3-contributor-free.toml | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
index 4350c4496ee..df9405a4823 100644
--- a/providers/opencode-go/models/muse-spark-1.3-contributor.toml
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -9,7 +9,9 @@ release_date = "2026-09-02"
 last_updated = "2026-09-02"
 attachment = true
 reasoning = true
+temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
index 9db748eb3da..829fdeae0b1 100644
--- a/providers/opencode/models/muse-spark-1.3-contributor-free.toml
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -9,7 +9,9 @@ release_date = "2026-09-02"
 last_updated = "2026-09-02"
 attachment = true
 reasoning = true
+temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]

From e73f6fd2eacd84f92e6da3aae7b0be87d22fa856 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 11:43:25 -0500
Subject: [PATCH 0482/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#6099)

* chore(sync): update OVHcloud AI Endpoints model catalog

* fix(ovhcloud): correct Qwen3.8 reasoning options

* fix(ovhcloud): emit zero costs when pricing is absent

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/ovhcloud.ts | 15 +++---
 packages/core/test/ovhcloud.test.ts          | 55 ++++++++++++++++++++
 providers/ovhcloud/models/qwen3.8-27b.toml   | 22 ++++++++
 3 files changed, 83 insertions(+), 9 deletions(-)
 create mode 100644 packages/core/test/ovhcloud.test.ts
 create mode 100644 providers/ovhcloud/models/qwen3.8-27b.toml

diff --git a/packages/core/src/sync/providers/ovhcloud.ts b/packages/core/src/sync/providers/ovhcloud.ts
index a7ff6eee316..60b1446e65e 100644
--- a/packages/core/src/sync/providers/ovhcloud.ts
+++ b/packages/core/src/sync/providers/ovhcloud.ts
@@ -103,15 +103,12 @@ export function buildOvhcloudModel(
   const outputCost = price(model.pricing?.completion);
   const cacheRead = price(model.pricing?.input_cache_reads);
   const cacheWrite = price(model.pricing?.input_cache_writes);
-  const cost =
-    (inputCost ?? 0) > 0 || (outputCost ?? 0) > 0
-      ? {
-          input: inputCost ?? 0,
-          output: outputCost ?? 0,
-          cache_read: cacheRead !== undefined && cacheRead > 0 ? cacheRead : undefined,
-          cache_write: cacheWrite !== undefined && cacheWrite > 0 ? cacheWrite : undefined,
-        }
-      : undefined;
+  const cost = {
+    input: inputCost ?? 0,
+    output: outputCost ?? 0,
+    cache_read: cacheRead !== undefined && cacheRead > 0 ? cacheRead : undefined,
+    cache_write: cacheWrite !== undefined && cacheWrite > 0 ? cacheWrite : undefined,
+  };
 
   return {
     base_model: existing?.base_model,
diff --git a/packages/core/test/ovhcloud.test.ts b/packages/core/test/ovhcloud.test.ts
new file mode 100644
index 00000000000..f12cc932898
--- /dev/null
+++ b/packages/core/test/ovhcloud.test.ts
@@ -0,0 +1,55 @@
+import { expect, test } from "bun:test";
+
+import { formatToml } from "../src/sync/index.js";
+import { buildOvhcloudModel, type OvhcloudModel } from "../src/sync/providers/ovhcloud.js";
+
+function model(pricing?: OvhcloudModel["pricing"]): OvhcloudModel {
+  return {
+    id: "example-model",
+    name: "Example Model",
+    created: Date.parse("2026-09-02T00:00:00Z") / 1_000,
+    context_length: 262_144,
+    pricing,
+  };
+}
+
+test.each([
+  { label: "absent pricing", pricing: undefined },
+  { label: "empty pricing", pricing: {} },
+  { label: "blank rates", pricing: { prompt: "", completion: "" } },
+  { label: "whitespace rates", pricing: { prompt: " ", completion: "\t" } },
+  { label: "explicit zero rates", pricing: { prompt: "0", completion: "0" } },
+])("OVHcloud serializes zero costs for $label", ({ pricing }) => {
+  const result = buildOvhcloudModel(model(pricing), undefined);
+  const content = formatToml({ id: "example-model", ...result });
+
+  expect(Bun.TOML.parse(content).cost).toEqual({ input: 0, output: 0 });
+});
+
+test("OVHcloud converts paid and cache rates to per-million costs", () => {
+  const result = buildOvhcloudModel(model({
+    prompt: "0.00000047",
+    completion: "0.00000319",
+    input_cache_reads: "0.00000009",
+    input_cache_writes: "0.00000012",
+  }), undefined);
+
+  expect(result.cost).toEqual({ input: 0.47, output: 3.19, cache_read: 0.09, cache_write: 0.12 });
+});
+
+test.each([
+  { pricing: { prompt: "0.00000047" }, expected: { input: 0.47, output: 0 } },
+  { pricing: { completion: "0.00000319" }, expected: { input: 0, output: 3.19 } },
+])("OVHcloud preserves a supplied rate when the other is missing: %j", ({ pricing, expected }) => {
+  const result = buildOvhcloudModel(model(pricing), undefined);
+  const content = formatToml({ id: "example-model", ...result });
+
+  expect(Bun.TOML.parse(content).cost).toEqual(expected);
+});
+
+test("OVHcloud replaces prior costs with zeros when pricing is absent", () => {
+  const result = buildOvhcloudModel(model(), { cost: { input: 0.47, output: 3.19 } });
+  const content = formatToml({ id: "example-model", ...result });
+
+  expect(Bun.TOML.parse(content).cost).toEqual({ input: 0, output: 0 });
+});
diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
new file mode 100644
index 00000000000..396aa69ac92
--- /dev/null
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -0,0 +1,22 @@
+# Sources (accessed 2026-09-02):
+# https://oai.endpoints.kepler.ai.cloud.ovh.net/doc/Qwen3.8-27B/openapi.json
+# https://catalog.endpoints.ai.ovh.net/rest/v2/openrouter
+# Effort: reasoning_effort = none|minimal|low|medium|high (OVHcloud's API enum,
+# not the lab's low|medium|xhigh). Responses uses reasoning.effort.
+# Blank catalog prompt/completion rates default to zero in the sync.
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8-27B"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+output = 262_144
+
+[modalities]
+input = ["text", "image"]

From eda91d0a0bdca5001f5aefbd5f62084de885b4af Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 3 Sep 2026 00:54:24 +0800
Subject: [PATCH 0483/2012] deprecate muse-spark-1.3

---
 providers/opencode-go/models/muse-spark-1.3-contributor.toml   | 1 +
 providers/opencode/models/muse-spark-1.3-contributor-free.toml | 1 +
 2 files changed, 2 insertions(+)

diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
index df9405a4823..d3878e3fcf4 100644
--- a/providers/opencode-go/models/muse-spark-1.3-contributor.toml
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -13,6 +13,7 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
+status = "deprecated"
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
index 829fdeae0b1..51a294ffc00 100644
--- a/providers/opencode/models/muse-spark-1.3-contributor-free.toml
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -13,6 +13,7 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
+status = "deprecated"
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 

From 04a4343c336d50c8b240033cf21f0013c09fbcbe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:24:29 +0000
Subject: [PATCH 0484/2012] chore(sync): update OpenRouter model catalog
 (#6116)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576

From e32e8d9fec2914046f5f0e1109fd24c07d539843 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:24:39 +0000
Subject: [PATCH 0485/2012] chore(sync): update NanoGPT model catalog (#6115)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/openai/o1-preview.toml    | 28 -------------------
 1 file changed, 28 deletions(-)
 delete mode 100644 providers/nano-gpt/models/openai/o1-preview.toml

diff --git a/providers/nano-gpt/models/openai/o1-preview.toml b/providers/nano-gpt/models/openai/o1-preview.toml
deleted file mode 100644
index 21b58841092..00000000000
--- a/providers/nano-gpt/models/openai/o1-preview.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "OpenAI o1-preview"
-description = "O-series reasoning model for hard analysis, math, coding, and planning"
-family = "o"
-release_date = "2024-01-01"
-last_updated = "2024-09-12"
-attachment = false
-reasoning = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 15
-output = 60
-cache_read = 7.5
-
-[limit]
-context = 128_000
-input = 128_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 339e1a8a8d0a08df5cc9d598bf1ea65b35e6c10a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:34:39 +0000
Subject: [PATCH 0486/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#6114)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ovhcloud/models/qwen3.8-27b.toml         | 14 +++++++++++++-
 providers/ovhcloud/models/qwen3guard-gen-0.6b.toml |  4 ++++
 providers/ovhcloud/models/qwen3guard-gen-8b.toml   |  4 ++++
 3 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 396aa69ac92..45ae7920a40 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -7,16 +7,28 @@
 base_model = "alibaba/qwen3.8-27b"
 name = "Qwen3.8-27B"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "qwen"
 release_date = "2026-09-02"
 last_updated = "2026-09-02"
-reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0
 output = 0
 
 [limit]
+context = 262_144
 output = 262_144
 
 [modalities]
 input = ["text", "image"]
+output = ["text"]
diff --git a/providers/ovhcloud/models/qwen3guard-gen-0.6b.toml b/providers/ovhcloud/models/qwen3guard-gen-0.6b.toml
index 6e1f4318b3b..46c9923c1db 100644
--- a/providers/ovhcloud/models/qwen3guard-gen-0.6b.toml
+++ b/providers/ovhcloud/models/qwen3guard-gen-0.6b.toml
@@ -8,6 +8,10 @@ temperature = true
 tool_call = false
 open_weights = true
 
+[cost]
+input = 0
+output = 0
+
 [limit]
 context = 32_768
 output = 16_384
diff --git a/providers/ovhcloud/models/qwen3guard-gen-8b.toml b/providers/ovhcloud/models/qwen3guard-gen-8b.toml
index b26f56bc363..f7f4eddbb8b 100644
--- a/providers/ovhcloud/models/qwen3guard-gen-8b.toml
+++ b/providers/ovhcloud/models/qwen3guard-gen-8b.toml
@@ -8,6 +8,10 @@ temperature = true
 tool_call = false
 open_weights = true
 
+[cost]
+input = 0
+output = 0
+
 [limit]
 context = 32_768
 output = 16_384

From b467f1764015774d84c228559d86cb2c53f59ed2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 18:32:51 +0000
Subject: [PATCH 0487/2012] chore(sync): update OpenRouter model catalog
 (#6118)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576

From 7b5ab02ddc79eccdea4afd60f41f78bf25199b9e Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Wed, 2 Sep 2026 14:53:03 -0400
Subject: [PATCH 0488/2012] fix model naming

---
 .../models/flexai/DeepSeek-V4-Flash-0731.toml      | 14 --------------
 .../models/flexai/deepseek-v4-flash-0731.toml      | 11 -----------
 2 files changed, 25 deletions(-)
 delete mode 100644 providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
 delete mode 100644 providers/edenai/models/flexai/deepseek-v4-flash-0731.toml

diff --git a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
deleted file mode 100644
index 55b4248ec19..00000000000
--- a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-0731"
-tool_call = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0.08
-output = 0.18
-
-[limit]
-context = 786_432
\ No newline at end of file
diff --git a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml b/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
deleted file mode 100644
index 57e0752af57..00000000000
--- a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-0731"
-tool_call = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0.08
-output = 0.18

From a1e288b829e14e3b89d2f80fe21fc1baefe5fee4 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Wed, 2 Sep 2026 14:54:16 -0400
Subject: [PATCH 0489/2012] update zen models

---
 .../models/muse-spark-1.3-contributor.toml    | 34 -------------------
 .../muse-spark-1.3-contributor-free.toml      | 34 -------------------
 2 files changed, 68 deletions(-)
 delete mode 100644 providers/opencode-go/models/muse-spark-1.3-contributor.toml
 delete mode 100644 providers/opencode/models/muse-spark-1.3-contributor-free.toml

diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
deleted file mode 100644
index d3878e3fcf4..00000000000
--- a/providers/opencode-go/models/muse-spark-1.3-contributor.toml
+++ /dev/null
@@ -1,34 +0,0 @@
-# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
-# Dates reflect first confirmed availability, not a public Meta announcement.
-# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
-# https://dev.meta.ai/docs/models
-name = "Muse Spark 1.3 Contributor"
-description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-status = "deprecated"
-
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.002
-
-[provider]
-npm = "@ai-sdk/openai"
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
deleted file mode 100644
index 51a294ffc00..00000000000
--- a/providers/opencode/models/muse-spark-1.3-contributor-free.toml
+++ /dev/null
@@ -1,34 +0,0 @@
-# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
-# Dates reflect first confirmed availability, not a public Meta announcement.
-# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
-# https://dev.meta.ai/docs/models
-name = "Muse Spark 1.3 Free"
-description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
-family = "muse-free"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-status = "deprecated"
-
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[provider]
-npm = "@ai-sdk/openai"

From ee9c7735300771f4c4cb0bfcecf6a7a2291ba154 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 19:23:08 +0000
Subject: [PATCH 0490/2012] chore(sync): update OpenRouter model catalog
 (#6119)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576

From 657801b2ad76eb8d3c8931b08da1075777f05445 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Wed, 2 Sep 2026 15:25:57 -0400
Subject: [PATCH 0491/2012] Revert "update zen models"

This reverts commit a1e288b829e14e3b89d2f80fe21fc1baefe5fee4.
---
 .../models/muse-spark-1.3-contributor.toml    | 34 +++++++++++++++++++
 .../muse-spark-1.3-contributor-free.toml      | 34 +++++++++++++++++++
 2 files changed, 68 insertions(+)
 create mode 100644 providers/opencode-go/models/muse-spark-1.3-contributor.toml
 create mode 100644 providers/opencode/models/muse-spark-1.3-contributor-free.toml

diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..d3878e3fcf4
--- /dev/null
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -0,0 +1,34 @@
+# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
+# Dates reflect first confirmed availability, not a public Meta announcement.
+# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
+# https://dev.meta.ai/docs/models
+name = "Muse Spark 1.3 Contributor"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+status = "deprecated"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[provider]
+npm = "@ai-sdk/openai"
diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
new file mode 100644
index 00000000000..51a294ffc00
--- /dev/null
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -0,0 +1,34 @@
+# Capability metadata follows Muse Spark 1.2 pending public 1.3 specifications.
+# Dates reflect first confirmed availability, not a public Meta announcement.
+# https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
+# https://dev.meta.ai/docs/models
+name = "Muse Spark 1.3 Free"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for coding and agentic workflows."
+family = "muse-free"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+status = "deprecated"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[provider]
+npm = "@ai-sdk/openai"

From 25055fcdcfe5a50ddb5ab4088f73eea1d7027b2c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 20:27:05 +0000
Subject: [PATCH 0492/2012] chore(sync): update Venice model catalog (#6122)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/gemini-3-8-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/venice/models/gemini-3-8-flash.toml

diff --git a/providers/venice/models/gemini-3-8-flash.toml b/providers/venice/models/gemini-3-8-flash.toml
new file mode 100644
index 00000000000..7c878c334ad
--- /dev/null
+++ b/providers/venice/models/gemini-3-8-flash.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.8-flash"
+description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
+reasoning_options = []
+
+[cost]
+input = 0.9375
+output = 4.6875
+cache_read = 0.09375
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image", "audio", "video"]

From 43b7381a5e8de2bfa252b20e09897ee29ef9de23 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 20:27:39 +0000
Subject: [PATCH 0493/2012] chore(sync): update OpenRouter model catalog
 (#6125)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../meta/muse-spark-1.3-contributor.toml      | 28 +++++++++++++++++++
 .../models/meta/muse-spark-1.3.toml           | 28 +++++++++++++++++++
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    |  8 +++---
 4 files changed, 63 insertions(+), 7 deletions(-)
 create mode 100644 providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
 create mode 100644 providers/openrouter/models/meta/muse-spark-1.3.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 07235ccdb08..92d05b69dd5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.030776
-output = 2.061552
-cache_read = 0.085898
+input = 1.02747
+output = 2.05494
+cache_read = 0.085623
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml b/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..747c8a17772
--- /dev/null
+++ b/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3 Contributor"
+description = "Open Llama multimodal model for image understanding and text reasoning"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
diff --git a/providers/openrouter/models/meta/muse-spark-1.3.toml b/providers/openrouter/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..1349149c70b
--- /dev/null
+++ b/providers/openrouter/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3"
+description = "Open Llama multimodal model for image understanding and text reasoning"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520

From 8efaa670e4c10aac2eada6e35d9f29ae42341511 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 20:27:44 +0000
Subject: [PATCH 0494/2012] chore(sync): update Charm Hyper model catalog
 (#6123)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/kimi-k2-thinking.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/hyper/models/kimi-k2-thinking.toml

diff --git a/providers/hyper/models/kimi-k2-thinking.toml b/providers/hyper/models/kimi-k2-thinking.toml
new file mode 100644
index 00000000000..cb77c77068c
--- /dev/null
+++ b/providers/hyper/models/kimi-k2-thinking.toml
@@ -0,0 +1,12 @@
+base_model = "moonshotai/kimi-k2-thinking"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+reasoning_options = []
+
+[cost]
+input = 0.6
+output = 2.5
+cache_write = 0.3
+
+[limit]
+output = 26_214

From 3fbb24e652e471994a66f3e4d676d4e5825b78d1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 20:27:56 +0000
Subject: [PATCH 0495/2012] chore(sync): update Kilo model catalog (#6124)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../meta/muse-spark-1.3-contributor.toml      | 28 +++++++++++++++++++
 .../kilo/models/meta/muse-spark-1.3.toml      | 28 +++++++++++++++++++
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    |  4 +--
 3 files changed, 58 insertions(+), 2 deletions(-)
 create mode 100644 providers/kilo/models/meta/muse-spark-1.3-contributor.toml
 create mode 100644 providers/kilo/models/meta/muse-spark-1.3.toml

diff --git a/providers/kilo/models/meta/muse-spark-1.3-contributor.toml b/providers/kilo/models/meta/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..04d56d4a05e
--- /dev/null
+++ b/providers/kilo/models/meta/muse-spark-1.3-contributor.toml
@@ -0,0 +1,28 @@
+name = "Meta: Muse Spark 1.3 Contributor"
+description = "Muse Spark 1.3 Contributor is the cost-efficient contributor tier of Meta’s multimodal reasoning model for experimentation, learning, and early-stage agentic, multi-agent, and coding workflows. It is designed to track information..."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
diff --git a/providers/kilo/models/meta/muse-spark-1.3.toml b/providers/kilo/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..59644af30d8
--- /dev/null
+++ b/providers/kilo/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,28 @@
+name = "Meta: Muse Spark 1.3"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It is designed to keep track of information across extended tasks, work through..."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520

From d42b90fd654e066d5aadc7648292541fe7b45f94 Mon Sep 17 00:00:00 2001
From: Filip Hejmowski <fhejmowski@simplito.com>
Date: Wed, 2 Sep 2026 22:59:20 +0200
Subject: [PATCH 0496/2012] fix(tencent): add Hy3 input token limit

---
 models/tencent/hy3.toml | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/models/tencent/hy3.toml b/models/tencent/hy3.toml
index 34f40d92c92..f1308871be2 100644
--- a/models/tencent/hy3.toml
+++ b/models/tencent/hy3.toml
@@ -1,3 +1,5 @@
+# Tencent documents a 192K maximum input for Hy3.
+# https://cloud.tencent.com/document/product/1823/130051
 name = "Hy3"
 description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
 family = "Hy"
@@ -11,6 +13,7 @@ open_weights = true
 
 [limit]
 context = 256_000
+input = 192_000
 output = 64_000
 
 [modalities]

From b215f8e0fbc02a5e9b1ac0ad0193dd75a9096b32 Mon Sep 17 00:00:00 2001
From: Filip Hejmowski <fhejmowski@simplito.com>
Date: Wed, 2 Sep 2026 23:00:08 +0200
Subject: [PATCH 0497/2012] fix(tencent): correct Hy3 output token limit

---
 models/tencent/hy3.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/models/tencent/hy3.toml b/models/tencent/hy3.toml
index f1308871be2..5fbfaacf1c1 100644
--- a/models/tencent/hy3.toml
+++ b/models/tencent/hy3.toml
@@ -1,4 +1,4 @@
-# Tencent documents a 192K maximum input for Hy3.
+# Tencent documents a 192K maximum input and 128K maximum output for Hy3.
 # https://cloud.tencent.com/document/product/1823/130051
 name = "Hy3"
 description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
@@ -14,7 +14,7 @@ open_weights = true
 [limit]
 context = 256_000
 input = 192_000
-output = 64_000
+output = 128_000
 
 [modalities]
 input = ["text"]

From 0736ee8cc5b90d8086341efff3c397424c50ec1f Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:04:06 -0500
Subject: [PATCH 0498/2012] Remove deprecated status from contributor
 configuration

Remove deprecated status from muse-spark-1.3-contributor.toml
---
 providers/opencode-go/models/muse-spark-1.3-contributor.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/opencode-go/models/muse-spark-1.3-contributor.toml b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
index d3878e3fcf4..df9405a4823 100644
--- a/providers/opencode-go/models/muse-spark-1.3-contributor.toml
+++ b/providers/opencode-go/models/muse-spark-1.3-contributor.toml
@@ -13,7 +13,6 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-status = "deprecated"
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 

From 3e3cee1f07aae149afbcf6cfcabad5b826c2d287 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Wed, 2 Sep 2026 16:05:17 -0500
Subject: [PATCH 0499/2012] Remove deprecated status field (#6129)

Remove deprecated status from muse-spark-1.3-contributor-free.toml
---
 providers/opencode/models/muse-spark-1.3-contributor-free.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/opencode/models/muse-spark-1.3-contributor-free.toml b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
index 51a294ffc00..829fdeae0b1 100644
--- a/providers/opencode/models/muse-spark-1.3-contributor-free.toml
+++ b/providers/opencode/models/muse-spark-1.3-contributor-free.toml
@@ -13,7 +13,6 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
-status = "deprecated"
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 

From 5c5e7787a094367f10d7f590deaa2baf398de6c8 Mon Sep 17 00:00:00 2001
From: Filip <34747899+neriousy@users.noreply.github.com>
Date: Wed, 2 Sep 2026 23:11:16 +0200
Subject: [PATCH 0500/2012] Update hy3.toml

---
 models/tencent/hy3.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/models/tencent/hy3.toml b/models/tencent/hy3.toml
index 5fbfaacf1c1..f5d2f84d73e 100644
--- a/models/tencent/hy3.toml
+++ b/models/tencent/hy3.toml
@@ -1,4 +1,3 @@
-# Tencent documents a 192K maximum input and 128K maximum output for Hy3.
 # https://cloud.tencent.com/document/product/1823/130051
 name = "Hy3"
 description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"

From 1f5e847cc2f490560566d11054966010a03e59b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 21:23:22 +0000
Subject: [PATCH 0501/2012] chore(sync): update OpenRouter model catalog
 (#6133)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 1 -
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 3b9b7ca7964..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -12,4 +12,3 @@ cache_read = 0.020625
 
 [limit]
 context = 262_144
-output = 128_000

From fba9ae638207c9d6ce93dab3e8b948448ee717cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 21:23:26 +0000
Subject: [PATCH 0502/2012] chore(sync): update Eden AI model catalog (#6132)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/deepinfra/tencent/Hy3.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/edenai/models/deepinfra/tencent/Hy3.toml b/providers/edenai/models/deepinfra/tencent/Hy3.toml
index 6048f17c085..ae1d1975c90 100644
--- a/providers/edenai/models/deepinfra/tencent/Hy3.toml
+++ b/providers/edenai/models/deepinfra/tencent/Hy3.toml
@@ -1,4 +1,5 @@
 base_model = "tencent/hy3"
+base_model_omit = ["limit.input"]
 structured_output = true
 
 [[reasoning_options]]

From 0eadd07c6de33bad815be03b004548d1da6313a4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 21:23:34 +0000
Subject: [PATCH 0503/2012] chore(sync): update Kilo model catalog (#6131)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/google/gemini-3-pro-image.toml | 3 ---
 providers/kilo/models/tencent/hy3.toml               | 1 -
 2 files changed, 4 deletions(-)

diff --git a/providers/kilo/models/google/gemini-3-pro-image.toml b/providers/kilo/models/google/gemini-3-pro-image.toml
index b4f1e62ee4d..13d9172697e 100644
--- a/providers/kilo/models/google/gemini-3-pro-image.toml
+++ b/providers/kilo/models/google/gemini-3-pro-image.toml
@@ -14,8 +14,5 @@ reasoning = 12
 cache_read = 0.2
 cache_write = 0.375
 
-[limit]
-context = 131_072
-
 [modalities]
 input = ["image", "text", "pdf"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index d43287f7c3c..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -13,4 +13,3 @@ cache_read = 0.020625
 
 [limit]
 context = 262_144
-output = 128_000

From b4284a8c61b3a0838c917731f8146c2cb8f8a9da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 21:24:12 +0000
Subject: [PATCH 0504/2012] chore(sync): update NanoGPT model catalog (#6134)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../meta/muse-spark-1.3-contributor.toml      | 28 +++++++++++++++++++
 .../nano-gpt/models/meta/muse-spark-1.3.toml  | 28 +++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 providers/nano-gpt/models/meta/muse-spark-1.3-contributor.toml
 create mode 100644 providers/nano-gpt/models/meta/muse-spark-1.3.toml

diff --git a/providers/nano-gpt/models/meta/muse-spark-1.3-contributor.toml b/providers/nano-gpt/models/meta/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..5bf15ba2287
--- /dev/null
+++ b/providers/nano-gpt/models/meta/muse-spark-1.3-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3 Contributor"
+description = "Meta's Muse Spark 1.3 Contributor is a frontier multimodal reasoning model for long-horizon coding and agentic workflows, with strong gains in computer use, browsing, professional tool use, codebase understanding, and million-token retrieval. It accepts text, images, audio, video, and files, supports tool calling and structured output, and always reasons before answering. Prompts and outputs may be used by Meta for training and to improve its products."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+input = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/meta/muse-spark-1.3.toml b/providers/nano-gpt/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..75ee44dfd23
--- /dev/null
+++ b/providers/nano-gpt/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3"
+description = "Meta's Muse Spark 1.3 is a frontier multimodal reasoning model for long-horizon coding and agentic workflows, with strong gains in computer use, browsing, professional tool use, codebase understanding, instruction following, and million-token retrieval. It accepts text, images, audio, video, and files, supports tool calling and structured output, and always reasons before answering."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[limit]
+context = 1_048_576
+input = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]
+output = ["text"]

From 6238115199471e694c656d92d4fe97b18365f11d Mon Sep 17 00:00:00 2001
From: Denis K <denis.kot@makersite.de>
Date: Thu, 3 Sep 2026 00:20:37 +0200
Subject: [PATCH 0505/2012] fix(providers/azure): align gpt-5.6-sol pricing
 with Azure promotional rate (#5992)

Azure OpenAI metered API pricing for GPT-5.6 Sol is reduced to
$4.00/MTok input and $20.00/MTok output, effective September 1,
2026 through at least November 30, 2026.

Source: https://azure.microsoft.com/en-us/blog/gpt-5-6-now-available-in-microsoft-foundry/
---
 providers/azure/models/gpt-5.6-sol.toml | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/azure/models/gpt-5.6-sol.toml b/providers/azure/models/gpt-5.6-sol.toml
index 1db85b549ef..2e42e9c1929 100644
--- a/providers/azure/models/gpt-5.6-sol.toml
+++ b/providers/azure/models/gpt-5.6-sol.toml
@@ -1,17 +1,18 @@
-# Pricing: https://developers.openai.com/api/docs/pricing
+# Pricing: https://azure.microsoft.com/en-us/blog/gpt-5-6-now-available-in-microsoft-foundry/
+# Azure OpenAI metered API pricing for the GPT-5.6 Sol model will be reduced to $4.00 per 1M input tokens and $20.00 per 1M output tokens (20% reduction on input pricing and 33.3% reduction on output pricing). This promotional pricing will run from September 1, 2026, through at least November 30, 2026.
 base_model = "openai/gpt-5.6-sol"
 status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
-input = 5.00
-output = 30.00
-cache_read = 0.50
-cache_write = 6.25
+input = 4.00
+output = 20.00
+cache_read = 0.40
+cache_write = 5.00
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 10.00
-output = 45.00
-cache_read = 1.00
-cache_write = 12.50
+input = 8.00
+output = 30.00
+cache_read = 0.80
+cache_write = 10.00

From 9a7b45f2b09c6bf225e3e7aeb0335270f540dec8 Mon Sep 17 00:00:00 2001
From: gabrielberget <gabriel@berget.ai>
Date: Thu, 3 Sep 2026 00:21:16 +0200
Subject: [PATCH 0506/2012] chore(berget): remove deprecated models, add
 Qwen3.8-27B (#5985)

* chore(berget): remove deprecated models, add Qwen3.8-27B

Sync berget provider with https://api.berget.ai/v1/models.

Removed (deprecated in API or absent from catalog):
- openai/gpt-oss-120b (lifecycle_state=deprecated, EOL 2026-09-11)
- zai-org/GLM-4.7 (lifecycle_state=deprecated, EOL 2026-09-04)
- meta-llama/Llama-3.3-70B-Instruct (lifecycle_state=deprecated)
- zai-org/GLM-5.3 (not in /v1/models; only GLM-5.3-Flash remains)
- mistralai/Mistral-Medium-3.5-128B (not in /v1/models)
- moonshotai/Kimi-K2.6 (not in /v1/models; superseded by Kimi-K3)

Added:
- Qwen/Qwen3.8-27B-FP8 (lifecycle_state=eval, vision=true)

Prices converted from EUR to USD at ECB rate 1.1596 (2026-08-31).
Validated with `bun validate`.

* fix(berget): drop inherited fields from Qwen3.8-27B override

Keep only real deltas from the lab entry: limit.output (65_536 vs lab
32_768) and modalities.input (text+image, no video). Drop context
(matches lab) and modalities.output (matches lab).

* fix(berget): raise GLM-5.3-Flash context to 512K

Needle-validated to 512K on the Berget B300 deployment.
---
 .../berget/models/Qwen/Qwen3.8-27B-FP8.toml   | 28 +++++++++++++++++++
 .../meta-llama/Llama-3.3-70B-Instruct.toml    | 27 ------------------
 .../mistralai/Mistral-Medium-3.5-128B.toml    | 28 -------------------
 .../berget/models/moonshotai/Kimi-K2.6.toml   | 15 ----------
 .../berget/models/openai/gpt-oss-120b.toml    | 15 ----------
 providers/berget/models/zai-org/GLM-4.7.toml  | 27 ------------------
 .../berget/models/zai-org/GLM-5.3-Flash.toml  |  6 ++--
 providers/berget/models/zai-org/GLM-5.3.toml  | 26 -----------------
 8 files changed, 31 insertions(+), 141 deletions(-)
 create mode 100644 providers/berget/models/Qwen/Qwen3.8-27B-FP8.toml
 delete mode 100644 providers/berget/models/meta-llama/Llama-3.3-70B-Instruct.toml
 delete mode 100644 providers/berget/models/mistralai/Mistral-Medium-3.5-128B.toml
 delete mode 100644 providers/berget/models/moonshotai/Kimi-K2.6.toml
 delete mode 100644 providers/berget/models/openai/gpt-oss-120b.toml
 delete mode 100644 providers/berget/models/zai-org/GLM-4.7.toml
 delete mode 100644 providers/berget/models/zai-org/GLM-5.3.toml

diff --git a/providers/berget/models/Qwen/Qwen3.8-27B-FP8.toml b/providers/berget/models/Qwen/Qwen3.8-27B-FP8.toml
new file mode 100644
index 00000000000..4fcaf5e45d7
--- /dev/null
+++ b/providers/berget/models/Qwen/Qwen3.8-27B-FP8.toml
@@ -0,0 +1,28 @@
+# Qwen3.8 27B FP8 served on Berget AI's Swedish infrastructure.
+# Multimodal (text + image), reasoning on by default with effort levels
+# none|low|medium|xhigh (matched against the live API 2026-09-01).
+# https://api.berget.ai/openapi.json
+#
+# Berget bills in EUR: 0.4 EUR/M input, 3 EUR/M output.
+# Converted to USD at 1.1596 (ECB rate, 2026-08-31):
+#   0.4 EUR/M -> 0.46 USD/M input
+#   3 EUR/M   -> 3.48 USD/M output
+base_model = "alibaba/qwen3.8-27b"
+last_updated = "2026-09-01"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.46
+output = 3.48
+
+[limit]
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/berget/models/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/berget/models/meta-llama/Llama-3.3-70B-Instruct.toml
deleted file mode 100644
index 791ddd2eb47..00000000000
--- a/providers/berget/models/meta-llama/Llama-3.3-70B-Instruct.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Llama 3.3 70B Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2025-04-27"
-last_updated = "2025-04-27"
-attachment = false
-reasoning = true
-# Berget documents no meaningful toggle, effort subset, or budget for this model.
-# https://api.berget.ai/openapi.json
-reasoning_options = []
-temperature = true
-knowledge = "2023-12"
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.99
-output = 0.99
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/berget/models/mistralai/Mistral-Medium-3.5-128B.toml b/providers/berget/models/mistralai/Mistral-Medium-3.5-128B.toml
deleted file mode 100644
index 4185041f1cd..00000000000
--- a/providers/berget/models/mistralai/Mistral-Medium-3.5-128B.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Mistral Medium 3.5 128B"
-description = "Mistral model for multilingual chat, reasoning, and tool-assisted workflows"
-family = "mistral-medium"
-release_date = "2026-04-29"
-last_updated = "2026-04-29"
-attachment = true
-reasoning = true
-# Berget's schema accepts none/low/medium/high, but documents no Mistral-specific
-# mapping; schema acceptance alone does not establish meaningful support.
-# https://api.berget.ai/openapi.json
-reasoning_options = [{ type = "effort", values = ["none", "high"] }]
-temperature = true
-knowledge = "2026-04"
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 1.65
-output = 5.5
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["image", "text"]
-output = ["text"]
diff --git a/providers/berget/models/moonshotai/Kimi-K2.6.toml b/providers/berget/models/moonshotai/Kimi-K2.6.toml
deleted file mode 100644
index 6d5e7455434..00000000000
--- a/providers/berget/models/moonshotai/Kimi-K2.6.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "moonshotai/kimi-k2.6"
-# Kimi K2.6 can disable reasoning with reasoning_effort="none" or
-# thinking.type="disabled"; enabled/adaptive are also schema-accepted.
-# https://api.berget.ai/openapi.json
-reasoning_options = [{ type = "toggle" }]
-release_date = "2026-05-07"
-last_updated = "2026-05-07"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.83
-output = 3.85
-cache_read = 0.16
diff --git a/providers/berget/models/openai/gpt-oss-120b.toml b/providers/berget/models/openai/gpt-oss-120b.toml
deleted file mode 100644
index d9c3aec2f59..00000000000
--- a/providers/berget/models/openai/gpt-oss-120b.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "openai/gpt-oss-120b"
-name = "GPT-OSS-120B"
-knowledge = "2025-08"
-# The request schema accepts reasoning_effort low/medium/high, but publishes no
-# GPT OSS-specific behavioral mapping or reasoning-token budget.
-# https://api.berget.ai/openapi.json
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-
-[cost]
-input = 0.22
-output = 0.83
-
-[limit]
-context = 128_000
-output = 8_192
diff --git a/providers/berget/models/zai-org/GLM-4.7.toml b/providers/berget/models/zai-org/GLM-4.7.toml
deleted file mode 100644
index 0e7421a344b..00000000000
--- a/providers/berget/models/zai-org/GLM-4.7.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "GLM 4.7"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-# Berget documents no meaningful toggle, effort subset, or budget for this model.
-# https://api.berget.ai/openapi.json
-reasoning_options = []
-temperature = true
-knowledge = "2025-12"
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.77
-output = 2.75
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/berget/models/zai-org/GLM-5.3-Flash.toml b/providers/berget/models/zai-org/GLM-5.3-Flash.toml
index 6550f13b057..62405e7d432 100644
--- a/providers/berget/models/zai-org/GLM-5.3-Flash.toml
+++ b/providers/berget/models/zai-org/GLM-5.3-Flash.toml
@@ -1,6 +1,6 @@
 # GLM-5.3-Flash served on Berget AI's Swedish infrastructure (NVIDIA B300,
 # SGLang TP4). Multimodal (text + image), reasoning on by default with effort
-# levels low/high/max. Needle-validated to 327K context.
+# levels low/high/max. Needle-validated to 512K context.
 # https://api.berget.ai/openapi.json
 #
 # Berget bills in EUR: 0.25 EUR/M input, 0.5 EUR/M output.
@@ -8,7 +8,7 @@
 #   0.25 EUR/M -> 0.29 USD/M input
 #   0.5 EUR/M  -> 0.58 USD/M output
 base_model = "zhipuai/glm-5.3-flash"
-last_updated = "2026-08-29"
+last_updated = "2026-09-01"
 
 [[reasoning_options]]
 type = "effort"
@@ -22,7 +22,7 @@ input = 0.29
 output = 0.58
 
 [limit]
-context = 327_680
+context = 524_288
 output = 16_384
 
 # The lab model accepts text/image/video/pdf; the Berget SGLang deployment
diff --git a/providers/berget/models/zai-org/GLM-5.3.toml b/providers/berget/models/zai-org/GLM-5.3.toml
deleted file mode 100644
index 355bc31b861..00000000000
--- a/providers/berget/models/zai-org/GLM-5.3.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-# GLM-5.3 served on Berget AI's Swedish infrastructure (NVIDIA B300, SGLang
-# TP4). Reasoning always on, effort levels low/high/max (default max), returned
-# in message.reasoning_content. Needle-validated to 327K context.
-# https://api.berget.ai/openapi.json
-#
-# Berget bills in EUR: 1.5 EUR/M input, 5.0 EUR/M output.
-# Converted to USD at 1.1643 (ECB rate, 2026-08-28):
-#   1.5 EUR/M  -> 1.75 USD/M input
-#   5.0 EUR/M  -> 5.82 USD/M output
-base_model = "zhipuai/glm-5.3"
-last_updated = "2026-08-29"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.75
-output = 5.82
-
-[limit]
-context = 327_680
-output = 32_768

From 54dc89dcf5c3105b932338b6e3a72f27c586b55e Mon Sep 17 00:00:00 2001
From: Christopher Petito <47751006+krissetto@users.noreply.github.com>
Date: Thu, 3 Sep 2026 00:21:34 +0200
Subject: [PATCH 0507/2012] deprecate old gpt models (#6088)

---
 providers/openai/models/gpt-5.2-chat-latest.toml | 1 +
 providers/openai/models/gpt-5.3-chat-latest.toml | 1 +
 2 files changed, 2 insertions(+)

diff --git a/providers/openai/models/gpt-5.2-chat-latest.toml b/providers/openai/models/gpt-5.2-chat-latest.toml
index f4508392b7a..2fd185477b9 100644
--- a/providers/openai/models/gpt-5.2-chat-latest.toml
+++ b/providers/openai/models/gpt-5.2-chat-latest.toml
@@ -11,6 +11,7 @@ knowledge = "2025-08-31"
 tool_call = true
 structured_output = true
 open_weights = false
+status = "deprecated"
 
 [cost]
 input = 1.75
diff --git a/providers/openai/models/gpt-5.3-chat-latest.toml b/providers/openai/models/gpt-5.3-chat-latest.toml
index fda276bb0d2..1f4bcb461fa 100644
--- a/providers/openai/models/gpt-5.3-chat-latest.toml
+++ b/providers/openai/models/gpt-5.3-chat-latest.toml
@@ -10,6 +10,7 @@ knowledge = "2025-08-31"
 tool_call = true
 structured_output = true
 open_weights = false
+status = "deprecated"
 
 [cost]
 input = 1.75

From 7a3b1bb2355797f7c940a3c0903b61e0c8a74c46 Mon Sep 17 00:00:00 2001
From: Bowen Liang <liang.bowen.123@qq.com>
Date: Thu, 3 Sep 2026 06:22:45 +0800
Subject: [PATCH 0508/2012] fix(kimi-for-coding): update doc URL to models.html
 (old link returned 404) (#6077)

---
 providers/kimi-for-coding/provider.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kimi-for-coding/provider.toml b/providers/kimi-for-coding/provider.toml
index 0fbd0638ba4..f8d71f33de3 100644
--- a/providers/kimi-for-coding/provider.toml
+++ b/providers/kimi-for-coding/provider.toml
@@ -6,5 +6,5 @@
 name = "Kimi For Coding"
 env = ["KIMI_API_KEY"]
 npm = "@ai-sdk/anthropic"
-doc = "https://www.kimi.com/code/docs/en/third-party-tools/other-coding-agents.html"
+doc = "https://www.kimi.com/code/docs/en/kimi-code/models.html"
 api = "https://api.kimi.com/coding/v1"

From e2adf2428a5184304ff80b23524cf9800c2fa244 Mon Sep 17 00:00:00 2001
From: Sujee Maniyam <sujee@sujee.dev>
Date: Wed, 2 Sep 2026 15:24:08 -0700
Subject: [PATCH 0509/2012] nebius: sync catalog with Token Factory, remove
 deprecated models (#5802)

* nebius: sync catalog with Token Factory, remove deprecated models

- Update context limits from the Token Factory models list
  (GLM-5.2 and DeepSeek-V4-Pro to 1M, Qwen3-30B to 262K, gpt-oss-120b /
  Hermes-4-405B to 131072, embedding model to 40960)
- Add DeepSeek-V4-Flash-0731, Kimi-K2.6, Nemotron-3-Ultra-550b-a55b,
  Nemotron-3_5-Lightning, GLM-5.1, GLM-5.3-Flash
- Remove deprecated and superseded endpoints (V3.2, K2.5, M2.5,
  fast variants, GLM-5, INTELLECT-3, Nano/Omni/253B, Qwen2.5-VL,
  Qwen3-32B, Qwen3-Next-80B, V4-Flash, Hermes-4-70B, Llama-3.3-70B)

* chore(sync): update OpenRouter model catalog (#5801)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

* nebius: address review feedback on reasoning options and limits

- Nemotron-3-Ultra / 3.5-Lightning: replace empty reasoning_options with the
  host chat_template_kwargs.enable_thinking toggle; raise limit.output to the
  1M served window
- GLM-5.3-Flash: effort values low|high|max per lab (no invented medium);
  document text-only serving and source URLs
- Qwen3-30B-A3B-Instruct-2507: align limit.input (262_144) with the new 262K context
- Drop Nebius Kimi-K2.6 and GLM-5.1 entries

* nebius: verify reasoning controls and limits live against Token Factory

- Nemotron-3-Ultra / 3.5-Lightning: replace unverified chat_template_kwargs
  toggle with the host-documented reasoning_effort control (probed live:
  none disables thinking; the enable_thinking kwarg is forwarded but
  undocumented); keep limits at the served 1M max_model_len
- DeepSeek-V4-Pro: raise limit.output to the served 1_048_576 window; effort
  none|low|high|max verified live (max distinctly heavier; mid tiers flat;
  thinking.type inert on this host)
- DeepSeek-V4-Flash-0731: document intentional 1_024_000 per-model
  max_model_len

* updated model cards according to spec requirements

* nebius: document Nemotron reasoning toggle

* nebius: inherit Nemotron Lightning description

* nebius: model Nemotron reasoning as effort

---------

Co-authored-by: opencode-agent[bot] <219766164+opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/MiniMaxAI/MiniMax-M2.5-fast.toml   | 28 -------------
 .../nebius/models/MiniMaxAI/MiniMax-M2.5.toml | 27 -------------
 .../models/NousResearch/Hermes-4-405B.toml    |  2 +-
 .../models/NousResearch/Hermes-4-70B.toml     | 31 --------------
 .../models/PrimeIntellect/INTELLECT-3.toml    | 27 -------------
 .../models/Qwen/Qwen2.5-VL-72B-Instruct.toml  | 26 ------------
 .../Qwen3-235B-A22B-Thinking-2507-fast.toml   | 28 -------------
 .../Qwen/Qwen3-30B-A3B-Instruct-2507.toml     |  4 +-
 providers/nebius/models/Qwen/Qwen3-32B.toml   | 26 ------------
 .../models/Qwen/Qwen3-Embedding-8B.toml       |  4 +-
 .../Qwen3-Next-80B-A3B-Thinking-fast.toml     | 28 -------------
 .../Qwen/Qwen3-Next-80B-A3B-Thinking.toml     | 31 --------------
 .../models/Qwen/Qwen3.5-397B-A17B-fast.toml   | 28 -------------
 .../deepseek-ai/DeepSeek-V3.2-fast.toml       | 28 -------------
 .../models/deepseek-ai/DeepSeek-V3.2.toml     | 32 ---------------
 .../deepseek-ai/DeepSeek-V4-Flash-0731.toml   | 26 ++++++++++++
 .../models/deepseek-ai/DeepSeek-V4-Flash.toml | 40 -------------------
 .../models/deepseek-ai/DeepSeek-V4-Pro.toml   | 21 ++++++++--
 .../meta-llama/Llama-3.3-70B-Instruct.toml    | 26 ------------
 .../models/moonshotai/Kimi-K2.5-fast.toml     | 32 ---------------
 .../nebius/models/moonshotai/Kimi-K2.5.toml   | 33 ---------------
 .../Llama-3_1-Nemotron-Ultra-253B-v1.toml     | 26 ------------
 .../NVIDIA-Nemotron-3-Nano-30B-A3B.toml       | 26 ------------
 .../models/nvidia/Nemotron-3-Nano-Omni.toml   | 27 -------------
 .../nvidia/Nemotron-3-Ultra-550b-a55b.toml    | 28 +++++++++++++
 .../models/nvidia/Nemotron-3_5-Lightning.toml | 27 +++++++++++++
 .../nvidia/nemotron-3-super-120b-a12b.toml    |  4 +-
 .../models/openai/gpt-oss-120b-fast.toml      | 28 -------------
 .../nebius/models/openai/gpt-oss-120b.toml    |  2 +-
 providers/nebius/models/zai-org/GLM-5.2.toml  |  4 +-
 .../nebius/models/zai-org/GLM-5.3-Flash.toml  | 35 ++++++++++++++++
 providers/nebius/models/zai-org/GLM-5.toml    | 31 --------------
 32 files changed, 143 insertions(+), 623 deletions(-)
 delete mode 100644 providers/nebius/models/MiniMaxAI/MiniMax-M2.5-fast.toml
 delete mode 100644 providers/nebius/models/MiniMaxAI/MiniMax-M2.5.toml
 delete mode 100644 providers/nebius/models/NousResearch/Hermes-4-70B.toml
 delete mode 100644 providers/nebius/models/PrimeIntellect/INTELLECT-3.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen2.5-VL-72B-Instruct.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen3-235B-A22B-Thinking-2507-fast.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen3-32B.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking-fast.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking.toml
 delete mode 100644 providers/nebius/models/Qwen/Qwen3.5-397B-A17B-fast.toml
 delete mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V3.2-fast.toml
 delete mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V3.2.toml
 create mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
 delete mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash.toml
 delete mode 100644 providers/nebius/models/meta-llama/Llama-3.3-70B-Instruct.toml
 delete mode 100644 providers/nebius/models/moonshotai/Kimi-K2.5-fast.toml
 delete mode 100644 providers/nebius/models/moonshotai/Kimi-K2.5.toml
 delete mode 100644 providers/nebius/models/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1.toml
 delete mode 100644 providers/nebius/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
 delete mode 100644 providers/nebius/models/nvidia/Nemotron-3-Nano-Omni.toml
 create mode 100644 providers/nebius/models/nvidia/Nemotron-3-Ultra-550b-a55b.toml
 create mode 100644 providers/nebius/models/nvidia/Nemotron-3_5-Lightning.toml
 delete mode 100644 providers/nebius/models/openai/gpt-oss-120b-fast.toml
 create mode 100644 providers/nebius/models/zai-org/GLM-5.3-Flash.toml
 delete mode 100644 providers/nebius/models/zai-org/GLM-5.toml

diff --git a/providers/nebius/models/MiniMaxAI/MiniMax-M2.5-fast.toml b/providers/nebius/models/MiniMaxAI/MiniMax-M2.5-fast.toml
deleted file mode 100644
index 3c3e5aefbdf..00000000000
--- a/providers/nebius/models/MiniMaxAI/MiniMax-M2.5-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "MiniMax-M2.5-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = []
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-01"
-release_date = "2025-01-20"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.30
-output = 1.20
-cache_read = 0.03
-cache_write = 0.375
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/MiniMaxAI/MiniMax-M2.5.toml b/providers/nebius/models/MiniMaxAI/MiniMax-M2.5.toml
deleted file mode 100644
index 92789b73aa7..00000000000
--- a/providers/nebius/models/MiniMaxAI/MiniMax-M2.5.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "MiniMax-M2.5"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-attachment = false
-reasoning = true
-reasoning_options = []
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-01"
-release_date = "2025-01-20"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.30
-output = 1.20
-cache_read = 0.03
-cache_write = 0.375
-
-[limit]
-context = 196_608
-input = 190_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/NousResearch/Hermes-4-405B.toml b/providers/nebius/models/NousResearch/Hermes-4-405B.toml
index 12013b7634b..41f4352bf13 100644
--- a/providers/nebius/models/NousResearch/Hermes-4-405B.toml
+++ b/providers/nebius/models/NousResearch/Hermes-4-405B.toml
@@ -19,7 +19,7 @@ cache_read = 0.10
 cache_write = 1.25
 
 [limit]
-context = 128_000
+context = 131_072
 input = 120_000
 output = 8_192
 
diff --git a/providers/nebius/models/NousResearch/Hermes-4-70B.toml b/providers/nebius/models/NousResearch/Hermes-4-70B.toml
deleted file mode 100644
index aeecd381857..00000000000
--- a/providers/nebius/models/NousResearch/Hermes-4-70B.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Hermes-4-70B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-attachment = false
-reasoning = true
-reasoning_options = []
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-11"
-release_date = "2026-01-30"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.13
-output = 0.40
-reasoning = 0.40
-cache_read = 0.013
-cache_write = 0.16
-
-[limit]
-context = 128_000
-input = 120_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/nebius/models/PrimeIntellect/INTELLECT-3.toml b/providers/nebius/models/PrimeIntellect/INTELLECT-3.toml
deleted file mode 100644
index 5d1c260122b..00000000000
--- a/providers/nebius/models/PrimeIntellect/INTELLECT-3.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "INTELLECT-3"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = false
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-10"
-release_date = "2026-01-25"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.20
-output = 1.10
-cache_read = 0.02
-cache_write = 0.25
-
-[limit]
-context = 128_000
-input = 120_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/Qwen/Qwen2.5-VL-72B-Instruct.toml b/providers/nebius/models/Qwen/Qwen2.5-VL-72B-Instruct.toml
deleted file mode 100644
index 4249901650f..00000000000
--- a/providers/nebius/models/Qwen/Qwen2.5-VL-72B-Instruct.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "Qwen2.5-VL-72B-Instruct"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-attachment = true
-reasoning = false
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2024-12"
-release_date = "2025-01-20"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.25
-output = 0.75
-cache_read = 0.025
-cache_write = 0.31
-
-[limit]
-context = 128_000
-input = 120_000
-output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/nebius/models/Qwen/Qwen3-235B-A22B-Thinking-2507-fast.toml b/providers/nebius/models/Qwen/Qwen3-235B-A22B-Thinking-2507-fast.toml
deleted file mode 100644
index f8ddac4f0c0..00000000000
--- a/providers/nebius/models/Qwen/Qwen3-235B-A22B-Thinking-2507-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Qwen3-235B-A22B-Thinking-2507-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = []
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-07"
-release_date = "2025-07-25"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.50
-output = 2.00
-cache_read = 0.05
-cache_write = 0.625
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml b/providers/nebius/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
index 2f9d2aefd72..620865c9b83 100644
--- a/providers/nebius/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
+++ b/providers/nebius/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
@@ -17,8 +17,8 @@ cache_read = 0.01
 cache_write = 0.125
 
 [limit]
-context = 128_000
-input = 120_000
+context = 262_144
+input = 262_144
 output = 8_192
 
 [modalities]
diff --git a/providers/nebius/models/Qwen/Qwen3-32B.toml b/providers/nebius/models/Qwen/Qwen3-32B.toml
deleted file mode 100644
index 30d6abf2a3d..00000000000
--- a/providers/nebius/models/Qwen/Qwen3-32B.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "Qwen3-32B"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-12"
-release_date = "2026-01-28"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.10
-output = 0.30
-cache_read = 0.01
-cache_write = 0.125
-
-[limit]
-context = 128_000
-input = 120_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/nebius/models/Qwen/Qwen3-Embedding-8B.toml b/providers/nebius/models/Qwen/Qwen3-Embedding-8B.toml
index 77091ddbda9..f75aafb1c3d 100644
--- a/providers/nebius/models/Qwen/Qwen3-Embedding-8B.toml
+++ b/providers/nebius/models/Qwen/Qwen3-Embedding-8B.toml
@@ -16,8 +16,8 @@ input = 0.01
 output = 0.00
 
 [limit]
-context = 32_768
-input = 32_768
+context = 40_960
+input = 40_960
 output = 0
 
 [modalities]
diff --git a/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking-fast.toml b/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking-fast.toml
deleted file mode 100644
index cabba5ab2fa..00000000000
--- a/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Qwen3-Next-80B-A3B-Thinking-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = []
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-07"
-release_date = "2025-07-25"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.15
-output = 1.20
-cache_read = 0.015
-cache_write = 0.1875
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking.toml b/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking.toml
deleted file mode 100644
index 241156973ec..00000000000
--- a/providers/nebius/models/Qwen/Qwen3-Next-80B-A3B-Thinking.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Qwen3-Next-80B-A3B-Thinking"
-description = "Qwen reasoning model for deliberate problem solving, math, and coding"
-attachment = false
-reasoning = true
-reasoning_options = []
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-12"
-release_date = "2026-01-28"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.15
-output = 1.20
-reasoning = 1.20
-cache_read = 0.015
-cache_write = 0.18
-
-[limit]
-context = 128_000
-input = 120_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/nebius/models/Qwen/Qwen3.5-397B-A17B-fast.toml b/providers/nebius/models/Qwen/Qwen3.5-397B-A17B-fast.toml
deleted file mode 100644
index 7ba4ee4e80a..00000000000
--- a/providers/nebius/models/Qwen/Qwen3.5-397B-A17B-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Qwen3.5-397B-A17B-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-07"
-release_date = "2025-07-15"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.60
-output = 3.60
-cache_read = 0.06
-cache_write = 0.75
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V3.2-fast.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V3.2-fast.toml
deleted file mode 100644
index f0837c4519f..00000000000
--- a/providers/nebius/models/deepseek-ai/DeepSeek-V3.2-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "DeepSeek-V3.2-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-01"
-release_date = "2025-01-27"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.40
-output = 2.00
-cache_read = 0.04
-cache_write = 0.5
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V3.2.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V3.2.toml
deleted file mode 100644
index 05ee4141350..00000000000
--- a/providers/nebius/models/deepseek-ai/DeepSeek-V3.2.toml
+++ /dev/null
@@ -1,32 +0,0 @@
-name = "DeepSeek-V3.2"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-11"
-release_date = "2026-01-20"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.30
-output = 0.45
-reasoning = 0.45
-cache_read = 0.03
-cache_write = 0.375
-
-[limit]
-context = 163_000
-input = 160_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
new file mode 100644
index 00000000000..776b5cfce81
--- /dev/null
+++ b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -0,0 +1,26 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/api/public/models_info (pricing, context, max length)
+# Accessed 2026-08-29.
+# Same gateway effort behavior as DeepSeek-V4-Flash on this host: reasoning_effort
+# accepts none/minimal/low/medium/high/xhigh/max, but minimal=low=medium and
+# high=xhigh collapse, leaving none/low/high/max as the distinct levels.
+# No discounted prompt-cache tier on Nebius: cached input is billed at the full
+# input rate, so cache_read equals input (same rationale as Kimi-K3 / V4-Flash).
+# Context/output use this model's exact models_info max_model_len (1_024_000), which
+# intentionally differs from other "1M" IDs on this host (DeepSeek-V4-Pro, GLM-5.2,
+# Nemotron entries serve max_model_len = 1_048_576).
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.14
+
+[limit]
+context = 1_024_000
+output = 1_024_000
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash.toml
deleted file mode 100644
index f91f083ce89..00000000000
--- a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Flash.toml
+++ /dev/null
@@ -1,40 +0,0 @@
-# Sources:
-# - https://tokenfactory.nebius.com/api/public/models_info (pricing, 131K context)
-# - https://tokenfactory.nebius.com/model-catalog.md
-# Accessed 2026-08-09.
-# Nebius truncates the 1M-context base model to a 131K window.
-# max_tokens has no documented ceiling other than the context window, so output
-# is set to the context size (same as the other Nebius base_model entries).
-# https://docs.tokenfactory.nebius.com/api-reference/inference/create-chat-completion
-# Effort: reasoning_effort. Verified live against
-# api.tokenfactory.nebius.com/v1/chat/completions on 2026-08-09, temperature 0,
-# two prompts. The gateway accepts all seven documented values, but only four
-# distinct behaviours exist on this model, and they line up with lab Flash's
-# low/high/max (https://api-docs.deepseek.com/guides/thinking_mode/) plus a
-# Nebius-side off switch:
-#   none                  -> reasoning_tokens 0, no reasoning_content
-#   minimal = low = medium -> identical output (677 / 133 reasoning tokens)
-#   high = xhigh          -> identical output (490 / 138 reasoning tokens)
-#   max                   -> distinct        (1102 / 154 reasoning tokens)
-# No toggle: `none` is the off switch, and DeepSeek's native
-# `thinking.type = disabled` is accepted but ignored here (reasoning_content
-# still comes back), so it is not a real caller control on this host.
-# Nebius has no discounted prompt-cache tier for this model: models_info exposes
-# no cache fields, and cached and fresh input tokens are billed identically at the
-# full input rate. cache_read is therefore set equal to input rather than left
-# unset, which downstream consumers (e.g. opencode) treat as $0/M. Same reasoning
-# as PR #3956 for Kimi-K3.
-base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.14
-
-[limit]
-context = 131_072
-output = 131_072
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro.toml
index 948f4d822dd..0957cac0a05 100644
--- a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro.toml
+++ b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro.toml
@@ -1,8 +1,17 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/api/public/models_info (pricing, context, max length)
+# Accessed 2026-08-29.
+# Effort: reasoning_effort = none|low|high|max (none = thinking off). Probed live against
+# api.tokenfactory.nebius.com chat completions on 2026-08-29: every schema effort value is
+# accepted, none disables thinking (default baseline is non-thinking), the mid tiers run
+# flat (same-family collapse note on DeepSeek-V4-Flash-0731: minimal=low=medium,
+# high=xhigh), and max is distinctly heavier. The lab's thinking.type field is accepted
+# but inert here (type=enabled does not turn thinking on), so reasoning_effort expresses off.
+# Served window: models_info max_model_len = 1_048_576 for this model with no separate
+# output cap, so context and output both take the full served window.
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["low", "medium", "high"] },
-]
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -11,3 +20,7 @@ field = "reasoning_content"
 input = 1.75
 output = 3.5
 cache_read = 0.15
+
+[limit]
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/nebius/models/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/nebius/models/meta-llama/Llama-3.3-70B-Instruct.toml
deleted file mode 100644
index 46b58975c5d..00000000000
--- a/providers/nebius/models/meta-llama/Llama-3.3-70B-Instruct.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "Llama-3.3-70B-Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-08"
-release_date = "2025-12-05"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.13
-output = 0.40
-cache_read = 0.013
-cache_write = 0.16
-
-[limit]
-context = 128_000
-input = 120_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/nebius/models/moonshotai/Kimi-K2.5-fast.toml b/providers/nebius/models/moonshotai/Kimi-K2.5-fast.toml
deleted file mode 100644
index d839ec35e0e..00000000000
--- a/providers/nebius/models/moonshotai/Kimi-K2.5-fast.toml
+++ /dev/null
@@ -1,32 +0,0 @@
-name = "Kimi-K2.5-fast"
-description = "Legacy model retained for compatibility with older integrations"
-family = "kimi-k2"
-release_date = "2025-12-15"
-last_updated = "2026-02-04"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-06"
-open_weights = true
-
-[cost]
-input = 0.50
-output = 2.50
-cache_read = 0.05
-cache_write = 0.625
-
-[limit]
-context = 256_000
-input = 256_000
-output = 8192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/nebius/models/moonshotai/Kimi-K2.5.toml b/providers/nebius/models/moonshotai/Kimi-K2.5.toml
deleted file mode 100644
index 3cc4c1e26d3..00000000000
--- a/providers/nebius/models/moonshotai/Kimi-K2.5.toml
+++ /dev/null
@@ -1,33 +0,0 @@
-name = "Kimi-K2.5"
-description = "Legacy model retained for compatibility with older integrations"
-family = "kimi-k2"
-release_date = "2025-12-15"
-last_updated = "2026-02-04"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-06"
-open_weights = true
-
-[cost]
-input = 0.50
-output = 2.50
-reasoning = 2.50
-cache_read = 0.05
-cache_write = 0.625
-
-[limit]
-context = 256_000
-input = 256_000
-output = 8192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/nebius/models/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1.toml b/providers/nebius/models/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1.toml
deleted file mode 100644
index 47b8177e2e2..00000000000
--- a/providers/nebius/models/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "Llama-3.1-Nemotron-Ultra-253B-v1"
-base_model = "nvidia/llama-3.1-nemotron-ultra-253b"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2024-12"
-release_date = "2025-01-15"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.60
-output = 1.80
-cache_read = 0.06
-cache_write = 0.75
-
-[limit]
-context = 128_000
-input = 120_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/nebius/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml b/providers/nebius/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
deleted file mode 100644
index f24f5276feb..00000000000
--- a/providers/nebius/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "Nemotron-3-Nano-30B-A3B"
-base_model = "nvidia/nemotron-3-nano-30b-a3b"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-05"
-release_date = "2025-08-10"
-last_updated = "2026-02-04"
-open_weights = true
-
-[cost]
-input = 0.06
-output = 0.24
-cache_read = 0.006
-cache_write = 0.075
-
-[limit]
-context = 32_000
-input = 30_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/nebius/models/nvidia/Nemotron-3-Nano-Omni.toml b/providers/nebius/models/nvidia/Nemotron-3-Nano-Omni.toml
deleted file mode 100644
index 114b15c2086..00000000000
--- a/providers/nebius/models/nvidia/Nemotron-3-Nano-Omni.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Nemotron-3-Nano-Omni"
-base_model = "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning"
-attachment = false
-reasoning = true
-reasoning_options = []
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-01"
-release_date = "2025-01-20"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.06
-output = 0.24
-cache_read = 0.006
-cache_write = 0.075
-
-[limit]
-context = 65_536
-input = 60_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/nvidia/Nemotron-3-Ultra-550b-a55b.toml b/providers/nebius/models/nvidia/Nemotron-3-Ultra-550b-a55b.toml
new file mode 100644
index 00000000000..1dd55593b4d
--- /dev/null
+++ b/providers/nebius/models/nvidia/Nemotron-3-Ultra-550b-a55b.toml
@@ -0,0 +1,28 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/model-catalog.md (pricing)
+# - https://tokenfactory.nebius.com/api/public/models_info (context, max length)
+# Accessed 2026-08-29.
+# Effort: reasoning_effort = none|high (none = thinking off). Probed live against
+# api.tokenfactory.nebius.com chat completions on 2026-08-29: thinking is ON by default,
+# reasoning_effort=none returns zero reasoning tokens, and high enables thinking. Other
+# accepted non-none levels did not show a distinct graded effect at probe scale, so they
+# are not exposed. The undocumented chat_template_kwargs.enable_thinking kwarg is also
+# forwarded, but reasoning_effort is the host's documented control.
+# Served window: models_info max_model_len = 1_048_576 with no separate output cap, so
+# context and output both take the full served window.
+
+base_model = "nvidia/nemotron-3-ultra-550b-a55b"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 1
+output = 3
+cache_read = 1
+
+[limit]
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/nebius/models/nvidia/Nemotron-3_5-Lightning.toml b/providers/nebius/models/nvidia/Nemotron-3_5-Lightning.toml
new file mode 100644
index 00000000000..125a90b4500
--- /dev/null
+++ b/providers/nebius/models/nvidia/Nemotron-3_5-Lightning.toml
@@ -0,0 +1,27 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/model-catalog.md (pricing)
+# - https://tokenfactory.nebius.com/api/public/models_info (context, max length)
+# Accessed 2026-08-29.
+# Effort: reasoning_effort = none|high (none = thinking off). Probed live against
+# api.tokenfactory.nebius.com chat completions on 2026-08-29: thinking is ON by default,
+# reasoning_effort=none returns zero reasoning tokens, and high enables thinking. Other
+# accepted non-none levels did not show a distinct graded effect at probe scale, so they
+# are not exposed.
+# Served window: models_info max_model_len = 1_048_576 with no separate output cap, so
+# context and output both take the full served window.
+
+base_model = "nvidia/nemotron-3.5-lightning"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.06
+output = 0.24
+cache_read = 0.06
+
+[limit]
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/nebius/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/nebius/models/nvidia/nemotron-3-super-120b-a12b.toml
index ae14965ad6d..9f3c68bd81d 100644
--- a/providers/nebius/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/nebius/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -16,8 +16,8 @@ input = 0.30
 output = 0.90
 
 [limit]
-context = 256_000
-input = 256_000
+context = 262_144
+input = 262_144
 output = 32_768
 
 [modalities]
diff --git a/providers/nebius/models/openai/gpt-oss-120b-fast.toml b/providers/nebius/models/openai/gpt-oss-120b-fast.toml
deleted file mode 100644
index b777d0e6a2b..00000000000
--- a/providers/nebius/models/openai/gpt-oss-120b-fast.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "gpt-oss-120b-fast"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2025-06"
-release_date = "2025-06-10"
-last_updated = "2026-05-07"
-open_weights = true
-
-[cost]
-input = 0.10
-output = 0.50
-cache_read = 0.01
-cache_write = 0.125
-
-[limit]
-context = 8_000
-input = 7_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nebius/models/openai/gpt-oss-120b.toml b/providers/nebius/models/openai/gpt-oss-120b.toml
index f1c2df98c95..6375d4a34d1 100644
--- a/providers/nebius/models/openai/gpt-oss-120b.toml
+++ b/providers/nebius/models/openai/gpt-oss-120b.toml
@@ -19,7 +19,7 @@ cache_read = 0.015
 cache_write = 0.18
 
 [limit]
-context = 128_000
+context = 131_072
 input = 124_000
 output = 8_192
 
diff --git a/providers/nebius/models/zai-org/GLM-5.2.toml b/providers/nebius/models/zai-org/GLM-5.2.toml
index d50ba738521..4a33cde6bde 100644
--- a/providers/nebius/models/zai-org/GLM-5.2.toml
+++ b/providers/nebius/models/zai-org/GLM-5.2.toml
@@ -12,5 +12,5 @@ input = 1.4
 output = 4.4
 
 [limit]
-context = 432_000
-output = 432_000
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/nebius/models/zai-org/GLM-5.3-Flash.toml b/providers/nebius/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..d01a5c3f2b2
--- /dev/null
+++ b/providers/nebius/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,35 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/model-catalog.md (pricing)
+# - https://tokenfactory.nebius.com/api/public/models_info (context, max length)
+# - https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash (effort set; thinking
+#   cannot be disabled, so there is no toggle — effort per lab: low|high|max, default max)
+# Accessed 2026-08-29.
+# Text-only on this host: Token Factory serves GLM-5.3-Flash as text2text with no
+# image/video use cases, unlike the multimodal lab model — attachment stays false.
+# No discounted prompt-cache tier in the public catalog (no cache rate listed for this
+# model): cached and fresh input are billed identically, so cache_read = input
+# (same rationale as Kimi-K3 / DeepSeek-V4-Flash-0731).
+
+base_model = "zhipuai/glm-5.3-flash"
+attachment = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.15
+
+[limit]
+context = 1_024_000
+output = 1_024_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nebius/models/zai-org/GLM-5.toml b/providers/nebius/models/zai-org/GLM-5.toml
deleted file mode 100644
index e0a31716e51..00000000000
--- a/providers/nebius/models/zai-org/GLM-5.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "GLM-5"
-description = "Legacy model retained for compatibility with older integrations"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-status = "deprecated"
-tool_call = true
-structured_output = true
-temperature = true
-knowledge = "2026-01"
-release_date = "2026-03-01"
-last_updated = "2026-03-10"
-open_weights = false
-
-[cost]
-input = 1.00
-output = 3.20
-cache_read = 0.10
-cache_write = 1.00
-
-[limit]
-context = 200_000
-input = 200_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

From 6207248da4ae32b05b94dcccf040e5074db2ae10 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:25:08 -0500
Subject: [PATCH 0510/2012] chore(sync): update Requesty model catalog (#6081)

* chore(sync): update Requesty model catalog

* fix(requesty): resolve dotted Claude base models

Resolve bare Claude IDs through the Anthropic matcher so dotted versions use canonical lab metadata instead of falling back to inline definitions. Factor both Fable 5.1 routes and cover regional variants and existing aliases.

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/requesty.ts  |  3 ++
 packages/core/test/requesty.test.ts           | 51 +++++++++++++++++++
 .../requesty/models/claude-fable-5.1.toml     | 15 ++++++
 .../requesty/models/claude-fable-5.1@eu.toml  | 16 ++++++
 .../models/deepseek-v4-flash-0731.toml        | 15 ++++--
 .../requesty/models/deepseek-v4-flash.toml    | 19 +++++--
 .../models/deepseek-v4-pro-0813@eu.toml       | 18 +++++++
 .../requesty/models/deepseek-v4-pro.toml      |  3 ++
 providers/requesty/models/fugu-ultra.toml     |  8 ++-
 .../requesty/models/gemini-3.8-flash.toml     | 16 ++++++
 .../requesty/models/gemini-3.8-flash@eu.toml  | 17 +++++++
 providers/requesty/models/glm-5.3-flash.toml  |  8 +--
 .../requesty/models/glm-5.3-flash@eu.toml     | 17 +++++++
 providers/requesty/models/hy3.toml            |  1 +
 providers/requesty/models/kimi-k2.7-code.toml |  1 +
 providers/requesty/models/mimo-v2.5-pro.toml  |  8 ++-
 providers/requesty/models/mimo-v2.5.toml      |  8 ++-
 .../requesty/models/qwen3.8-2.4T-A95B@eu.toml | 18 +++++++
 .../requesty/models/qwen3.8-flash-next.toml   | 16 ++++++
 .../models/qwen3.8-flash-next@eu.toml         | 17 +++++++
 providers/requesty/models/qwen3.8-flash.toml  | 17 +++++++
 21 files changed, 276 insertions(+), 16 deletions(-)
 create mode 100644 packages/core/test/requesty.test.ts
 create mode 100644 providers/requesty/models/claude-fable-5.1.toml
 create mode 100644 providers/requesty/models/claude-fable-5.1@eu.toml
 create mode 100644 providers/requesty/models/deepseek-v4-pro-0813@eu.toml
 create mode 100644 providers/requesty/models/gemini-3.8-flash.toml
 create mode 100644 providers/requesty/models/gemini-3.8-flash@eu.toml
 create mode 100644 providers/requesty/models/glm-5.3-flash@eu.toml
 create mode 100644 providers/requesty/models/qwen3.8-2.4T-A95B@eu.toml
 create mode 100644 providers/requesty/models/qwen3.8-flash-next.toml
 create mode 100644 providers/requesty/models/qwen3.8-flash-next@eu.toml
 create mode 100644 providers/requesty/models/qwen3.8-flash.toml

diff --git a/packages/core/src/sync/providers/requesty.ts b/packages/core/src/sync/providers/requesty.ts
index 7022e530406..00f6dbcc990 100644
--- a/packages/core/src/sync/providers/requesty.ts
+++ b/packages/core/src/sync/providers/requesty.ts
@@ -133,6 +133,9 @@ export function resolveRequestyBaseModel(id: string) {
     const bare = id.replace(REGION_SUFFIX, "");
     return (
         resolveModelMetadataBaseModel(bare) ??
+        (bare.startsWith("claude-")
+            ? resolveModelMetadataBaseModel(`anthropic/${bare}`)
+            : undefined) ??
         (ANTHROPIC_DOT_ZERO.test(bare)
             ? resolveModelMetadataBaseModel(`${bare}-0`)
             : undefined)
diff --git a/packages/core/test/requesty.test.ts b/packages/core/test/requesty.test.ts
new file mode 100644
index 00000000000..597b6de65c8
--- /dev/null
+++ b/packages/core/test/requesty.test.ts
@@ -0,0 +1,51 @@
+import { expect, test } from "bun:test";
+
+import { buildRequestyModel, RequestyModel, resolveRequestyBaseModel } from "../src/sync/providers/requesty.js";
+
+test.each([
+  ["claude-fable-5.1", "anthropic/claude-fable-5-1"],
+  ["claude-fable-5.1@eu", "anthropic/claude-fable-5-1"],
+  ["claude-sonnet-4.6", "anthropic/claude-sonnet-4-6"],
+  ["claude-sonnet-4", "anthropic/claude-sonnet-4-0"],
+  ["claude-opus-4-7", "anthropic/claude-opus-4-7"],
+  ["gemini-3.8-flash@eu", "google/gemini-3.8-flash"],
+  ["qwen3.8-2.4T-A95B@eu", "alibaba/qwen3.8-2.4t-a95b"],
+])("resolves Requesty %s to %s", (id, expected) => {
+  expect(resolveRequestyBaseModel(id)).toBe(expected);
+});
+
+test("does not invent a base model for unknown Claude releases", () => {
+  expect(resolveRequestyBaseModel("claude-fable-999.1@eu")).toBeUndefined();
+});
+
+test.each(["claude-fable-5.1", "claude-fable-5.1@eu"])(
+  "keeps %s override-only",
+  (id) => {
+    const model = buildRequestyModel(RequestyModel.parse({
+      id,
+      created: Date.parse("2026-09-01") / 1_000,
+      description: "Requesty description",
+      context_window: 1_000_000,
+      max_output_tokens: 128_000,
+      supports_vision: true,
+      supports_reasoning: true,
+      supports_tool_calling: true,
+      supports_output_json_schema: true,
+      input_price: 0.00001,
+      output_price: 0.00005,
+      cached_price: 0.00000025,
+      caching_price: 0.0000125,
+    }));
+
+    expect(JSON.parse(JSON.stringify(model))).toEqual({
+      base_model: "anthropic/claude-fable-5-1",
+      ...(id.endsWith("@eu") ? { name: "Claude Fable 5.1 (EU)" } : {}),
+      structured_output: true,
+      reasoning_options: [
+        { type: "effort", values: ["none", "low", "medium", "high", "max"] },
+        { type: "budget_tokens" },
+      ],
+      cost: { input: 10, output: 50, cache_read: 0.25, cache_write: 12.5 },
+    });
+  },
+);
diff --git a/providers/requesty/models/claude-fable-5.1.toml b/providers/requesty/models/claude-fable-5.1.toml
new file mode 100644
index 00000000000..9d623a2b2bf
--- /dev/null
+++ b/providers/requesty/models/claude-fable-5.1.toml
@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/requesty/models/claude-fable-5.1@eu.toml b/providers/requesty/models/claude-fable-5.1@eu.toml
new file mode 100644
index 00000000000..4a2be775eb0
--- /dev/null
+++ b/providers/requesty/models/claude-fable-5.1@eu.toml
@@ -0,0 +1,16 @@
+base_model = "anthropic/claude-fable-5-1"
+name = "Claude Fable 5.1 (EU)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 11
+output = 55
+cache_read = 0.275
+cache_write = 13.75
diff --git a/providers/requesty/models/deepseek-v4-flash-0731.toml b/providers/requesty/models/deepseek-v4-flash-0731.toml
index d5b37a5c0f8..d73d297b44e 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731.toml
@@ -1,10 +1,17 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
-reasoning = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.076
-output = 0.153
-cache_read = 0.014
+input = 0.14
+output = 0.28
+cache_read = 0.07
 
 [limit]
 context = 1_048_576
+output = 131_072
diff --git a/providers/requesty/models/deepseek-v4-flash.toml b/providers/requesty/models/deepseek-v4-flash.toml
index 991114cd014..415669c9c5e 100644
--- a/providers/requesty/models/deepseek-v4-flash.toml
+++ b/providers/requesty/models/deepseek-v4-flash.toml
@@ -1,8 +1,17 @@
 base_model = "deepseek/deepseek-v4-flash"
-reasoning = false
-structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.14
+output = 0.28
+cache_read = 0.07
+
+[limit]
+context = 1_048_576
+output = 131_072
diff --git a/providers/requesty/models/deepseek-v4-pro-0813@eu.toml b/providers/requesty/models/deepseek-v4-pro-0813@eu.toml
new file mode 100644
index 00000000000..52382440332
--- /dev/null
+++ b/providers/requesty/models/deepseek-v4-pro-0813@eu.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 1.75
+output = 3.5
+cache_read = 0.44
+
+[limit]
+context = 1_048_576
+output = 1_048_576
diff --git a/providers/requesty/models/deepseek-v4-pro.toml b/providers/requesty/models/deepseek-v4-pro.toml
index 6b68c88eba5..3f837b71894 100644
--- a/providers/requesty/models/deepseek-v4-pro.toml
+++ b/providers/requesty/models/deepseek-v4-pro.toml
@@ -11,3 +11,6 @@ type = "budget_tokens"
 input = 1.32
 output = 3.96
 cache_read = 0.044
+
+[limit]
+output = 131_072
diff --git a/providers/requesty/models/fugu-ultra.toml b/providers/requesty/models/fugu-ultra.toml
index bc82853c108..f26c055ac34 100644
--- a/providers/requesty/models/fugu-ultra.toml
+++ b/providers/requesty/models/fugu-ultra.toml
@@ -1,7 +1,13 @@
 base_model = "sakana/fugu-ultra"
-reasoning = false
 structured_output = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 5
 output = 30
diff --git a/providers/requesty/models/gemini-3.8-flash.toml b/providers/requesty/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..05011d631a8
--- /dev/null
+++ b/providers/requesty/models/gemini-3.8-flash.toml
@@ -0,0 +1,16 @@
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+
+[limit]
+output = 65_535
diff --git a/providers/requesty/models/gemini-3.8-flash@eu.toml b/providers/requesty/models/gemini-3.8-flash@eu.toml
new file mode 100644
index 00000000000..09f662715c8
--- /dev/null
+++ b/providers/requesty/models/gemini-3.8-flash@eu.toml
@@ -0,0 +1,17 @@
+base_model = "google/gemini-3.8-flash"
+name = "Gemini 3.8 Flash (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.825
+output = 4.125
+cache_read = 0.0825
+
+[limit]
+output = 65_535
diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
index 12d3bbf1031..45ac44632bb 100644
--- a/providers/requesty/models/glm-5.3-flash.toml
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.2
+output = 0.5
+cache_read = 0.07
 
 [limit]
-context = 1_048_576
+output = 262_144
diff --git a/providers/requesty/models/glm-5.3-flash@eu.toml b/providers/requesty/models/glm-5.3-flash@eu.toml
new file mode 100644
index 00000000000..d58a76089fe
--- /dev/null
+++ b/providers/requesty/models/glm-5.3-flash@eu.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3-Flash (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.07
+
+[limit]
+output = 262_144
diff --git a/providers/requesty/models/hy3.toml b/providers/requesty/models/hy3.toml
index c6be2366f65..e5668d835a2 100644
--- a/providers/requesty/models/hy3.toml
+++ b/providers/requesty/models/hy3.toml
@@ -1,4 +1,5 @@
 base_model = "tencent/hy3"
+base_model_omit = ["limit.input"]
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/requesty/models/kimi-k2.7-code.toml b/providers/requesty/models/kimi-k2.7-code.toml
index 7dae44e6ff6..22d04d34e8a 100644
--- a/providers/requesty/models/kimi-k2.7-code.toml
+++ b/providers/requesty/models/kimi-k2.7-code.toml
@@ -1,4 +1,5 @@
 base_model = "moonshotai/kimi-k2.7-code"
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/requesty/models/mimo-v2.5-pro.toml b/providers/requesty/models/mimo-v2.5-pro.toml
index 96a7a3599ce..7f3a4a10bb6 100644
--- a/providers/requesty/models/mimo-v2.5-pro.toml
+++ b/providers/requesty/models/mimo-v2.5-pro.toml
@@ -1,7 +1,13 @@
 base_model = "xiaomi/mimo-v2.5-pro"
-reasoning = false
 structured_output = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.435
 output = 0.87
diff --git a/providers/requesty/models/mimo-v2.5.toml b/providers/requesty/models/mimo-v2.5.toml
index 4ba7b3143bc..962066d1fc5 100644
--- a/providers/requesty/models/mimo-v2.5.toml
+++ b/providers/requesty/models/mimo-v2.5.toml
@@ -1,7 +1,13 @@
 base_model = "xiaomi/mimo-v2.5"
-reasoning = false
 structured_output = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.14
 output = 0.28
diff --git a/providers/requesty/models/qwen3.8-2.4T-A95B@eu.toml b/providers/requesty/models/qwen3.8-2.4T-A95B@eu.toml
new file mode 100644
index 00000000000..4c87846e555
--- /dev/null
+++ b/providers/requesty/models/qwen3.8-2.4T-A95B@eu.toml
@@ -0,0 +1,18 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8 2.4T A95B (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 2.5
+output = 6
+cache_read = 0.63
+
+[limit]
+context = 1_000_000
+output = 262_144
diff --git a/providers/requesty/models/qwen3.8-flash-next.toml b/providers/requesty/models/qwen3.8-flash-next.toml
new file mode 100644
index 00000000000..e7792ebeda9
--- /dev/null
+++ b/providers/requesty/models/qwen3.8-flash-next.toml
@@ -0,0 +1,16 @@
+base_model = "alibaba/qwen3.8-flash-next"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+output = 262_144
diff --git a/providers/requesty/models/qwen3.8-flash-next@eu.toml b/providers/requesty/models/qwen3.8-flash-next@eu.toml
new file mode 100644
index 00000000000..43a5b2489c6
--- /dev/null
+++ b/providers/requesty/models/qwen3.8-flash-next@eu.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-flash-next"
+name = "Qwen3.8 Flash Next (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+output = 262_144
diff --git a/providers/requesty/models/qwen3.8-flash.toml b/providers/requesty/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..322e429cce3
--- /dev/null
+++ b/providers/requesty/models/qwen3.8-flash.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
+
+[limit]
+context = 1_048_576

From 35418aed613fb1326c09f648d4034ee2716d9d49 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:25:43 -0500
Subject: [PATCH 0511/2012] fix: Add deepseek-v4-flash-vision-exp to
 fireworks-ai (#6128)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-flash-vision-exp.toml  | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..e3955a13717
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,20 @@
+# Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
+# Effort: reasoning_effort = low|high|max (same surface as deepseek-v4-flash-0731 on Fireworks)
+# Pricing/model: https://fireworks.ai/models/deepseek-ai/deepseek-v4-flash-vision-exp (accessed 2026-09-02)
+# Peer: providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007

From f9ac5b398bf2a7ccb3b70521d295588e4f71e09b Mon Sep 17 00:00:00 2001
From: Vladimir Glafirov <vglafirov@gitlab.com>
Date: Thu, 3 Sep 2026 00:25:55 +0200
Subject: [PATCH 0512/2012] feat: add gitlab duo-chat-fable-5-1 model (#6121)

---
 providers/gitlab/models/duo-chat-fable-5-1.toml | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/gitlab/models/duo-chat-fable-5-1.toml

diff --git a/providers/gitlab/models/duo-chat-fable-5-1.toml b/providers/gitlab/models/duo-chat-fable-5-1.toml
new file mode 100644
index 00000000000..1379d3f11bf
--- /dev/null
+++ b/providers/gitlab/models/duo-chat-fable-5-1.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-fable-5-1"
+name = "Agentic Chat (Claude Fable 5.1)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 24d7e90b26857db559f2c1261bd43069a7337b9a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 22:26:10 +0000
Subject: [PATCH 0513/2012] chore(sync): update OpenRouter model catalog
 (#6137)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92d05b69dd5..a010dd38ba2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02747
-output = 2.05494
-cache_read = 0.085623
+input = 0.87
+output = 1.74
+cache_read = 0.0725
 
 [limit]
 context = 1_048_576

From 41e2ba6e5a70a62b2fcffbe3efbeb9b40f1b0976 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:26:20 -0500
Subject: [PATCH 0514/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6070)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-max-0902.toml      | 24 +++++++++++++++++++
 .../models/google/gemini-3.8-flash.toml       | 13 ++++++++++
 .../meta/muse-spark-1.3-contributor.toml      | 23 ++++++++++++++++++
 .../vercel/models/meta/muse-spark-1.3.toml    | 23 ++++++++++++++++++
 providers/vercel/models/tencent/hy3.toml      |  1 +
 .../vercel/models/zai/glm-5.3-promo-50.toml   | 23 ++++++++++++++++++
 6 files changed, 107 insertions(+)
 create mode 100644 providers/vercel/models/alibaba/qwen3.8-max-0902.toml
 create mode 100644 providers/vercel/models/google/gemini-3.8-flash.toml
 create mode 100644 providers/vercel/models/meta/muse-spark-1.3-contributor.toml
 create mode 100644 providers/vercel/models/meta/muse-spark-1.3.toml
 create mode 100644 providers/vercel/models/zai/glm-5.3-promo-50.toml

diff --git a/providers/vercel/models/alibaba/qwen3.8-max-0902.toml b/providers/vercel/models/alibaba/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..e1289bef6ef
--- /dev/null
+++ b/providers/vercel/models/alibaba/qwen3.8-max-0902.toml
@@ -0,0 +1,24 @@
+name = "Qwen3.8 Max 0902"
+description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
+family = "qwen3.8-max"
+release_date = "2026-09-01"
+last_updated = "2026-09-01"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+cache_write = 2.5
+
+[limit]
+context = 991_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/vercel/models/google/gemini-3.8-flash.toml b/providers/vercel/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..32fddda9360
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,13 @@
+base_model = "google/gemini-3.8-flash"
+reasoning_options = []
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/meta/muse-spark-1.3-contributor.toml b/providers/vercel/models/meta/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..2decd48753b
--- /dev/null
+++ b/providers/vercel/models/meta/muse-spark-1.3-contributor.toml
@@ -0,0 +1,23 @@
+name = "Muse Spark 1.3 Contributor"
+description = "Open Llama multimodal model for image understanding and text reasoning"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/vercel/models/meta/muse-spark-1.3.toml b/providers/vercel/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..1ef5327f57f
--- /dev/null
+++ b/providers/vercel/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,23 @@
+name = "Muse Spark 1.3"
+description = "Open Llama multimodal model for image understanding and text reasoning"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/vercel/models/tencent/hy3.toml b/providers/vercel/models/tencent/hy3.toml
index 066f2f8f4ec..dde3c6bfc50 100644
--- a/providers/vercel/models/tencent/hy3.toml
+++ b/providers/vercel/models/tencent/hy3.toml
@@ -6,6 +6,7 @@
 # API: {"chat_template_kwargs": {"reasoning_effort": "no_think"|"low"|"high"}}
 # mapped: no_think → none
 base_model = "tencent/hy3"
+base_model_omit = ["limit.input"]
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/vercel/models/zai/glm-5.3-promo-50.toml b/providers/vercel/models/zai/glm-5.3-promo-50.toml
new file mode 100644
index 00000000000..3d49e4731b7
--- /dev/null
+++ b/providers/vercel/models/zai/glm-5.3-promo-50.toml
@@ -0,0 +1,23 @@
+name = "GLM 5.3 (50% off)"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+family = "glm"
+release_date = "2026-08-18"
+last_updated = "2026-08-18"
+attachment = false
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.7
+output = 2.2
+cache_read = 0.13
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text"]
+output = ["text"]

From d037fa0d63600b4b8a753fdf8f3cbb243478b2ce Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:27:24 -0500
Subject: [PATCH 0515/2012] fix(azure): use documented model limits and pricing
 (#6136)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
Co-authored-by: DenisMkS <179223937+DenisMkS@users.noreply.github.com>
---
 providers/azure/models/gpt-5.6-sol.toml | 16 +++++++++-------
 providers/azure/models/grok-4.6.toml    | 14 ++++++++++++++
 2 files changed, 23 insertions(+), 7 deletions(-)
 create mode 100644 providers/azure/models/grok-4.6.toml

diff --git a/providers/azure/models/gpt-5.6-sol.toml b/providers/azure/models/gpt-5.6-sol.toml
index 2e42e9c1929..2eb13824b76 100644
--- a/providers/azure/models/gpt-5.6-sol.toml
+++ b/providers/azure/models/gpt-5.6-sol.toml
@@ -1,5 +1,7 @@
 # Pricing: https://azure.microsoft.com/en-us/blog/gpt-5-6-now-available-in-microsoft-foundry/
-# Azure OpenAI metered API pricing for the GPT-5.6 Sol model will be reduced to $4.00 per 1M input tokens and $20.00 per 1M output tokens (20% reduction on input pricing and 33.3% reduction on output pricing). This promotional pricing will run from September 1, 2026, through at least November 30, 2026.
+# Rates: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+# $4 input / $20 output promotion: 2026-09-01 through at least 2026-11-30.
+# Keep published cache and long-context rates; the announcement does not specify discounts for them.
 base_model = "openai/gpt-5.6-sol"
 status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
@@ -7,12 +9,12 @@ reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high
 [cost]
 input = 4.00
 output = 20.00
-cache_read = 0.40
-cache_write = 5.00
+cache_read = 0.50
+cache_write = 6.25
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 8.00
-output = 30.00
-cache_read = 0.80
-cache_write = 10.00
+input = 10.00
+output = 45.00
+cache_read = 1.00
+cache_write = 12.50
diff --git a/providers/azure/models/grok-4.6.toml b/providers/azure/models/grok-4.6.toml
new file mode 100644
index 00000000000..cde00e1c7ee
--- /dev/null
+++ b/providers/azure/models/grok-4.6.toml
@@ -0,0 +1,14 @@
+# Source: https://learn.microsoft.com/en-us/azure/foundry/foundry-models/how-to/use-foundry-models-grok
+# Pricing: https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/grok/
+# Azure's public pricing page does not list Grok 4.6 rates as of 2026-09-02.
+# Omit cost rather than assume xAI's token/cache rates or >200k context tier.
+base_model = "xai/grok-4.6"
+status = "beta"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[limit]
+context = 200_000
+output = 128_000

From f35c204dad68bd8635bbb2370b5da12bc13b85ed Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 17:39:26 -0500
Subject: [PATCH 0516/2012] feat(sync): sync Meta pricing and context from
 public docs (#6130)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/index.ts           |   5 +-
 packages/core/src/sync/providers/meta.ts  | 125 ++++++++++++++
 packages/core/test/meta.test.ts           | 191 ++++++++++++++++++++++
 providers/meta/models/muse-spark-1.1.toml |   9 +-
 sync.md                                   |   9 +
 5 files changed, 336 insertions(+), 3 deletions(-)
 create mode 100644 packages/core/src/sync/providers/meta.ts
 create mode 100644 packages/core/test/meta.test.ts

diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index f08532afa14..2e151aa702b 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -25,6 +25,7 @@ import { inceptron } from "./providers/inceptron.js";
 import { kilo } from "./providers/kilo.js";
 import { llmgateway, llmgatewayProviders } from "./providers/llmgateway.js";
 import { mergeGateway } from "./providers/merge-gateway.js";
+import { meta } from "./providers/meta.js";
 import { nanoGpt } from "./providers/nano-gpt.js";
 import { openai } from "./providers/openai.js";
 import { ofox } from "./providers/ofox.js";
@@ -149,6 +150,7 @@ export const providers: {
   llmgateway: SyncProvider<any>;
   "llmgateway-providers": SyncProvider<any>;
   "merge-gateway": SyncProvider<any>;
+  meta: SyncProvider<any>;
   "nano-gpt": SyncProvider<any>;
   ofox: SyncProvider<any>;
   openai: SyncProvider<any>;
@@ -183,6 +185,7 @@ export const providers: {
   llmgateway,
   "llmgateway-providers": llmgatewayProviders,
   "merge-gateway": mergeGateway,
+  meta,
   "nano-gpt": nanoGpt,
   ofox,
   openai,
@@ -215,7 +218,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "meta", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
diff --git a/packages/core/src/sync/providers/meta.ts b/packages/core/src/sync/providers/meta.ts
new file mode 100644
index 00000000000..7c56b9be2ca
--- /dev/null
+++ b/packages/core/src/sync/providers/meta.ts
@@ -0,0 +1,125 @@
+import { z } from "zod";
+
+import type { SyncProvider, SyncedModel } from "../index.js";
+
+const MODELS_ENDPOINT = "https://dev.meta.ai/docs/models.md";
+const PRICING_ENDPOINT = "https://dev.meta.ai/docs/pricing-rate-limits.md";
+
+const MetaResponse = z.object({ models: z.string(), pricing: z.string() });
+const MetaModel = z.object({
+  id: z.string().regex(/^[a-z0-9][a-z0-9.-]*$/),
+  context: z.number().int().positive().safe(),
+  cost: z.object({
+    input: z.number().finite().nonnegative(),
+    output: z.number().finite().nonnegative(),
+    cache_read: z.number().finite().nonnegative(),
+  }),
+});
+
+export type MetaModel = z.infer<typeof MetaModel>;
+
+function cells(line: string) {
+  return line.trim().split("|").slice(1, -1).map((cell) => cell.trim());
+}
+
+function table(markdown: string, header: string[]) {
+  const lines = markdown.split("\n");
+  const start = lines.findIndex((line) => JSON.stringify(cells(line)) === JSON.stringify(header));
+  if (start < 0) throw new Error(`Meta docs are missing the ${header.join(" / ")} table`);
+  const separator = cells(lines[start + 1] ?? "");
+  if (separator.length !== header.length || separator.some((cell) => !/^:?-+:?$/.test(cell))) {
+    throw new Error("Meta docs have an invalid table separator");
+  }
+  const rows: string[][] = [];
+  for (const line of lines.slice(start + 2)) {
+    if (!line.trim().startsWith("|")) break;
+    const row = cells(line);
+    if (row.length !== header.length) throw new Error("Meta docs have an invalid table row");
+    rows.push(row);
+  }
+  if (rows.length === 0) throw new Error("Meta docs table is empty");
+  return rows;
+}
+
+function tierCost(markdown: string, tier: string) {
+  const sections = markdown.split(/^### /m).filter((section) => section.split("\n")[0]?.includes(`{#${tier}}`));
+  if (sections.length !== 1) throw new Error(`Meta docs need exactly one pricing section for ${tier}`);
+  const prices = new Map<string, number>();
+  for (const [usage, price] of table(sections[0]!, ["Usage", "Price per 1M tokens"])) {
+    if (!/^\$\d+(?:\.\d+)?$/.test(price!) || prices.has(usage!)) {
+      throw new Error(`Meta docs have invalid or duplicate ${tier} pricing`);
+    }
+    prices.set(usage!, Number(price!.slice(1)));
+  }
+  return {
+    input: prices.get("Input"),
+    output: prices.get("Output"),
+    cache_read: prices.get("Cached input"),
+  };
+}
+
+export function parseMetaModels(raw: unknown): MetaModel[] {
+  const response = MetaResponse.parse(raw);
+  const rows = table(response.models, ["Model ID", "Tier", "Input modalities", "Output modalities", "Context window"]);
+  const ids = new Set<string>();
+  return rows.map(([model, tierLink, _input, output, window]) => {
+    const id = model?.match(/^`([^`]+)`$/)?.[1];
+    const tier = tierLink?.match(/^\[[^\]]+\]\([^)]*#([a-z-]+)\)$/)?.[1];
+    const tokens = window?.match(/^(\d+|\d{1,3}(?:,\d{3})+) tokens$/)?.[1];
+    if (id === undefined || tier === undefined || tokens === undefined || output !== "Text") {
+      throw new Error("Meta docs have an unsupported token-priced model row");
+    }
+    if (ids.has(id)) throw new Error(`Meta docs have a duplicate model: ${id}`);
+    ids.add(id);
+    const parsed = MetaModel.safeParse({
+      id,
+      context: Number(tokens.replaceAll(",", "")),
+      cost: tierCost(response.pricing, tier),
+    });
+    if (!parsed.success) {
+      parsed.error.cause = { provider: "meta", model: id };
+      throw parsed.error;
+    }
+    return parsed.data;
+  });
+}
+
+export async function fetchMetaModels(fetcher: typeof fetch = fetch) {
+  const [models, pricing] = await Promise.all([MODELS_ENDPOINT, PRICING_ENDPOINT].map(async (url) => {
+    const response = await fetcher(url);
+    if (!response.ok) throw new Error(`Meta docs request failed: ${response.status} ${response.statusText}`);
+    return response.text();
+  }));
+  return { models, pricing };
+}
+
+export const meta = {
+  id: "meta",
+  name: "Meta",
+  modelsDir: "providers/meta/models",
+  skipCreates: true,
+  deleteMissing: false,
+  sourceID(model) {
+    return model.id;
+  },
+  skippedNotice(ids) {
+    return ids.length === 0 ? [] : [
+      `Meta's public docs list models requiring hand-authored metadata: ${ids.map((id) => `\`${id}\``).join(", ")}`,
+    ];
+  },
+  fetchModels: fetchMetaModels,
+  parseModels: parseMetaModels,
+  translateModel(model, context) {
+    const authored = context.authored(model.id);
+    if (authored === undefined) return undefined;
+    // Only the documented token rates and context window are authoritative.
+    // Keep output limits, modalities, reasoning controls, dates, and base_model.
+    const limit = context.existing(model.id)?.limit?.context === model.context
+      ? authored.limit
+      : { ...authored.limit, context: model.context };
+    return {
+      id: model.id,
+      model: { ...authored, limit, cost: { ...authored.cost, ...model.cost } } as SyncedModel,
+    };
+  },
+} satisfies SyncProvider<MetaModel>;
diff --git a/packages/core/test/meta.test.ts b/packages/core/test/meta.test.ts
new file mode 100644
index 00000000000..7163c343f54
--- /dev/null
+++ b/packages/core/test/meta.test.ts
@@ -0,0 +1,191 @@
+import { expect, test } from "bun:test";
+import { mkdtemp, rm } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import path from "node:path";
+
+import { groups, providers, syncProvider, type ExistingModel } from "../src/sync/index.js";
+import { fetchMetaModels, meta, parseMetaModels } from "../src/sync/providers/meta.js";
+
+// Public docs format; intentionally exclude account-scoped API responses.
+const models = `
+## Muse Spark {#muse-spark}
+
+| Model ID | Tier | Input modalities | Output modalities | Context window |
+| :---- | :---- | :---- | :---- | :---- |
+| \`muse-spark-1.2\` | [Standard](/docs/pricing-rate-limits#standard-tier) | Text, image, video, audio, PDF | Text | 1,048,576 tokens |
+| \`muse-spark-1.2-contributor\` | [Contributor](/docs/pricing-rate-limits#contributor-tier) | Text, image, video, audio, PDF | Text | 1,048,576 tokens |
+
+## Muse Image
+
+| Model ID | Family | Input | Output |
+| :---- | :---- | :---- | :---- |
+| \`muse-image-1.0\` | Muse Image | Text, image | Image |
+`;
+
+const pricing = `
+### Standard tier {#standard-tier}
+
+| Usage | Price per 1M tokens |
+| :---- | :---- |
+| Cached input | $0.15 |
+| Input | $1.25 |
+| Output | $4.25 |
+
+### Contributor tier {#contributor-tier}
+
+| Usage | Price per 1M tokens |
+| :---- | :---- |
+| Cached input | $0.002 |
+| Input | $0.10 |
+| Output | $0.20 |
+
+### Image generation
+
+Muse Image costs $0.01 per image, not per token.
+`;
+
+const source = { models, pricing };
+
+test("Meta sync is registered for direct and hourly runs", () => {
+  expect(providers.meta).toBe(meta);
+  expect(groups.direct).toContain("meta");
+  expect(meta.skipCreates).toBe(true);
+  expect(meta.deleteMissing).toBe(false);
+});
+
+test("parses public text models, tier prices in USD/MTok, and context windows", () => {
+  expect(parseMetaModels(source)).toEqual([
+    { id: "muse-spark-1.2", context: 1_048_576, cost: { input: 1.25, output: 4.25, cache_read: 0.15 } },
+    { id: "muse-spark-1.2-contributor", context: 1_048_576, cost: { input: 0.1, output: 0.2, cache_read: 0.002 } },
+  ]);
+  expect(parseMetaModels({ ...source, pricing: pricing.replace("$0.002", "$0") })[1]?.cost.cache_read).toBe(0);
+  expect(parseMetaModels({ ...source, models: models.replaceAll("1,048,576", "1048576") })[0]?.context).toBe(1_048_576);
+});
+
+test("rejects incomplete or changed docs rather than guessing prices or limits", () => {
+  for (const bad of [
+    { ...source, models: "<html>Unavailable</html>" },
+    { ...source, pricing: "" },
+    { ...source, models: models.replace("1,048,576 tokens", "Unknown") },
+    { ...source, models: models.replace("1,048,576 tokens", "0 tokens") },
+    { ...source, models: models.replace("1,048,576 tokens", "1,04,8576 tokens") },
+    { ...source, models: models.replace("#standard-tier)", "#unknown-tier)") },
+    { ...source, models: models.replace("`muse-spark-1.2`", "`../private`") },
+    { ...source, pricing: pricing.replace("$1.25", "€1.25") },
+    { ...source, pricing: pricing.replace("$1.25", "$-1") },
+    { ...source, pricing: pricing.replace("Price per 1M tokens", "Price per 1K tokens") },
+    { ...source, pricing: pricing.replace("| Output | $4.25 |", "") },
+    { ...source, pricing: pricing.replace("| Output | $4.25 |", "| Output | $4.25 |\n| Input | $2 |") },
+    { ...source, models: models.replace("muse-spark-1.2-contributor", "muse-spark-1.2") },
+    { ...source, models: models.replace("| :---- | :---- | :---- | :---- | :---- |", "| broken |") },
+  ]) {
+    expect(() => parseMetaModels(bad)).toThrow();
+  }
+});
+
+test("updates only authoritative fields without expanding inherited metadata", () => {
+  const authored: ExistingModel = {
+    base_model: "meta/muse-spark-1.2",
+    base_model_omit: ["limit.input"],
+    reasoning_options: [{ type: "effort", values: ["minimal", "low", "medium", "high", "xhigh"] }],
+    cost: { input: 9, output: 9, cache_read: 9, input_audio: 2, reasoning: 3 },
+    status: "beta",
+  };
+  const existing: ExistingModel = {
+    ...authored,
+    name: "Muse Spark 1.2",
+    reasoning: true,
+    limit: { context: 1_048_576, output: 131_072 },
+    modalities: { input: ["text", "image", "pdf", "video"], output: ["text"] },
+  };
+  const model = parseMetaModels(source)[0]!;
+  const context = { authored: () => authored, existing: () => existing };
+  const translated = meta.translateModel(model, context)!;
+  expect(translated.model).toEqual({
+    ...authored,
+    limit: undefined,
+    cost: { input: 1.25, output: 4.25, cache_read: 0.15, input_audio: 2, reasoning: 3 },
+  });
+  expect(meta.translateModel({ ...model, context: 2_000_000 }, context)?.model.limit).toEqual({ context: 2_000_000 });
+  expect(authored.cost?.input).toBe(9);
+});
+
+test("unknown documented models are reported, never synthesized", () => {
+  const model = parseMetaModels(source)[0]!;
+  expect(meta.translateModel(model, { authored: () => undefined, existing: () => undefined })).toBeUndefined();
+  expect(meta.sourceID(model)).toBe(model.id);
+  expect(meta.skippedNotice([model.id]).join(" ")).toContain(model.id);
+});
+
+test("fetches only public docs without credentials and rejects HTTP failures", async () => {
+  const urls: string[] = [];
+  const fetcher = (async (url: string, init?: RequestInit) => {
+    urls.push(url);
+    expect(init).toBeUndefined();
+    return new Response(url.endsWith("/models.md") ? models : pricing);
+  }) as typeof fetch;
+  expect(await fetchMetaModels(fetcher)).toEqual(source);
+  expect(urls).toEqual(["https://dev.meta.ai/docs/models.md", "https://dev.meta.ai/docs/pricing-rate-limits.md"]);
+  const failing = (async () => new Response("Unavailable", { status: 503 })) as typeof fetch;
+  await expect(fetchMetaModels(failing)).rejects.toThrow("Meta docs request failed: 503");
+});
+
+test("runner preserves inherited controls, retains absent models, and is idempotent", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "models-dev-meta-"));
+  const modelsDir = path.join(root, "providers", "meta", "models");
+  const filename = path.join(modelsDir, "muse-spark-1.2.toml");
+  const original = `# Keep this authored source comment.
+base_model = "meta/muse-spark-1.2"
+base_model_omit = ["limit.input"]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+[cost]
+input = 9
+output = 9
+cache_read = 9
+`;
+  const absent = path.join(modelsDir, "muse-spark-1.1.toml");
+  try {
+    await Bun.write(path.join(root, "models", "meta", "muse-spark-1.2.toml"), `
+name = "Muse Spark 1.2"
+description = "Test fixture"
+release_date = "2026-01-01"
+last_updated = "2026-01-01"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+[limit]
+context = 1048576
+input = 1048576
+output = 131072
+[modalities]
+input = ["text", "image", "pdf", "video"]
+output = ["text"]
+`);
+    await Bun.write(filename, original);
+    await Bun.write(absent, original);
+    const provider = { ...meta, modelsDir, fetchModels: async () => source };
+    const dry = await syncProvider(provider, { dryRun: true, openIssues: false });
+    expect(dry.updated).toBe(1);
+    expect(await Bun.file(filename).text()).toBe(original);
+    const first = await syncProvider(provider, { openIssues: false });
+    expect(first).toMatchObject({ created: 0, updated: 1, deleted: 0, unchanged: 1 });
+    expect(first.notices.join(" ")).toContain("muse-spark-1.2-contributor");
+    expect(await Bun.file(absent).text()).toBe(original);
+    const text = await Bun.file(filename).text();
+    expect(text).toStartWith("# Keep this authored source comment.");
+    const result = Bun.TOML.parse(text);
+    expect(result.base_model).toBe("meta/muse-spark-1.2");
+    expect(result.base_model_omit).toEqual(["limit.input"]);
+    expect(result.reasoning_options).toEqual(Bun.TOML.parse(original).reasoning_options);
+    expect(result.limit).toBeUndefined();
+    expect(result.cost).toEqual({ input: 1.25, output: 4.25, cache_read: 0.15 });
+    const second = await syncProvider(provider, { openIssues: false });
+    expect(second).toMatchObject({ created: 0, updated: 0, deleted: 0, unchanged: 2 });
+    expect(await Bun.file(filename).text()).toBe(text);
+    await expect(syncProvider({ ...provider, fetchModels: async () => ({ ...source, pricing: "" }) })).rejects.toThrow();
+    expect(await Bun.file(filename).text()).toBe(text);
+  } finally {
+    await rm(root, { recursive: true, force: true });
+  }
+});
diff --git a/providers/meta/models/muse-spark-1.1.toml b/providers/meta/models/muse-spark-1.1.toml
index 78580fce3d0..d4a75e74f68 100644
--- a/providers/meta/models/muse-spark-1.1.toml
+++ b/providers/meta/models/muse-spark-1.1.toml
@@ -1,12 +1,17 @@
 # Sources:
+# https://dev.meta.ai/docs/models/
 # https://ai.developer.meta.com/docs/getting-started/pricing-rate-limits/
 # https://developer.meta.com/ai/resources/blog/build-with-muse-spark/
-
 base_model = "meta/muse-spark-1.1"
 
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25
 output = 4.25
 cache_read = 0.15
+
+[limit]
+context = 1_048_576
diff --git a/sync.md b/sync.md
index 2aa0ac34651..2acf38d148a 100644
--- a/sync.md
+++ b/sync.md
@@ -264,6 +264,15 @@ xAI is implemented in `packages/core/src/sync/providers/xai.ts`.
 - The endpoint is used only to monitor catalog availability. Existing TOMLs are preserved byte-for-byte, including models absent from the response, because model access can be scoped to the API project.
 - Fine-tuned and customer-owned models are excluded. Unknown first-party models are ignored because the endpoint does not provide enough lifecycle or visibility metadata to distinguish public catalog additions.
 
+## Meta Notes
+
+- Run with `bun models:sync meta` or as part of the `direct` group. Registration also enables the hourly provider-specific workflow; no new secret is required.
+- Sources: `https://dev.meta.ai/docs/models.md` and `https://dev.meta.ai/docs/pricing-rate-limits.md`.
+- Meta's `/v1/models` endpoint is team-scoped and exposes IDs and registry timestamps, not pricing or limits. Use the public documentation instead of treating account-visible IDs as public catalog additions.
+- Sync only the token-priced text models in the public model table. Update standard/contributor input, output, and cached-input USD/MTok prices and context windows. Keep output limits, modalities (including audio support caveats), reasoning controls, dates, inheritance, and other authored fields unchanged.
+- New documented models open deduped missing-model issues for manual authoring (`skipCreates`); local models absent from the docs are retained (`deleteMissing: false`). Image generation, transcription, and self-hosted models are outside this sync's scope.
+- Missing tables, unknown pricing tiers, invalid prices/limits, and duplicate model rows fail before writing. Documentation format changes require updating the parser, not guessing defaults.
+
 ## OVHcloud Notes
 
 OVHcloud AI Endpoints is implemented in `packages/core/src/sync/providers/ovhcloud.ts`.

From c6b157f690a27c1b8bde5c1d4083397773823fb6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 2 Sep 2026 23:23:07 +0000
Subject: [PATCH 0517/2012] chore(sync): update OpenRouter model catalog
 (#6140)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576

From fac23bc1a963cba99b4a82356b7e96fe246600eb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:01:23 +0000
Subject: [PATCH 0518/2012] chore(sync): update EmpirioLabs AI model catalog
 (#6145)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../empiriolabs/models/qwen3-8-max-0902.toml  | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/empiriolabs/models/qwen3-8-max-0902.toml

diff --git a/providers/empiriolabs/models/qwen3-8-max-0902.toml b/providers/empiriolabs/models/qwen3-8-max-0902.toml
new file mode 100644
index 00000000000..1ecde38c227
--- /dev/null
+++ b/providers/empiriolabs/models/qwen3-8-max-0902.toml
@@ -0,0 +1,20 @@
+base_model = "alibaba/qwen3.8-max"
+name = "Qwen3.8 Max 0902"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 262_144
+
+[cost]
+input = 2
+output = 6
+cache_read = 2
+
+[modalities]
+input = ["text", "image", "video"]

From 9f24871f7133ff784ddd2f2db489765f3e2eaf78 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:01:25 +0000
Subject: [PATCH 0519/2012] chore(sync): update Kilo model catalog (#6147)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/meta-llama/llama-3.3-70b-instruct.toml |  3 ++-
 .../kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml    |  2 +-
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml |  4 ++--
 providers/kilo/models/qwen/qwen3-14b.toml              |  4 ++--
 providers/kilo/models/qwen/qwen3.5-397b-a17b.toml      |  3 +++
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                | 10 +++-------
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 8 files changed, 21 insertions(+), 21 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
index 1651fa89774..878106e6be1 100644
--- a/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
@@ -8,4 +8,5 @@ input = 0.1
 output = 0.32
 
 [limit]
-output = 115_200
+context = 131_072
+output = 16_384
diff --git a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
index 526a9ca0646..dbe7bd30394 100644
--- a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -11,4 +11,4 @@ output = 0.2
 cache_read = 0.025
 
 [limit]
-output = 228_000
+output = 235_929
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 25c72a0b0fb..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 202_800
-output = 182_520
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 2ce8bc46042..67828967f41 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 40_960
-output = 16_384
+context = 131_072
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
index d0e349f341a..3d76879dbef 100644
--- a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
@@ -9,5 +9,8 @@ values = ["none", "high"]
 input = 0.39
 output = 2.34
 
+[limit]
+output = 235_929
+
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fee21b1f2d8 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,6 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-context = 198_000
-output = 16_384
+input = 0.55
+output = 2.2
+cache_read = 0.11
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 82fdc44362b..7c559a202b6 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.15
-output = 3.5
-cache_read = 0.1
+input = 1.092
+output = 3.432
+cache_read = 0.1794
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From a0e210e8dd233b50e76a85d07169cae9f60c0fe2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:01:40 +0000
Subject: [PATCH 0520/2012] chore(sync): update NanoGPT model catalog (#6146)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From d2bdede48299e6932d790a27de16a22fb4b86746 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:01:42 +0000
Subject: [PATCH 0521/2012] chore(sync): update OpenRouter model catalog
 (#6148)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml  |  6 +++---
 .../models/deepseek/deepseek-v4-pro-0813.toml          |  6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml    |  6 +++---
 .../models/meta-llama/llama-3.3-70b-instruct.toml      |  7 +++----
 .../models/nvidia/nemotron-3-nano-30b-a3b.toml         |  4 ++--
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml      |  8 ++++----
 providers/openrouter/models/qwen/qwen3-14b.toml        |  8 +++++---
 .../openrouter/models/qwen/qwen3.5-397b-a17b.toml      | 10 ++++++++--
 providers/openrouter/models/tencent/hy3.toml           |  6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml          |  9 +++------
 providers/openrouter/models/~z-ai/glm-latest.toml      |  8 ++++----
 11 files changed, 41 insertions(+), 37 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a010dd38ba2..66443ade4c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.0725
+input = 1.04226
+output = 2.08452
+cache_read = 0.086855
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
index c6879803b3d..5080de4489c 100644
--- a/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
@@ -3,10 +3,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.71
-output = 0.71
-cache_read = 0.71
+input = 0.1
+output = 0.32
 
 [limit]
 context = 131_072
-output = 115_200
+output = 16_384
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
index 6469684457d..bf732029012 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -9,7 +9,7 @@ type = "toggle"
 [cost]
 input = 0.05
 output = 0.2
-cache_read = 0.025
+cache_read = 0.03
 
 [limit]
-output = 228_000
+output = 235_929
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 5dcd117448e..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 182_520
+output = 32_768
diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index 2e47169c42b..b6ce24c73e0 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "Qwen3 14B"
 description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
 family = "qwen"
@@ -15,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.12
-output = 0.24
+input = 0.2275
+output = 0.91
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
index a73eaf00e0e..a5c5b5bf91f 100644
--- a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
@@ -1,11 +1,17 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.5-397b-a17b"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.39
-output = 2.34
+input = 0.55
+output = 3.5
+cache_read = 0.225
+
+[limit]
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..6e636f26b9c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.55
+output = 2.2
+cache_read = 0.11
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 51793fd9815..3bce75132bc 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.15
-output = 3.5
-cache_read = 0.1
+input = 1.092
+output = 3.432
+cache_read = 0.1794
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From e2863218777046f5d8839a08f7a1c8adc85ff2f2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:40:30 +0000
Subject: [PATCH 0522/2012] chore(sync): update Kilo model catalog (#6150)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta/muse-glimmer-30b.toml             | 2 +-
 providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml    | 2 +-
 providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml | 4 ++--
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml            | 4 ++++
 providers/kilo/models/~z-ai/glm-flash-latest.toml            | 2 +-
 5 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/meta/muse-glimmer-30b.toml b/providers/kilo/models/meta/muse-glimmer-30b.toml
index aa01201020a..2c352c0d845 100644
--- a/providers/kilo/models/meta/muse-glimmer-30b.toml
+++ b/providers/kilo/models/meta/muse-glimmer-30b.toml
@@ -11,7 +11,7 @@ output = 1.1
 cache_read = 0.04
 
 [limit]
-output = 16_384
+output = 117_964
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
index dbe7bd30394..ecf7bafae91 100644
--- a/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -8,7 +8,7 @@ values = ["none", "high"]
 [cost]
 input = 0.05
 output = 0.2
-cache_read = 0.025
+cache_read = 0.03
 
 [limit]
 output = 235_929
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index ad993c8f794..c79bdfce589 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -13,3 +13,7 @@ input = 2
 output = 6
 cache_read = 0.25
 cache_write = 2.5
+
+[limit]
+context = 1_000_000
+output = 262_144
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index e4ce4e5dfe6..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From 4b4df4f19a438e289602a2af964f1268372c4793 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 01:40:36 +0000
Subject: [PATCH 0523/2012] chore(sync): update OpenRouter model catalog
 (#6149)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml     | 6 +++---
 providers/openrouter/models/meta/muse-glimmer-30b.toml    | 4 ++--
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml   | 3 ++-
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 2 +-
 5 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index d4077f81891..12578d9072c 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -11,8 +11,8 @@ values = ["low", "medium", "high", "xhigh"]
 
 [cost]
 input = 0.3
-output = 1.2
+output = 1.1
 cache_read = 0.04
 
 [limit]
-output = 16_384
+output = 117_964
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index 51225871db8..abacdd68c88 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -14,7 +14,8 @@ values = ["low", "medium", "xhigh"]
 [cost]
 input = 2
 output = 6
-cache_read = 0.2
+cache_read = 0.25
 
 [limit]
 context = 1_048_576
+output = 262_144
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index c32fca67d4b..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From fda29e543c98a5a561b5e59baea07cf615ca4c0d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 05:23:52 +0000
Subject: [PATCH 0524/2012] chore(sync): update OpenRouter model catalog
 (#6158)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From ff645778468651e196ed433b70d3709455da1651 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 05:24:07 +0000
Subject: [PATCH 0525/2012] chore(sync): update NanoGPT model catalog (#6160)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 1998f2b80a15182c1a26f40d36efdce8a1d9a6ae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 06:41:14 +0000
Subject: [PATCH 0526/2012] chore(sync): update Kilo model catalog (#6163)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7c559a202b6..2406d89fc54 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.092
-output = 3.432
-cache_read = 0.1794
+input = 1.106
+output = 3.476
+cache_read = 0.1817
 
 [limit]
 context = 1_048_576

From 3c1dcb31396b0463995f492b9699aa67845c18d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 06:41:26 +0000
Subject: [PATCH 0527/2012] chore(sync): update NanoGPT model catalog (#6164)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 9fafe02a2bebceabf2b758c6729b533b5ddd67d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 06:41:44 +0000
Subject: [PATCH 0528/2012] chore(sync): update OpenRouter model catalog
 (#6162)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3bce75132bc..d2137e25375 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.092
-output = 3.432
-cache_read = 0.1794
+input = 1.106
+output = 3.476
+cache_read = 0.1817
 
 [limit]
 context = 1_310_720

From 3e947122644bba96b8d4fbc3f0606b9055cdba0c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 07:29:08 +0000
Subject: [PATCH 0529/2012] chore(sync): update OpenRouter model catalog
 (#6167)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/qwen/qwen3-14b.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index b6ce24c73e0..b10030292e3 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -17,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.2275
-output = 0.91
+input = 0.12
+output = 0.24
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384
 
 [modalities]
 input = ["text"]

From 0bf1fe3ebb561f771f3f68563dbbd16e3e0eeb2a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 07:29:12 +0000
Subject: [PATCH 0530/2012] chore(sync): update Kilo model catalog (#6166)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-14b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 67828967f41..2ce8bc46042 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 131_072
-output = 8_192
+context = 40_960
+output = 16_384
 
 [modalities]
 input = ["text"]

From 85f23ac5def39da8bbe8723e00225bc17a14ea9e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 08:34:39 +0000
Subject: [PATCH 0531/2012] chore(sync): update OpenRouter model catalog
 (#6171)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 55248129361..0269e6a5aba 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -8,7 +8,8 @@ values = ["low", "high", "max"]
 [cost]
 input = 1.4
 output = 4.4
-cache_read = 0.26
+cache_read = 0.14
 
 [limit]
 context = 1_310_720
+output = 262_144

From 348892f1861906f714a506b8c5ebe818f2ffb4ac Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 08:34:55 +0000
Subject: [PATCH 0532/2012] chore(sync): update Kilo model catalog (#6169)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.3.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index e9d5bdad891..9e9823ba910 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -12,3 +12,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
+output = 262_144

From 7b80eef3ef3d0eecffd6c0c1e562f0250408e221 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 08:37:04 +0000
Subject: [PATCH 0533/2012] chore(sync): update NanoGPT model catalog (#6170)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/abliteration-ai/abliterated-model-large-v2.toml      | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
index 0333560b0f1..2228e8805bd 100644
--- a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
@@ -4,7 +4,7 @@ release_date = "2026-08-31"
 last_updated = "2026-08-31"
 attachment = false
 reasoning = true
-tool_call = false
+tool_call = true
 structured_output = true
 open_weights = false
 

From 30c484cc0715ef6eb90c56497f9e5b8b49ff2a3a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 09:30:02 +0000
Subject: [PATCH 0534/2012] chore(sync): update Kilo model catalog (#6173)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/ibm-granite/granite-4.2-8b.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/ibm-granite/granite-4.2-8b.toml b/providers/kilo/models/ibm-granite/granite-4.2-8b.toml
index c99650d60ce..f7ead7bbbc0 100644
--- a/providers/kilo/models/ibm-granite/granite-4.2-8b.toml
+++ b/providers/kilo/models/ibm-granite/granite-4.2-8b.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.1
-output = 0.15
-cache_read = 0.05
+input = 0.06
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 131_072

From c26f5067f34682924fe9469772d07a23e44bc96f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:27:48 +0000
Subject: [PATCH 0535/2012] chore(sync): update OpenRouter model catalog
 (#6175)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-chat.toml     | 6 +++---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat.toml b/providers/openrouter/models/deepseek/deepseek-chat.toml
index a01fc1ca488..7768aadee98 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat.toml
@@ -3,9 +3,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.2574
-output = 1.0287
+input = 0.32
+output = 0.89
 
 [limit]
 context = 163_840
-output = 16_000
+output = 16_384
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index d2137e25375..4d5bf738309 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.106
-output = 3.476
-cache_read = 0.1817
+input = 1.12
+output = 3.52
+cache_read = 0.184
 
 [limit]
 context = 1_310_720

From c893b8938cadb23b95848ef88d94b54f7ecf4230 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:27:59 +0000
Subject: [PATCH 0536/2012] chore(sync): update NanoGPT model catalog (#6176)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 6aee14da1f8ba0a27eadb70734243cb44748decf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:28:19 +0000
Subject: [PATCH 0537/2012] chore(sync): update Kilo model catalog (#6174)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat.toml | 8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml       | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat.toml b/providers/kilo/models/deepseek/deepseek-chat.toml
index f67bbe80e1f..42fda83c47b 100644
--- a/providers/kilo/models/deepseek/deepseek-chat.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat.toml
@@ -4,9 +4,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.2574
-output = 1.0287
+input = 0.32
+output = 0.89
 
 [limit]
-context = 128_000
-output = 16_000
+context = 163_840
+output = 16_384
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2406d89fc54..b10dbb745b5 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.106
-output = 3.476
-cache_read = 0.1817
+input = 1.12
+output = 3.52
+cache_read = 0.184
 
 [limit]
 context = 1_048_576

From 1c36ae7a6a11d7ea1603b442ae29492c637a9897 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 12:38:06 +0000
Subject: [PATCH 0538/2012] chore(sync): update Kilo model catalog (#6182)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
index 5441601c6e9..02ea9134945 100644
--- a/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -11,12 +11,13 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.25
-output = 0.75
+input = 0.8
+output = 1
+cache_read = 0.4
 
 [limit]
-context = 32_000
-output = 28_800
+context = 128_000
+output = 115_200
 
 [modalities]
 input = ["text", "image"]

From 38d5d4bcf52d543d39b6283908ee29c53b64246a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 12:38:12 +0000
Subject: [PATCH 0539/2012] chore(sync): update Eden AI model catalog (#6181)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../accounts/fireworks/models/gpt-oss-120b.toml        | 10 ----------
 1 file changed, 10 deletions(-)
 delete mode 100644 providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-120b.toml

diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-120b.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-120b.toml
deleted file mode 100644
index 293dd1bccfd..00000000000
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/gpt-oss-120b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "openai/gpt-oss-120b"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015

From 0126a435e4fd48a34d17f11e21e4e1297cfc1627 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 12:38:29 +0000
Subject: [PATCH 0540/2012] chore(sync): update OpenRouter model catalog
 (#6183)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml    | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
index c0bba4820f6..b32e69d51d6 100644
--- a/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen2.5-vl-72b-instruct.toml
@@ -12,12 +12,13 @@ knowledge = "2024-06-30"
 open_weights = true
 
 [cost]
-input = 0.25
-output = 0.75
+input = 0.8
+output = 1
+cache_read = 0.4
 
 [limit]
 context = 128_000
-output = 28_800
+output = 115_200
 
 [modalities]
 input = ["text", "image"]

From f5b0643936cb7a708922499603d3688ef586475e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 13:27:21 +0000
Subject: [PATCH 0541/2012] chore(sync): update Kilo model catalog (#6184)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.6-35b-a3b.toml | 7 +++----
 providers/kilo/models/~z-ai/glm-latest.toml     | 6 +++---
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index 1385363812a..26d30087a8e 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,12 +6,11 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.1
-output = 0.9
-cache_read = 0.05
+input = 0.05
+output = 0.7
 
 [limit]
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index b10dbb745b5..7e40b3c4374 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12
-output = 3.52
-cache_read = 0.184
+input = 1.134
+output = 3.564
+cache_read = 0.1863
 
 [limit]
 context = 1_048_576

From 60ae53364d11a9bd64f20390143690060bc61668 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 13:27:53 +0000
Subject: [PATCH 0542/2012] chore(sync): update OpenRouter model catalog
 (#6185)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml      | 7 +++----
 providers/openrouter/models/~z-ai/glm-latest.toml          | 6 +++---
 4 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 66443ade4c4..4bf670bbd00 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.04226
-output = 2.08452
-cache_read = 0.086855
+input = 1.039302
+output = 2.078604
+cache_read = 0.086609
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index ac95b282785..29c3f360042 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,12 +6,11 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.9
-cache_read = 0.05
+input = 0.05
+output = 0.7
 
 [limit]
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 4d5bf738309..4bdfd2001cf 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12
-output = 3.52
-cache_read = 0.184
+input = 1.134
+output = 3.564
+cache_read = 0.1863
 
 [limit]
 context = 1_310_720

From 9806fe89f030011e91d379c128854da040263172 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 13:28:04 +0000
Subject: [PATCH 0543/2012] chore(sync): update NanoGPT model catalog (#6186)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ornith-ai/ornith-1.5-9b.toml       | 25 +++++++++++++++++++
 .../ornith-ai/ornith-1.5-9b:thinking.toml     | 25 +++++++++++++++++++
 .../qwen/qwen3.6-35b-a3b-uncensored.toml      | 25 +++++++++++++++++++
 .../qwen3.6-35b-a3b-uncensored:thinking.toml  | 25 +++++++++++++++++++
 4 files changed, 100 insertions(+)
 create mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
 create mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml

diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
new file mode 100644
index 00000000000..5d0b5ebe016
--- /dev/null
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
@@ -0,0 +1,25 @@
+name = "Ornith 1.5 9B"
+description = "Ornith 1.5 9B is an FP8 dense open-weight model built for agentic coding, tool use, visual understanding, and efficient long-context work. This variant disables thinking for faster direct responses."
+family = "ornith"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
new file mode 100644
index 00000000000..ebbf9c7150b
--- /dev/null
+++ b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
@@ -0,0 +1,25 @@
+name = "Ornith 1.5 9B Thinking"
+description = "Ornith 1.5 9B with thinking enabled for more deliberate agentic coding, tool use, visual understanding, and long-context work."
+family = "ornith"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
new file mode 100644
index 00000000000..bbb44da495d
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.6 35B A3B Uncensored"
+description = "Qwen 3.6 35B A3B Uncensored is an NVFP4 open-weight mixture-of-experts model LoRA-tuned for fewer refusals across chat, coding, tool use, and multimodal tasks."
+family = "qwen3.6"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.15
+output = 0.95
+cache_read = 0.075
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
new file mode 100644
index 00000000000..0cbe6601a36
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.6 35B A3B Uncensored Thinking"
+description = "Qwen 3.6 35B A3B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and complex chat tasks."
+family = "qwen3.6"
+release_date = "2026-07-29"
+last_updated = "2026-07-29"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.15
+output = 0.95
+cache_read = 0.075
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From b3e4ece63b2970470273969e3319b124b96a3011 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 14:29:13 +0000
Subject: [PATCH 0544/2012] chore(sync): update LLM Gateway model catalog
 (#6187)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/Qwen3.8-27B.toml | 27 +++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
new file mode 100644
index 00000000000..b5674abce63
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
@@ -0,0 +1,27 @@
+name = "Qwen3.8 27B (Consensus Protocol)"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.41
+output = 2.5
+cache_read = 0.08
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

From ae3d1cd509f61c56fec51a9b92d3d6e882a3eb3a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 14:29:16 +0000
Subject: [PATCH 0545/2012] chore(sync): update Merge Gateway model catalog
 (#6188)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..faa70b7c1e0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -11,8 +12,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 53944cf8bd773a1ab138f2693dafd9c6526fe9ee Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:25:23 -0500
Subject: [PATCH 0546/2012] fix: [missing-model] meta:
 muse-spark-1.3-contributor (#6144)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 models/meta/muse-spark-1.3.toml               | 24 +++++++++++++++++++
 .../models/muse-spark-1.3-contributor.toml    | 16 +++++++++++++
 2 files changed, 40 insertions(+)
 create mode 100644 models/meta/muse-spark-1.3.toml
 create mode 100644 providers/meta/models/muse-spark-1.3-contributor.toml

diff --git a/models/meta/muse-spark-1.3.toml b/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..6cb35e84582
--- /dev/null
+++ b/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,24 @@
+# Sources:
+# https://research.meta.ai/blog/introducing-muse-spark-1-3
+# https://dev.meta.ai/docs/models
+# https://openrouter.ai/meta/muse-spark-1.3 (OpenRouter Meta-hosted catalog snapshot, 2026-09-02)
+
+name = "Muse Spark 1.3"
+description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It improves long-horizon agent collaboration, instruction following, and coding efficiency relative to Muse Spark 1.2."
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf", "audio"]
+output = ["text"]
diff --git a/providers/meta/models/muse-spark-1.3-contributor.toml b/providers/meta/models/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..264b2fada97
--- /dev/null
+++ b/providers/meta/models/muse-spark-1.3-contributor.toml
@@ -0,0 +1,16 @@
+# Sources:
+# https://research.meta.ai/blog/introducing-muse-spark-1-3
+# https://dev.meta.ai/docs/models
+# https://dev.meta.ai/docs/pricing-rate-limits
+# https://openrouter.ai/meta/muse-spark-1.3-contributor (OpenRouter Meta-hosted catalog snapshot, 2026-09-02)
+# Contributor tier prompts and completions may be used to improve Meta's products.
+
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 Contributor"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.10
+output = 0.20
+cache_read = 0.002

From 4685712c3fe56b3d120a074d2c81226139c4920f Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:26:26 -0500
Subject: [PATCH 0547/2012] fix: [missing-model] ofox: bailian/qwen3.8-max-0902
 (#6179)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 models/alibaba/qwen3.8-max-0902.toml          | 28 +++++++++++++++++++
 .../ofox/models/bailian/qwen3.8-max-0902.toml | 20 +++++++++++++
 2 files changed, 48 insertions(+)
 create mode 100644 models/alibaba/qwen3.8-max-0902.toml
 create mode 100644 providers/ofox/models/bailian/qwen3.8-max-0902.toml

diff --git a/models/alibaba/qwen3.8-max-0902.toml b/models/alibaba/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..aeb8160175c
--- /dev/null
+++ b/models/alibaba/qwen3.8-max-0902.toml
@@ -0,0 +1,28 @@
+# Sources (accessed 2026-09-03):
+# https://www.qwencloud.com/models/qwen3.8-max-0902
+# https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Snapshot of qwen3.8-max (alias qwen3.8-max-2026-09-02). Same limits,
+# hybrid thinking, modalities, and tool ecosystem as qwen3.8-max; post-trained
+# for stronger coding, collaborative agents, and vision/document understanding.
+# PDF input: inherits Completions PDF理解 / document parsing from the Max line
+# (same as models/alibaba/qwen3.8-max.toml).
+
+name = "Qwen3.8 Max 0902"
+description = "2026-09-02 upgraded snapshot of Qwen3.8 Max with stronger coding, collaborative agents, and multimodal document understanding"
+family = "qwen"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]
diff --git a/providers/ofox/models/bailian/qwen3.8-max-0902.toml b/providers/ofox/models/bailian/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..bcf8deebcd4
--- /dev/null
+++ b/providers/ofox/models/bailian/qwen3.8-max-0902.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-03):
+# https://ofox.ai/models/bailian/qwen3.8-max-0902
+# https://www.qwencloud.com/models/qwen3.8-max-0902
+# https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Toggle: enable_thinking true|false (hybrid; same Bailian/DashScope surface as ofox bailian/qwen3.8-max)
+# Effort: reasoning_effort = low|medium|xhigh (default xhigh); high accepted as alias → xhigh
+# Budget: thinking_budget (0..262144) cannot be combined with reasoning_effort
+base_model = "alibaba/qwen3.8-max-0902"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", min = 0, max = 262_144 },
+]
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.25
+cache_write = 2.5

From b890135233df27ad12e25de6542121e3bbed9ceb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rub=C3=A9n=20Solano?=
 <41839257+rsolanodev@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:27:25 +0200
Subject: [PATCH 0548/2012] feat(nan): add NaN provider with its LLM catalog
 (#6172)

* feat(nan): add NaN provider with its LLM catalog

NaN is a builders community that shares dedicated GPUs to serve
open-weight models through LiteLLM with an OpenAI-compatible API at
https://api.nan.builders/v1.

Adds the seven LLMs of the cluster, all via base_model since NaN is a
relay and not the lab: deepseek-v4-flash, mimo-v2.5, glm5.2,
qwen3.8-flash, glm5.3-flash, gemma4 and qwen3.6. The embedding, rerank,
TTS/STT and image models of the cluster are out of scope here.

Access is membership based (70 EUR/month, plus a 200 EUR/month tier that
unlocks glm5.2) and metered as monthly token quotas rather than billed
per token, so cost is 0 -- same convention as other plan-based hosts in
the catalog. Quotas and tpm limits are recorded in each file header.

NaN documents a reasoning mode on every LLM but publishes a request
field for it only on gemma4 and qwen3.6 (reasoning_config={}), and no
graded effort levels anywhere, so each model carries a toggle taken from
the lab entry and relay peers of the underlying model.

Sources: https://nan.builders/docs/models and
https://nan.builders/docs/getting-started (accessed 2026-09-03).
Co-Authored-By: Claude <noreply@anthropic.com>

* fix(nan): inherit the native context of mimo-v2.5

The file overrode limit.context to 1_000_000 against the 1_048_576 of
the lab entry. NaN publishes "1M tokens" for this model with the same
wording it uses for deepseek-v4-flash and glm5.3-flash, and documents no
trim of the native window, so the override asserted a 48_576 token cut
with nothing behind it. Inherit instead, and record the reasoning in the
file header.

Co-Authored-By: Claude <noreply@anthropic.com>

* fix(nan): record no caller reasoning control where NaN documents none

Review flagged five models as authoring a toggle that this host does not
expose. The headers themselves admitted NaN publishes no reasoning
request field for them, and AGENTS.md requires every toggle to name its
exact wire path, so the option was a lab/peer control that does not
reach this surface.

Set reasoning_options = [] on deepseek-v4-flash, mimo-v2.5, glm5.2,
qwen3.8-flash and glm5.3-flash, and reworded each header to state the
absence of a control affirmatively instead of hedging.

gemma4 and qwen3.6 keep their toggle: `reasoning_config={}` is the one
reasoning field NaN documents, and it is named in their leading comment.

Co-Authored-By: Claude <noreply@anthropic.com>

---------

Co-authored-by: Claude <noreply@anthropic.com>
---
 providers/nan/logo.svg                      |  1 +
 providers/nan/models/deepseek-v4-flash.toml | 16 ++++++++++++++++
 providers/nan/models/gemma4.toml            | 14 ++++++++++++++
 providers/nan/models/glm5.2.toml            | 18 ++++++++++++++++++
 providers/nan/models/glm5.3-flash.toml      | 16 ++++++++++++++++
 providers/nan/models/mimo-v2.5.toml         | 21 +++++++++++++++++++++
 providers/nan/models/qwen3.6.toml           | 19 +++++++++++++++++++
 providers/nan/models/qwen3.8-flash.toml     | 18 ++++++++++++++++++
 providers/nan/provider.toml                 | 11 +++++++++++
 9 files changed, 134 insertions(+)
 create mode 100644 providers/nan/logo.svg
 create mode 100644 providers/nan/models/deepseek-v4-flash.toml
 create mode 100644 providers/nan/models/gemma4.toml
 create mode 100644 providers/nan/models/glm5.2.toml
 create mode 100644 providers/nan/models/glm5.3-flash.toml
 create mode 100644 providers/nan/models/mimo-v2.5.toml
 create mode 100644 providers/nan/models/qwen3.6.toml
 create mode 100644 providers/nan/models/qwen3.8-flash.toml
 create mode 100644 providers/nan/provider.toml

diff --git a/providers/nan/logo.svg b/providers/nan/logo.svg
new file mode 100644
index 00000000000..04bead3f0e2
--- /dev/null
+++ b/providers/nan/logo.svg
@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 100 100"><path d="M54.33 11.5 L81.17 27 Q85.5 29.5 85.5 34.5 L85.5 65.5 Q85.5 70.5 81.17 73 L54.33 88.5 Q50 91 45.67 88.5 L18.83 73 Q14.5 70.5 14.5 65.5 L14.5 34.5 Q14.5 29.5 18.83 27 L45.67 11.5 Q50 9 54.33 11.5 Z" fill="none" stroke="currentColor" stroke-width="4.5"></path><path d="M24 50 C24 42.5 29.5 38 36 38 C43 38 46.5 43.5 50 50 C53.5 56.5 57 62 64 62 C70.5 62 76 57.5 76 50 C76 42.5 70.5 38 64 38 C57 38 53.5 43.5 50 50 C46.5 56.5 43 62 36 62 C29.5 62 24 57.5 24 50 Z" fill="none" stroke="currentColor" stroke-width="4.2"></path></svg>
diff --git a/providers/nan/models/deepseek-v4-flash.toml b/providers/nan/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..f232549beb2
--- /dev/null
+++ b/providers/nan/models/deepseek-v4-flash.toml
@@ -0,0 +1,16 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 284B-21B MoE, FP8, 1M context. 2B token monthly quota per member, 1.5M tpm.
+# Served from the shared community GPU cluster, so no per-token price: included
+# in the membership.
+# Reasoning: no caller control on this surface. NaN lists a reasoning mode but
+# exposes no request field for it anywhere in its docs (only gemma4/qwen3.6 show
+# `reasoning_config={}`) and no graded effort, so neither the DeepSeek toggle nor
+# its low|high|max levels reach this host.
+# The docs/examples opencode.json snippet caps contextWindow at 500000; the
+# models reference publishes 1M, which is what is recorded here.
+base_model = "deepseek/deepseek-v4-flash"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/nan/models/gemma4.toml b/providers/nan/models/gemma4.toml
new file mode 100644
index 00000000000..35f7fb1e841
--- /dev/null
+++ b/providers/nan/models/gemma4.toml
@@ -0,0 +1,14 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 26B-A4B MoE, FP8, 256K context, vision, XML tool calling. No token counter,
+# 1.5M tpm. Sampling defaults temp=0.6, top_p=0.95.
+# Included in the membership, so no per-token price.
+# Toggle: `reasoning_config={}` in the request body enables reasoning mode --
+# the only reasoning field NaN documents. No graded effort.
+base_model = "google/gemma-4-26b-a4b-it"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/nan/models/glm5.2.toml b/providers/nan/models/glm5.2.toml
new file mode 100644
index 00000000000..876a00ec212
--- /dev/null
+++ b/providers/nan/models/glm5.2.toml
@@ -0,0 +1,18 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# ~753B MoE, FP8, sparse attention, text only. Callable only with a key on the
+# 200 EUR/month GLM 5.2 premium membership: 3,000M tokens per billing period and
+# 400M tokens per rolling 4h window, 5 concurrent requests. Context is 500K on
+# this host, not the 1M of the weights. Included in the tier, so no per-token price.
+# Reasoning: no caller control on this surface. NaN documents a reasoning trace
+# but no request field for it, and the high|max effort levels of the first-party
+# Z.ai API are not exposed here.
+base_model = "zhipuai/glm-5.2"
+description = "Open flagship GLM for long-horizon coding agents, served at 500K context on the NaN premium tier"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 500_000
diff --git a/providers/nan/models/glm5.3-flash.toml b/providers/nan/models/glm5.3-flash.toml
new file mode 100644
index 00000000000..550bda3827e
--- /dev/null
+++ b/providers/nan/models/glm5.3-flash.toml
@@ -0,0 +1,16 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 320B-18B MoE, FP8, natively multimodal. 1M context. 500M token monthly quota
+# per member. MIT license. No video or PDF input on this host.
+# Included in the membership, so no per-token price.
+# Reasoning: no caller control on this surface. NaN lists a reasoning mode but
+# exposes no request field for it, and the low|high|max effort levels of the
+# first-party Z.ai API are not exposed here.
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/nan/models/mimo-v2.5.toml b/providers/nan/models/mimo-v2.5.toml
new file mode 100644
index 00000000000..882287acf30
--- /dev/null
+++ b/providers/nan/models/mimo-v2.5.toml
@@ -0,0 +1,21 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 310B-15B MoE, FP8, natively omnimodal with dedicated vision and audio encoders.
+# 1M context. 1.0B token monthly quota per member, 1.5M tpm. MIT license.
+# No video input on this host. Included in the membership, so no per-token price.
+# Reasoning: no caller control on this surface. NaN lists a reasoning mode
+# (recommends max_tokens >= 300) but exposes no request field for it and no
+# graded effort.
+# Context is inherited: NaN publishes "1M tokens", the same wording it uses for
+# deepseek-v4-flash and glm5.3-flash, with no documented trim of the model's
+# native 1_048_576 window. The docs/examples opencode.json snippet caps
+# contextWindow at 500000, but that snippet also trims maxTokens to a
+# per-request value rather than the ceiling.
+base_model = "xiaomi/mimo-v2.5"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[modalities]
+input = ["text", "image", "audio"]
diff --git a/providers/nan/models/qwen3.6.toml b/providers/nan/models/qwen3.6.toml
new file mode 100644
index 00000000000..fb47cc0b56a
--- /dev/null
+++ b/providers/nan/models/qwen3.6.toml
@@ -0,0 +1,19 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 35B total MoE, 3B active per token, FP8, 256K context, vision, XML tool
+# calling. MTP speculative decoding (~2x throughput). No token counter, 1.5M tpm.
+# Sampling defaults temp=0.6, top_p=0.95. NaN's own client configs confirm
+# maxTokens 65536 as the model ceiling. Included in the membership, so no
+# per-token price.
+# Toggle: `reasoning_config={}` in the request body enables reasoning mode --
+# the only reasoning field NaN documents. No graded effort.
+base_model = "alibaba/qwen3.6-35b-a3b"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/nan/models/qwen3.8-flash.toml b/providers/nan/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..dd94ab609d8
--- /dev/null
+++ b/providers/nan/models/qwen3.8-flash.toml
@@ -0,0 +1,18 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# 125B-6B MoE with vision, XML tool calling. 262K context, the model's native
+# window. 1B token monthly quota per member. License qwen-community-1.0.
+# Included in the membership, so no per-token price.
+# Reasoning: on by default, with no caller control on this surface. NaN exposes
+# no request field to disable it and no graded effort.
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/nan/provider.toml b/providers/nan/provider.toml
new file mode 100644
index 00000000000..6359c2fb7e6
--- /dev/null
+++ b/providers/nan/provider.toml
@@ -0,0 +1,11 @@
+# NaN is a builders community that shares dedicated GPUs to serve open-weight models.
+# Access is via LiteLLM with an OpenAI-compatible API at https://api.nan.builders/v1.
+# Membership based (70 EUR/month, plus a 200 EUR/month tier that unlocks glm5.2), so
+# per-token cost is 0: usage is metered as monthly token quotas, not billed per token.
+# Rate limits: 60 rpm and 5 concurrent requests per API key.
+# https://nan.builders/docs/getting-started (accessed 2026-09-03)
+name = "NaN"
+env = ["NAN_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.nan.builders/v1"
+doc = "https://nan.builders/docs/models"

From ef02c2d5d41514b5a5c7c281930344c7b38d3500 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:27:38 -0500
Subject: [PATCH 0549/2012] fix: [missing-model] tinfoil: glm-5-3-flash (#6161)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/glm-5-3-flash.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/tinfoil/models/glm-5-3-flash.toml

diff --git a/providers/tinfoil/models/glm-5-3-flash.toml b/providers/tinfoil/models/glm-5-3-flash.toml
new file mode 100644
index 00000000000..9762fb93463
--- /dev/null
+++ b/providers/tinfoil/models/glm-5-3-flash.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-03):
+# - https://inference.tinfoil.sh/v1/models
+# - https://docs.tinfoil.sh/models/chat.md (weights RedHatAI/GLM-5.3-Flash-NVFP4)
+# - https://docs.tinfoil.sh/guides/reasoning.md
+# Effort: reasoning_effort = low|high|max on Tinfoil; reasoning is always on (no none/toggle).
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.4
+output = 1.25
+cache_read = 0.1
+
+[limit]
+context = 1_048_576

From fb4a06f6f65ed07ea6ec3a2debc22e92baacd168 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:27:59 -0500
Subject: [PATCH 0550/2012] fix: [missing-model] meta: muse-spark-1.3 (#6143)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/meta/models/muse-spark-1.3.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/meta/models/muse-spark-1.3.toml

diff --git a/providers/meta/models/muse-spark-1.3.toml b/providers/meta/models/muse-spark-1.3.toml
new file mode 100644
index 00000000000..d2c1f6a4e45
--- /dev/null
+++ b/providers/meta/models/muse-spark-1.3.toml
@@ -0,0 +1,15 @@
+# Sources:
+# https://research.meta.ai/blog/introducing-muse-spark-1-3
+# https://dev.meta.ai/docs/pricing-rate-limits/
+# https://openrouter.ai/api/v1/models/meta/muse-spark-1.3/endpoints (Meta endpoint pricing; accessed 2026-09-02)
+# Standard-tier rates match Muse Spark 1.2; Axios reports 1.3 priced the same as predecessor.
+# Effort: reasoning_effort = minimal|low|medium|high|xhigh (max not yet live per Meta launch post)
+
+base_model = "meta/muse-spark-1.3"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15

From 544de959ba238e9693d46763f9026b8f76eb1b35 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:28:28 -0500
Subject: [PATCH 0551/2012] chore(sync): update Hugging Face model catalog
 (#6178)

* chore(sync): update Hugging Face model catalog

* fix(huggingface): correct DeepSeek V4 Flash Vision reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/huggingface/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml

diff --git a/providers/huggingface/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml b/providers/huggingface/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
new file mode 100644
index 00000000000..b6175614f9f
--- /dev/null
+++ b/providers/huggingface/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
@@ -0,0 +1,15 @@
+# HF relay: reasoning_effort = low|high|max, matching the native model and relay peers.
+# HF documents no shared thinking toggle or numeric reasoning budget.
+# https://huggingface.co/docs/inference-providers/en/tasks/chat-completion (accessed 2026-09-03)
+# https://api-docs.deepseek.com/guides/thinking_mode (accessed 2026-09-03)
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+open_weights = true
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 0.44
+output = 1.32
+
+[limit]
+context = 1_048_576

From 3acbeebfe843f397a2a5c53430bc8d9124d7cae5 Mon Sep 17 00:00:00 2001
From: MicroHEROX <140800253+MicroHEROX@users.noreply.github.com>
Date: Thu, 3 Sep 2026 23:29:24 +0800
Subject: [PATCH 0552/2012] feat: add AMD DeepSeek-V4-Flash-Vision-Exp and
 MiniCPM5-1B (#6155)

* test write access

* feat: add AMD DeepSeek-V4-Flash-Vision-Exp

* feat: add OpenBMB MiniCPM5-1B lab metadata

* feat: add AMD MiniCPM5-1B
---
 models/openbmb/minicpm5-1b.toml               | 25 +++++++++++++++++++
 .../models/DeepSeek-V4-Flash-Vision-Exp.toml  | 16 ++++++++++++
 providers/amd/models/MiniCPM5-1B.toml         | 17 +++++++++++++
 3 files changed, 58 insertions(+)
 create mode 100644 models/openbmb/minicpm5-1b.toml
 create mode 100644 providers/amd/models/DeepSeek-V4-Flash-Vision-Exp.toml
 create mode 100644 providers/amd/models/MiniCPM5-1B.toml

diff --git a/models/openbmb/minicpm5-1b.toml b/models/openbmb/minicpm5-1b.toml
new file mode 100644
index 00000000000..01f66f89903
--- /dev/null
+++ b/models/openbmb/minicpm5-1b.toml
@@ -0,0 +1,25 @@
+# Sources (accessed 2026-09-03):
+# https://huggingface.co/openbmb/MiniCPM5-1B (model card, config)
+# https://github.com/OpenBMB/MiniCPM (release announcement 2026-05-19)
+name = "MiniCPM5-1B"
+description = "Dense 1B-class open-source model for on-device and resource-constrained use, with native long-context support, Think / No Think chat modes, and tool calling"
+release_date = "2026-05-19"
+last_updated = "2026-05-19"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+license = "apache-2.0"
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/openbmb/MiniCPM5-1B"
\ No newline at end of file
diff --git a/providers/amd/models/DeepSeek-V4-Flash-Vision-Exp.toml b/providers/amd/models/DeepSeek-V4-Flash-Vision-Exp.toml
new file mode 100644
index 00000000000..fdf5436819c
--- /dev/null
+++ b/providers/amd/models/DeepSeek-V4-Flash-Vision-Exp.toml
@@ -0,0 +1,16 @@
+# Pricing and reasoning controls verified against the AMD Token Factory API
+# (accessed 2026-09-03): GET /v1/models and live POST /v1/chat/completions.
+# reasoning_effort is a validated enum none|minimal|low|medium|high|xhigh|max
+# (invalid values return 400) and affects generation; no reasoning_content is
+# ever returned (stream + non-stream), so no toggle and no interleaved channel.
+# Vision verified live: image input is accepted and understood.
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+reasoning_options = [
+  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
\ No newline at end of file
diff --git a/providers/amd/models/MiniCPM5-1B.toml b/providers/amd/models/MiniCPM5-1B.toml
new file mode 100644
index 00000000000..09d14d672f9
--- /dev/null
+++ b/providers/amd/models/MiniCPM5-1B.toml
@@ -0,0 +1,17 @@
+# Pricing and reasoning controls verified against the AMD Token Factory API
+# (accessed 2026-09-03): GET /v1/models and live POST /v1/chat/completions.
+# reasoning_effort accepts low|medium|high on this host (minimal, xhigh, and
+# max return 400). none is accepted but does not disable thinking - the model
+# keeps emitting reasoning (as <think> text in the content stream), and no
+# reasoning_content is ever returned, so there is no toggle and no
+# interleaved channel on this host.
+base_model = "openbmb/minicpm5-1b"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 0.124
+output = 0.7425
+cache_read = 0.124
\ No newline at end of file

From f6fdced5f6630b0cedf6f3a3dd1882fc04d63911 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 15:29:35 +0000
Subject: [PATCH 0553/2012] chore(sync): update NanoGPT model catalog (#6191)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-max-0902.toml         | 15 +--------------
 .../nano-gpt/models/meta/muse-spark-1.3.toml     | 16 +---------------
 2 files changed, 2 insertions(+), 29 deletions(-)

diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
index d1e3bd8866a..a8fda51f1ef 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
+++ b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
@@ -1,13 +1,5 @@
-name = "Qwen3.8 Max 0902"
-description = "Qwen3.8 Max 0902 is Alibaba's September 2 checkpoint of its flagship Qwen3.8 Max model for coding, knowledge work, data analysis, and long-running agent workflows. It supports text, image, video, PDF input, selectable thinking, tool calling, structured output, and a near-million-token context window."
-family = "qwen3.8-max"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-tool_call = true
+base_model = "alibaba/qwen3.8-max-0902"
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -22,8 +14,3 @@ cache_write = 2.5
 [limit]
 context = 991_808
 input = 991_808
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "pdf"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/meta/muse-spark-1.3.toml b/providers/nano-gpt/models/meta/muse-spark-1.3.toml
index 75ee44dfd23..7e3eca1e095 100644
--- a/providers/nano-gpt/models/meta/muse-spark-1.3.toml
+++ b/providers/nano-gpt/models/meta/muse-spark-1.3.toml
@@ -1,13 +1,4 @@
-name = "Muse Spark 1.3"
-description = "Meta's Muse Spark 1.3 is a frontier multimodal reasoning model for long-horizon coding and agentic workflows, with strong gains in computer use, browsing, professional tool use, codebase understanding, instruction following, and million-token retrieval. It accepts text, images, audio, video, and files, supports tool calling and structured output, and always reasons before answering."
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
+base_model = "meta/muse-spark-1.3"
 
 [[reasoning_options]]
 type = "effort"
@@ -19,10 +10,5 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 input = 1_048_576
 output = 943_718
-
-[modalities]
-input = ["text", "image", "video", "audio", "pdf"]
-output = ["text"]

From b5f8143fc966c658cd45a6401f00ea5435b7bcfe Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:29:37 -0500
Subject: [PATCH 0554/2012] fix: [missing-model] ofox:
 anthropic/claude-fable-5.1 (#6154)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../ofox/models/anthropic/claude-fable-5.1.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/ofox/models/anthropic/claude-fable-5.1.toml b/providers/ofox/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..1471ba3d420
--- /dev/null
+++ b/providers/ofox/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,17 @@
+# Sources:
+# - https://ofox.ai/models/anthropic/claude-fable-5.1 (pricing, context, capabilities; accessed 2026-09-03)
+# - https://api.ofox.ai/v2/models/catalog (remote id anthropic/claude-fable-5.1)
+# - Lab: models/anthropic/claude-fable-5-1.toml; first-party effort: providers/anthropic/models/claude-fable-5-1.toml
+# Effort: Anthropic adaptive thinking effort on Ofox Anthropic protocol (same surface as other ofox anthropic/* entries)
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.ofox.ai/anthropic/v1"

From 033edf0fc7a0b3b08b3bac9580e638c7b6bd2ab8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 15:29:51 +0000
Subject: [PATCH 0555/2012] chore(sync): update Eden AI model catalog (#6190)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepinfra/stepfun-ai/Step-3.7-Flash.toml       |  2 +-
 .../edenai/models/google/gemini-3.8-flash.toml     | 13 +++++++++++++
 .../ionos/meta-llama/Llama-3.3-70B-Instruct.toml   |  4 ++--
 .../edenai/models/ionos/openai/gpt-oss-120b.toml   |  4 ++--
 providers/edenai/models/qwen/qwen3.8-flash.toml    |  2 +-
 .../models/scaleway/deepseek-v4-flash-0731.toml    |  4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml |  4 ++--
 .../models/scaleway/llama-3.3-70b-instruct.toml    |  4 ++--
 .../edenai/models/vertex/gemini-3.8-flash.toml     | 13 +++++++++++++
 .../edenai/models/vertex/gemini-3.8-flash@eu.toml  | 14 ++++++++++++++
 .../edenai/models/vertex/gemini-3.8-flash@us.toml  | 14 ++++++++++++++
 11 files changed, 66 insertions(+), 12 deletions(-)
 create mode 100644 providers/edenai/models/google/gemini-3.8-flash.toml
 create mode 100644 providers/edenai/models/vertex/gemini-3.8-flash.toml
 create mode 100644 providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
 create mode 100644 providers/edenai/models/vertex/gemini-3.8-flash@us.toml

diff --git a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
index 123c6bc463b..d395a094601 100644
--- a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
+++ b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
@@ -1,6 +1,6 @@
 base_model = "stepfun/step-3.7-flash"
 base_model_omit = ["limit.input"]
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/google/gemini-3.8-flash.toml b/providers/edenai/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..edad81f8322
--- /dev/null
+++ b/providers/edenai/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,13 @@
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+input_audio = 1.5
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 465b5f9eb20..f41860fe6d8 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75257
-output = 0.75257
+input = 0.754975
+output = 0.754975
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 46d0db0b3d5..e453aff36f2 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17367
-output = 0.75257
+input = 0.174225
+output = 0.754975
diff --git a/providers/edenai/models/qwen/qwen3.8-flash.toml b/providers/edenai/models/qwen/qwen3.8-flash.toml
index 7b6150ebeeb..3432e836e33 100644
--- a/providers/edenai/models/qwen/qwen3.8-flash.toml
+++ b/providers/edenai/models/qwen/qwen3.8-flash.toml
@@ -5,7 +5,7 @@ type = "effort"
 values = ["none", "low", "medium", "xhigh"]
 
 [cost]
-input = 0.16
+input = 0.15
 output = 0.47
 cache_read = 0.016
 cache_write = 0.2
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 34338102b03..98720232880 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46312
-output = 0.92624
+input = 0.4646
+output = 0.9292
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index e17608fe294..b3e11f0e545 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17367
-output = 0.69468
+input = 0.174225
+output = 0.6969
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 818df2948ac..50d0c19815f 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04202
-output = 1.04202
+input = 1.04535
+output = 1.04535
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash.toml b/providers/edenai/models/vertex/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..edad81f8322
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.8-flash.toml
@@ -0,0 +1,13 @@
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+input_audio = 1.5
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
new file mode 100644
index 00000000000..0937543bf13
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.8-flash"
+name = "Gemini 3.8 Flash (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+input_audio = 1.5
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@us.toml b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
new file mode 100644
index 00000000000..9e5f2a9b227
--- /dev/null
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.8-flash"
+name = "Gemini 3.8 Flash (US)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+reasoning = 7.5
+cache_read = 0.15
+cache_write = 0.083333
+input_audio = 1.5

From c6eeb4e9852244b587c8b0c92a73902e62c35ed4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 15:30:07 +0000
Subject: [PATCH 0556/2012] chore(sync): update EmpirioLabs AI model catalog
 (#6194)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/muse-spark-1-3.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/empiriolabs/models/muse-spark-1-3.toml

diff --git a/providers/empiriolabs/models/muse-spark-1-3.toml b/providers/empiriolabs/models/muse-spark-1-3.toml
new file mode 100644
index 00000000000..3e6d5ba8d7d
--- /dev/null
+++ b/providers/empiriolabs/models/muse-spark-1-3.toml
@@ -0,0 +1,13 @@
+base_model = "meta/muse-spark-1.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 1
+
+[modalities]
+input = ["text", "image", "video"]

From 567ead50dbf26e916678fdb9c7f881d8b33e7c12 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 15:30:15 +0000
Subject: [PATCH 0557/2012] chore(sync): update Kilo model catalog (#6192)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta/muse-spark-1.3.toml  | 16 +---------------
 providers/kilo/models/qwen/qwen3.6-35b-a3b.toml |  7 ++++---
 2 files changed, 5 insertions(+), 18 deletions(-)

diff --git a/providers/kilo/models/meta/muse-spark-1.3.toml b/providers/kilo/models/meta/muse-spark-1.3.toml
index 59644af30d8..29b23e8ab3a 100644
--- a/providers/kilo/models/meta/muse-spark-1.3.toml
+++ b/providers/kilo/models/meta/muse-spark-1.3.toml
@@ -1,14 +1,5 @@
-name = "Meta: Muse Spark 1.3"
+base_model = "meta/muse-spark-1.3"
 description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It is designed to keep track of information across extended tasks, work through..."
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,9 +11,4 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 943_718
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index 26d30087a8e..1385363812a 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,11 +6,12 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.05
-output = 0.7
+input = 0.1
+output = 0.9
+cache_read = 0.05
 
 [limit]
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From 52790f5e3e4123c9200b24a4dfde282f294fb50b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 15:30:22 +0000
Subject: [PATCH 0558/2012] chore(sync): update OpenRouter model catalog
 (#6193)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml       |  6 +++---
 .../models/deepseek/deepseek-v4-pro-0813.toml    |  6 +++---
 .../models/deepseek/deepseek-v4-pro.toml         |  6 +++---
 .../openrouter/models/meta/muse-spark-1.3.toml   | 16 +---------------
 .../openrouter/models/qwen/qwen3.6-35b-a3b.toml  |  7 ++++---
 5 files changed, 14 insertions(+), 27 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4bf670bbd00..97904bfc0b8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.039302
-output = 2.078604
-cache_read = 0.086609
+input = 1.035996
+output = 2.071992
+cache_read = 0.086333
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta/muse-spark-1.3.toml b/providers/openrouter/models/meta/muse-spark-1.3.toml
index 1349149c70b..f5242622051 100644
--- a/providers/openrouter/models/meta/muse-spark-1.3.toml
+++ b/providers/openrouter/models/meta/muse-spark-1.3.toml
@@ -1,14 +1,5 @@
-name = "Muse Spark 1.3"
+base_model = "meta/muse-spark-1.3"
 description = "Open Llama multimodal model for image understanding and text reasoning"
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,9 +11,4 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 943_718
-
-[modalities]
-input = ["text", "image", "video", "pdf", "audio"]
-output = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index 29c3f360042..ac95b282785 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,11 +6,12 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.05
-output = 0.7
+input = 0.1
+output = 0.9
+cache_read = 0.05
 
 [limit]
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From 391db6d395a9d9af10d8a9ef19911b02fc9713e2 Mon Sep 17 00:00:00 2001
From: albertodepaola <albertodepaola@gmail.com>
Date: Thu, 3 Sep 2026 08:31:20 -0700
Subject: [PATCH 0559/2012] Update context and output limits in
 muse-spark-1.1.toml (#6059)

---
 models/meta/muse-spark-1.1.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/models/meta/muse-spark-1.1.toml b/models/meta/muse-spark-1.1.toml
index 8ddf89f2815..d9e527e59b0 100644
--- a/models/meta/muse-spark-1.1.toml
+++ b/models/meta/muse-spark-1.1.toml
@@ -11,8 +11,8 @@ structured_output = true
 open_weights = false
 
 [limit]
-context = 1_000_000
-output = 32_000
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text", "image", "pdf", "video"]

From 49c23c0bf571e3ea5cb33d35a45f100b35ba781f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:32:35 -0500
Subject: [PATCH 0560/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6159)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): correct Qwen, Muse and GLM reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-max-0902.toml      | 20 +++++++++----------
 .../vercel/models/meta/muse-spark-1.3.toml    | 16 +++++----------
 providers/vercel/models/zai/glm-5.3-fast.toml | 14 +++++++++++++
 providers/vercel/models/zai/glm-5.3.toml      |  6 +++---
 4 files changed, 32 insertions(+), 24 deletions(-)
 create mode 100644 providers/vercel/models/zai/glm-5.3-fast.toml

diff --git a/providers/vercel/models/alibaba/qwen3.8-max-0902.toml b/providers/vercel/models/alibaba/qwen3.8-max-0902.toml
index e1289bef6ef..f5cd157fc68 100644
--- a/providers/vercel/models/alibaba/qwen3.8-max-0902.toml
+++ b/providers/vercel/models/alibaba/qwen3.8-max-0902.toml
@@ -1,13 +1,14 @@
-name = "Qwen3.8 Max 0902"
+# Toggle: reasoning.enabled = true|false
+# Effort and budget match Qwen3.8 Max; reasoning.effort and reasoning.max_tokens are mutually exclusive.
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+base_model = "alibaba/qwen3.8-max-0902"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen3.8-max"
-release_date = "2026-09-01"
-last_updated = "2026-09-01"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", min = 0, max = 262_144 },
+]
 
 [cost]
 input = 2
@@ -21,4 +22,3 @@ output = 128_000
 
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/vercel/models/meta/muse-spark-1.3.toml b/providers/vercel/models/meta/muse-spark-1.3.toml
index 1ef5327f57f..13867dbedd3 100644
--- a/providers/vercel/models/meta/muse-spark-1.3.toml
+++ b/providers/vercel/models/meta/muse-spark-1.3.toml
@@ -1,13 +1,9 @@
-name = "Muse Spark 1.3"
+# Effort levels match Meta's native model and established relay peers; max is not yet available.
+# https://research.meta.ai/blog/introducing-muse-spark-1-3
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+base_model = "meta/muse-spark-1.3"
 description = "Open Llama multimodal model for image understanding and text reasoning"
-family = "muse"
-release_date = "2026-09-02"
-last_updated = "2026-09-02"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 1.25
@@ -15,9 +11,7 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 1_048_576
 
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/vercel/models/zai/glm-5.3-fast.toml b/providers/vercel/models/zai/glm-5.3-fast.toml
new file mode 100644
index 00000000000..734583acb2b
--- /dev/null
+++ b/providers/vercel/models/zai/glm-5.3-fast.toml
@@ -0,0 +1,14 @@
+# GLM-5.3 always reasons; effort levels match the native model and Vercel's standard GLM-5.3 route.
+# https://docs.z.ai/guides/llm/glm-5.3
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 Fast"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 2.1
+output = 6.6
+cache_read = 0.21
+
+[limit]
+context = 1_048_576
+output = 262_144
diff --git a/providers/vercel/models/zai/glm-5.3.toml b/providers/vercel/models/zai/glm-5.3.toml
index 05cdccc7976..ca69c65568b 100644
--- a/providers/vercel/models/zai/glm-5.3.toml
+++ b/providers/vercel/models/zai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.14
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
 output = 1_000_000

From 23e70734e07c0d19a53a11bceee53016422203f7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:32:57 -0500
Subject: [PATCH 0561/2012] chore(sync): update Deep Infra model catalog
 (#6157)

* chore(sync): update Deep Infra model catalog

* fix(deepinfra): correct DeepSeek V4 Flash Vision reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../DeepSeek-V4-Flash-Vision-Exp.toml          | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml

diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
new file mode 100644
index 00000000000..510ea23b43c
--- /dev/null
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
@@ -0,0 +1,18 @@
+# Effort: reasoning_effort = none|low|high|max; none disables reasoning.
+# Graded levels match native DeepSeek; reasoning.enabled = false is equivalent to none.
+# https://docs.deepinfra.com/chat/reasoning
+# https://api.deepinfra.com/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp/schema/openai-chat-completions
+# https://api-docs.deepseek.com/guides/thinking_mode
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.44
+output = 1.32
+cache_read = 0.14
+
+[limit]
+context = 1_048_576

From 618ace88fc173adc05e32e972865ccde58770796 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:36:42 -0500
Subject: [PATCH 0562/2012] fix(sync): normalize Anthropic pricing table
 headers (#6195)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/anthropic.ts | 12 ++++----
 packages/core/test/sync.test.ts               | 30 +++++++++++++++++++
 2 files changed, 36 insertions(+), 6 deletions(-)

diff --git a/packages/core/src/sync/providers/anthropic.ts b/packages/core/src/sync/providers/anthropic.ts
index fda7baf8de6..af568337ce0 100644
--- a/packages/core/src/sync/providers/anthropic.ts
+++ b/packages/core/src/sync/providers/anthropic.ts
@@ -210,15 +210,15 @@ export function parseAnthropicPricing(markdown: string, now = new Date()) {
 
   const table = section.split("\n").filter((line) => line.trimStart().startsWith("|"));
   const rows = table.map((line) => line.split("|").slice(1, -1).map((cell) => cell.trim()));
-  const header = rows[0]?.map(markdownText);
+  const header = rows[0]?.map((cell) => markdownText(cell).toLowerCase().replaceAll("&", "and"));
   if (header === undefined) throw new Error("Anthropic pricing page is missing the model pricing table");
 
   const indexes = {
-    model: header.indexOf("Model"),
-    input: header.indexOf("Base Input Tokens"),
-    cacheWrite: header.indexOf("5m Cache Writes"),
-    cacheRead: header.indexOf("Cache Hits & Refreshes"),
-    output: header.indexOf("Output Tokens"),
+    model: header.indexOf("model"),
+    input: header.indexOf("base input tokens"),
+    cacheWrite: header.indexOf("5m cache writes"),
+    cacheRead: header.indexOf("cache hits and refreshes"),
+    output: header.indexOf("output tokens"),
   };
   if (Object.values(indexes).some((index) => index < 0)) {
     throw new Error("Anthropic model pricing table has unexpected columns");
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index d04d7e75817..d92d231a0b6 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -923,6 +923,36 @@ test("parses current and future Anthropic pricing rows", () => {
   expect(standard.get("claude sonnet 5")).toMatchObject({ input: 3, output: 15 });
 });
 
+test.each([
+  "| Model | Base input tokens | 5m cache writes | 1h cache writes | Cache hits and refreshes | Output tokens |",
+  "| Model | Base input tokens | 5m cache writes | 1h cache writes | Cache hits & refreshes | Output tokens |",
+  "| Model | Base Input Tokens | 5m Cache Writes | 1h Cache Writes | Cache Hits and Refreshes | Output Tokens |",
+])("parses Anthropic pricing with header %s", (header) => {
+  const markdown = anthropicPricingMarkdown.replace(/^\| Model \|.*$/m, header);
+  const pricing = parseAnthropicPricing(markdown, new Date("2026-09-03T00:00:00Z"));
+
+  expect(pricing.size).toBe(5);
+  expect(pricing.get("claude opus 4.8")).toEqual({
+    input: 5,
+    output: 25,
+    cacheRead: 0.5,
+    cacheWrite: 6.25,
+    deprecated: false,
+  });
+});
+
+test.each([
+  "Model",
+  "Base Input Tokens",
+  "5m Cache Writes",
+  "Cache Hits & Refreshes",
+  "Output Tokens",
+])("rejects Anthropic pricing without the %s column", (column) => {
+  const markdown = anthropicPricingMarkdown.replace(`| ${column} |`, "| Unknown |");
+
+  expect(() => parseAnthropicPricing(markdown)).toThrow("Anthropic model pricing table has unexpected columns");
+});
+
 test("syncs Anthropic capabilities and exact effort levels", () => {
   const model = buildAnthropicModel(anthropicModel(), {
     name: "Claude Sonnet 5",

From cb52125dbcba14b9cb285d2ff5a8016363d5eabc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:44:00 -0500
Subject: [PATCH 0563/2012] fix(ci): watch model sync failures in CI fixer
 (#6197)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .github/workflows/ci-fixer.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/ci-fixer.yml b/.github/workflows/ci-fixer.yml
index 3f0112d1650..9d1205a126c 100644
--- a/.github/workflows/ci-fixer.yml
+++ b/.github/workflows/ci-fixer.yml
@@ -2,7 +2,7 @@ name: Dev CI Fixer
 
 on:
   workflow_run:
-    workflows: [Deploy]
+    workflows: [Deploy, Sync Model Catalogs]
     types: [completed]
   workflow_dispatch:
 

From 3b15628809c8f2f3ffe7d39bf0c6c23e7c70eb15 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 10:44:13 -0500
Subject: [PATCH 0564/2012] fix(sync): let last LLM Gateway case variant win
 (#6196)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/llmgateway.ts     |  4 +-
 packages/core/test/sync.test.ts               | 50 +++++++++++++++++++
 2 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/packages/core/src/sync/providers/llmgateway.ts b/packages/core/src/sync/providers/llmgateway.ts
index a6ea0b7ca57..3a379eac472 100644
--- a/packages/core/src/sync/providers/llmgateway.ts
+++ b/packages/core/src/sync/providers/llmgateway.ts
@@ -113,7 +113,9 @@ export const llmgateway = {
     if (data.length === 0) {
       throw new Error("LLM Gateway returned no text models");
     }
-    return data;
+    // Case-insensitive ID conflicts use the last entry, including its original
+    // casing and complete record; never mix metadata from different routes.
+    return [...new Map(data.map((model) => [model.id.toLowerCase(), model])).values()];
   },
   translateModel(model, context) {
     const translated = buildLLMGatewayModel(model, context.existing(model.id));
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index d92d231a0b6..c190f650eac 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -3636,6 +3636,56 @@ test("strips image input when the deployment has no vision", () => {
   });
 });
 
+test("keeps the last LLM Gateway entry for case-insensitive duplicate IDs", () => {
+  const first = llmGatewayModel({ id: "qwen3.8-27b", family: "alibaba" });
+  const other = llmGatewayModel();
+  for (const id of [first.id, "Qwen3.8-27B"]) {
+    const last = llmGatewayModel({
+      id,
+      family: "consensusprotocol",
+      context_length: 32_768,
+      pricing: { prompt: "0.41e-6", completion: "2.5e-6" },
+    });
+    expect(llmgateway.parseModels({ data: [first, other, last] })).toEqual([last, other]);
+    expect(llmgateway.parseModels({ data: [last, other, first] })).toEqual([first, other]);
+  }
+  const nonText = llmGatewayModel({
+    id: first.id,
+    architecture: { input_modalities: ["text"], output_modalities: ["image"] },
+  });
+  expect(llmgateway.parseModels({ data: [first, nonText] })).toEqual([first]);
+});
+
+test("syncs the last LLM Gateway case variant without mixing source records", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "models-dev-llmgateway-case-"));
+  const modelsDir = path.join(root, "providers", "llmgateway", "models");
+  await mkdir(modelsDir, { recursive: true });
+  const first = llmGatewayModel({ id: "qwen3.8-27b", family: undefined });
+  const last = llmGatewayModel({
+    id: "Qwen3.8-27B",
+    family: undefined,
+    context_length: 32_768,
+    pricing: { prompt: "0.41e-6", completion: "2.5e-6" },
+  });
+  const provider = { ...llmgateway, modelsDir, fetchModels: async () => ({ data: [first, last] }) };
+
+  try {
+    await syncProvider({ ...provider, fetchModels: async () => ({ data: [first] }) });
+    const result = await syncProvider(provider);
+    expect(result).toMatchObject({ created: 1, updated: 0, deleted: 1 });
+    expect(await Bun.file(path.join(modelsDir, `${first.id}.toml`)).exists()).toBe(false);
+    const written = Bun.TOML.parse(await readFile(path.join(modelsDir, `${last.id}.toml`), "utf8"));
+    expect(written).toMatchObject({
+      cost: { input: 0.41, output: 2.5 },
+      limit: { context: 32_768 },
+    });
+    expect(written.cost).not.toHaveProperty("cache_write");
+    expect(await syncProvider(provider)).toMatchObject({ created: 0, updated: 0, deleted: 0, unchanged: 1 });
+  } finally {
+    await rm(root, { recursive: true, force: true });
+  }
+});
+
 test("refuses empty responses in both LLM Gateway syncs", () => {
   expect(() => llmgateway.parseModels({ data: [] })).toThrow("no text models");
   expect(() => llmgatewayProviders.parseModels({ data: [] })).toThrow("mapped view unavailable");

From 7541e1bbc64172905db09971c13db90448511308 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:28:44 +0000
Subject: [PATCH 0565/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6203)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/Qwen3.8-27B.toml  | 27 +++++++++++++++++++
 .../llmgateway/models/muse-spark-1.1.toml     |  8 +++---
 providers/llmgateway/models/qwen3.8-27b.toml  | 10 -------
 3 files changed, 31 insertions(+), 14 deletions(-)
 create mode 100644 providers/llmgateway/models/Qwen3.8-27B.toml
 delete mode 100644 providers/llmgateway/models/qwen3.8-27b.toml

diff --git a/providers/llmgateway/models/Qwen3.8-27B.toml b/providers/llmgateway/models/Qwen3.8-27B.toml
new file mode 100644
index 00000000000..8e2a4937fda
--- /dev/null
+++ b/providers/llmgateway/models/Qwen3.8-27B.toml
@@ -0,0 +1,27 @@
+name = "Qwen3.8 27B"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.41
+output = 2.5
+cache_read = 0.08
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/llmgateway/models/muse-spark-1.1.toml b/providers/llmgateway/models/muse-spark-1.1.toml
index bbc612167da..f71eeb745a8 100644
--- a/providers/llmgateway/models/muse-spark-1.1.toml
+++ b/providers/llmgateway/models/muse-spark-1.1.toml
@@ -1,10 +1,10 @@
 base_model = "meta/muse-spark-1.1"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25
 output = 4.25
 cache_read = 0.15
-
-[limit]
-context = 1_048_576
diff --git a/providers/llmgateway/models/qwen3.8-27b.toml b/providers/llmgateway/models/qwen3.8-27b.toml
deleted file mode 100644
index 79646fa55fc..00000000000
--- a/providers/llmgateway/models/qwen3.8-27b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "alibaba/qwen3.8-27b"
-reasoning_options = []
-
-[cost]
-input = 0.42
-output = 3
-cache_read = 0.085
-
-[limit]
-context = 1_000_000

From f80bbfbe8d8bcc3e2c9383c11ec9883844104cf7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:28:53 +0000
Subject: [PATCH 0566/2012] chore(sync): update Pioneer model catalog (#6209)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/pioneer/models/meta/muse-spark-1.1.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/pioneer/models/meta/muse-spark-1.1.toml b/providers/pioneer/models/meta/muse-spark-1.1.toml
index e91da16f440..dd7a5d7a776 100644
--- a/providers/pioneer/models/meta/muse-spark-1.1.toml
+++ b/providers/pioneer/models/meta/muse-spark-1.1.toml
@@ -12,7 +12,3 @@ input = 1.25
 output = 4.25
 cache_read = 0.15
 cache_write = 1.25
-
-[limit]
-context = 1_048_576
-output = 131_072

From e399c286646b529077e7343d8db3cda05b4ee5dd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:06 +0000
Subject: [PATCH 0567/2012] chore(sync): update NanoGPT model catalog (#6201)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/meta/muse-spark-1.1.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/nano-gpt/models/meta/muse-spark-1.1.toml b/providers/nano-gpt/models/meta/muse-spark-1.1.toml
index 628d4ce3da9..b5a1847c173 100644
--- a/providers/nano-gpt/models/meta/muse-spark-1.1.toml
+++ b/providers/nano-gpt/models/meta/muse-spark-1.1.toml
@@ -12,6 +12,7 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
+context = 1_000_000
 input = 1_000_000
 output = 65_536
 

From dc15f51b25ecf1f0a3dba57b36d3b4d916683cb9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:10 +0000
Subject: [PATCH 0568/2012] chore(sync): update Kilo model catalog (#6207)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta/muse-spark-1.1.toml | 1 -
 providers/kilo/models/tencent/hy3.toml         | 6 +++---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/meta/muse-spark-1.1.toml b/providers/kilo/models/meta/muse-spark-1.1.toml
index 52d43f95e1f..68a20ab4888 100644
--- a/providers/kilo/models/meta/muse-spark-1.1.toml
+++ b/providers/kilo/models/meta/muse-spark-1.1.toml
@@ -11,7 +11,6 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 943_718
 
 [modalities]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 071f184222d99cc04afba1fd774f91d45c120727 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:21 +0000
Subject: [PATCH 0569/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6206)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/meta/muse-spark-1.1.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/vercel/models/meta/muse-spark-1.1.toml b/providers/vercel/models/meta/muse-spark-1.1.toml
index e0e2bdd14ce..6a2711cb198 100644
--- a/providers/vercel/models/meta/muse-spark-1.1.toml
+++ b/providers/vercel/models/meta/muse-spark-1.1.toml
@@ -12,7 +12,6 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 1_048_576
 
 [modalities]

From 031b56ce85d278010f0352c82c1178babe9cc18f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:24 +0000
Subject: [PATCH 0570/2012] chore(sync): update EmpirioLabs AI model catalog
 (#6204)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/muse-spark-1-1.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/empiriolabs/models/muse-spark-1-1.toml b/providers/empiriolabs/models/muse-spark-1-1.toml
index c6e27c90a7a..8db80eb737d 100644
--- a/providers/empiriolabs/models/muse-spark-1-1.toml
+++ b/providers/empiriolabs/models/muse-spark-1-1.toml
@@ -9,9 +9,5 @@ input = 1.25
 output = 4.25
 cache_read = 1
 
-[limit]
-context = 1_048_576
-output = 131_072
-
 [modalities]
 input = ["text", "image", "video", "audio"]

From e3e048927a2d37771b21152aee0d453b53765ae6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:27 +0000
Subject: [PATCH 0571/2012] chore(sync): update LLM Gateway model catalog
 (#6202)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/meta/muse-spark-1.1.toml      | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/llmgateway-providers/models/meta/muse-spark-1.1.toml b/providers/llmgateway-providers/models/meta/muse-spark-1.1.toml
index 760bbb92383..a8448c5cc19 100644
--- a/providers/llmgateway-providers/models/meta/muse-spark-1.1.toml
+++ b/providers/llmgateway-providers/models/meta/muse-spark-1.1.toml
@@ -9,7 +9,3 @@ values = ["minimal", "low", "medium", "high", "xhigh"]
 input = 1.25
 output = 4.25
 cache_read = 0.15
-
-[limit]
-context = 1_048_576
-output = 131_072

From eb9abca40277409180cc05e39f45461aeb547cb0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:33 +0000
Subject: [PATCH 0572/2012] chore(sync): update OpenRouter model catalog
 (#6208)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/meta/muse-spark-1.1.toml        | 1 -
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 3 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta/muse-spark-1.1.toml b/providers/openrouter/models/meta/muse-spark-1.1.toml
index be6bd0b377c..2e671bfddf0 100644
--- a/providers/openrouter/models/meta/muse-spark-1.1.toml
+++ b/providers/openrouter/models/meta/muse-spark-1.1.toml
@@ -11,7 +11,6 @@ output = 4.25
 cache_read = 0.15
 
 [limit]
-context = 1_048_576
 output = 943_718
 
 [modalities]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From c776288f419f9e697319eb9707e63872a6add93d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 16:29:40 +0000
Subject: [PATCH 0573/2012] chore(sync): update Tinfoil model catalog (#6205)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/glm-5-2.toml | 14 --------------
 1 file changed, 14 deletions(-)
 delete mode 100644 providers/tinfoil/models/glm-5-2.toml

diff --git a/providers/tinfoil/models/glm-5-2.toml b/providers/tinfoil/models/glm-5-2.toml
deleted file mode 100644
index e8f9e6bdc34..00000000000
--- a/providers/tinfoil/models/glm-5-2.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-# Pricing source: https://inference.tinfoil.sh/v1/models
-base_model = "zhipuai/glm-5.2"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 1.5
-output = 5.25
-cache_read = 0.375
-
-[limit]
-context = 393_216

From 2a90db72212cd5af26ccf037a637b6d4217aeb43 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 11:44:56 -0500
Subject: [PATCH 0574/2012] fix(sync): report missing Cloudflare reasoning
 options (#6198)

* fix(sync): send incomplete models to the issue fixer

* fix(sync): limit issue recovery to missing reasoning options

* refactor(sync): minimize the reasoning-options issue fix

---------

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../script/generate-cloudflare-ai-gateway.ts  |  4 ++
 packages/core/src/sync/index.ts               | 44 +++++++++++++------
 packages/core/src/sync/missing-issues.ts      | 16 +++++--
 .../src/sync/missing-reasoning-options.ts     |  7 +++
 .../sync/providers/cloudflare-ai-gateway.ts   | 10 +++--
 .../core/test/cloudflare-ai-gateway.test.ts   | 35 ++++++++++++++-
 6 files changed, 95 insertions(+), 21 deletions(-)
 create mode 100644 packages/core/src/sync/missing-reasoning-options.ts

diff --git a/packages/core/script/generate-cloudflare-ai-gateway.ts b/packages/core/script/generate-cloudflare-ai-gateway.ts
index b74d62a8c8f..447e05a4d26 100644
--- a/packages/core/script/generate-cloudflare-ai-gateway.ts
+++ b/packages/core/script/generate-cloudflare-ai-gateway.ts
@@ -6,6 +6,10 @@ const check = process.argv.includes("--check");
 const result = await syncProviderByID("cloudflare-ai-gateway", { dryRun: check });
 
 if (check) {
+  if (result.notices.length > 0) {
+    console.error(`--check: ${result.notices.join("\n")}`);
+    process.exit(1);
+  }
   if (result.files.length > 0) {
     console.error(`--check: ${result.files.length} file(s) out of date`);
     process.exit(1);
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 2e151aa702b..7f88aacb5bb 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -5,6 +5,7 @@ import { z } from "zod";
 
 import { AuthoredModel, AuthoredModelShape, ModelMetadata } from "../schema.js";
 import { openMissingModelIssues } from "./missing-issues.js";
+import { MissingReasoningOptionsError } from "./missing-reasoning-options.js";
 import { ambient } from "./providers/ambient.js";
 import { anthropic } from "./providers/anthropic.js";
 import { baseten } from "./providers/baseten.js";
@@ -251,16 +252,25 @@ export async function syncProvider<SourceModel>(
   const caseNormalizedDesiredPaths = new Map<string, string>();
   const desiredMetadata = new Map<string, { model: z.infer<typeof ModelMetadata>; content: string }>();
   const skippedRemote: string[] = [];
+  const missingReasoning = new Map<string, string>();
 
   for (const sourceModel of sourceModels) {
-    const translated = provider.translateModel(sourceModel, {
-      existing(id) {
-        return existing.get(`${id}.toml`)?.toml;
-      },
-      authored(id) {
-        return existing.get(`${id}.toml`)?.authored;
-      },
-    });
+    let translated: ReturnType<typeof provider.translateModel>;
+    try {
+      translated = provider.translateModel(sourceModel, {
+        existing(id) {
+          return existing.get(`${id}.toml`)?.toml;
+        },
+        authored(id) {
+          return existing.get(`${id}.toml`)?.authored;
+        },
+      });
+    } catch (error) {
+      if (!(error instanceof MissingReasoningOptionsError)) throw error;
+      missingReasoning.set(error.modelId, error.message);
+      console.warn(error.message);
+      continue;
+    }
     if (translated === undefined) {
       const skippedID = provider.sourceID?.(sourceModel);
       if (skippedID !== undefined) skippedRemote.push(skippedID);
@@ -443,6 +453,10 @@ export async function syncProvider<SourceModel>(
   const missingLocal: string[] = [];
   for (const relativePath of new Set([...existing.keys(), ...brokenSymlinks])) {
     if (desired.has(relativePath)) continue;
+    if (missingReasoning.has(relativePath.slice(0, -5))) {
+      unchanged++;
+      continue;
+    }
     if (provider.deleteMissing === false) {
       missingLocal.push(relativePath);
       console.log(`Retaining model missing from source: ${relativePath}`);
@@ -465,22 +479,26 @@ export async function syncProvider<SourceModel>(
   }
 
   const notices = [
+    ...missingReasoning.values(),
     ...provider.skippedNotice?.(skippedRemote) ?? [],
     ...provider.missingNotice?.(missingLocal) ?? [],
   ];
 
+  const issueModels = [
+    ...(provider.skipCreates === true ? skippedRemote : []),
+    ...missingReasoning.keys(),
+  ];
   if (
-    provider.skipCreates === true
-    && provider.trackMissingModels !== false
-    && skippedRemote.length > 0
+    provider.trackMissingModels !== false
+    && issueModels.length > 0
     && options.openIssues === true
   ) {
     try {
       notices.push(
         ...await openMissingModelIssues(
           { id: provider.id, name: provider.name, modelsDir: provider.modelsDir },
-          skippedRemote,
-          { dryRun: options.dryRun },
+          issueModels,
+          { dryRun: options.dryRun, reasons: Object.fromEntries(missingReasoning) },
         ),
       );
     } catch (error) {
diff --git a/packages/core/src/sync/missing-issues.ts b/packages/core/src/sync/missing-issues.ts
index fe008b074b9..81a01a3742e 100644
--- a/packages/core/src/sync/missing-issues.ts
+++ b/packages/core/src/sync/missing-issues.ts
@@ -6,15 +6,18 @@ export interface MissingModelIssueTarget {
 
 export interface OpenMissingModelIssuesOptions {
   dryRun?: boolean;
+  reasons?: Record<string, string>;
 }
 
 function issueTitle(providerId: string, modelId: string) {
   return `[missing-model] ${providerId}: ${modelId}`;
 }
 
-function issueBody(provider: MissingModelIssueTarget, modelId: string) {
+function issueBody(provider: MissingModelIssueTarget, modelId: string, reason?: string) {
   return [
-    `The **${provider.name}** catalog sync found remote model \`${modelId}\` that is not in the local catalog.`,
+    reason === undefined
+      ? `The **${provider.name}** catalog sync found remote model \`${modelId}\` that is not in the local catalog.`
+      : `The **${provider.name}** catalog sync is missing reasoning options for remote model \`${modelId}\`. Any existing local entry was left unchanged.`,
     "",
     `| Field | Value |`,
     `| --- | --- |`,
@@ -22,8 +25,13 @@ function issueBody(provider: MissingModelIssueTarget, modelId: string) {
     `| Model ID | \`${modelId}\` |`,
     `| Expected path | \`${provider.modelsDir}/${modelId}.toml\` |`,
     "",
-    "This provider uses `skipCreates` because the remote source is not enough to auto-author a full TOML.",
+    reason === undefined
+      ? "This provider uses `skipCreates` because the remote source is not enough to auto-author a full TOML."
+      : `Sync diagnostic: ${reason}`,
     "Add the model manually (prefer `base_model` when matching `models/` metadata exists).",
+    ...(reason === undefined ? [] : [
+      `Research the provider's reasoning controls; do not use an empty placeholder. Update \`providers/${provider.id}/curation.toml\` if present, including source URLs and wire paths in its \`note\` array, so the next sync retains the fix.`,
+    ]),
     "",
   ].join("\n");
 }
@@ -80,7 +88,7 @@ export async function openMissingModelIssues(
     }
 
     try {
-      const number = await createIssue(title, issueBody(provider, modelId), labels);
+      const number = await createIssue(title, issueBody(provider, modelId, options.reasons?.[modelId]), labels);
       existingByTitle.set(title, number);
       await dispatchIssueFixer(provider.id, number);
       const notice = `Opened GitHub issue #${number} and dispatched the issue fixer for missing model \`${modelId}\``;
diff --git a/packages/core/src/sync/missing-reasoning-options.ts b/packages/core/src/sync/missing-reasoning-options.ts
new file mode 100644
index 00000000000..756557ed86d
--- /dev/null
+++ b/packages/core/src/sync/missing-reasoning-options.ts
@@ -0,0 +1,7 @@
+/** A reasoning model needs researched provider-specific controls. */
+export class MissingReasoningOptionsError extends Error {
+  constructor(readonly modelId: string, reason: string) {
+    super(`${modelId}: ${reason}`);
+    this.name = "MissingReasoningOptionsError";
+  }
+}
diff --git a/packages/core/src/sync/providers/cloudflare-ai-gateway.ts b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
index 65c9499b7e3..775c8524f36 100644
--- a/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
+++ b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
@@ -4,6 +4,7 @@ import { z } from "zod";
 
 import { ReasoningOption } from "../../schema.js";
 import type { ExistingModel, SyncProvider, SyncedBaseModel } from "../index.js";
+import { MissingReasoningOptionsError } from "../missing-reasoning-options.js";
 
 const API_BASE = "https://api.cloudflare.com/client/v4/accounts";
 const PROVIDER_DIR = path.join(
@@ -148,6 +149,8 @@ export function buildCloudflareAiGatewayModel(
   existing?: ExistingModel,
 ): SyncedBaseModel {
   const id = catalog.model_id;
+  // Pricing failures must not be hidden by missing reasoning controls.
+  const cost = proxiedCost(catalog.pricing, id);
   const baseModel = curated.base_model ?? resolveBaseModel(id);
   if (baseModel === undefined) {
     throw new Error(`${id}: no lab file and no curated base_model; add it to skip or map it`);
@@ -163,14 +166,15 @@ export function buildCloudflareAiGatewayModel(
     const derived = deriveReasoningOptions(schemaInput);
     const reasoningOptions = curated.reasoning_options ?? (derived.length > 0 ? derived : undefined);
     if (reasoningOptions === undefined) {
-      throw new Error(
-        `${id}: base ${baseModel} reasons but the catalog schema and curation provide no reasoning_options`,
+      throw new MissingReasoningOptionsError(
+        id,
+        `base ${baseModel} reasons but the catalog schema and curation provide no reasoning_options`,
       );
     }
     model.reasoning_options = reasoningOptions;
   }
 
-  model.cost = proxiedCost(catalog.pricing, id);
+  model.cost = cost;
 
   const limit = {
     ...(catalog.context_length == null && existing?.limit?.context === undefined
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
index 3cf59f404c3..25db6a6e0fd 100644
--- a/packages/core/test/cloudflare-ai-gateway.test.ts
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -1,14 +1,47 @@
-import { expect, test } from "bun:test";
+import { expect, spyOn, test } from "bun:test";
 import { mkdir, mkdtemp, readFile, rm, symlink, writeFile } from "node:fs/promises";
 import path from "node:path";
 
 import { syncProvider } from "../src/sync/index.js";
+import * as missingIssues from "../src/sync/missing-issues.js";
 import {
   buildCloudflareAiGatewayModel,
   cloudflareAiGateway,
   deriveReasoningOptions,
 } from "../src/sync/providers/cloudflare-ai-gateway.js";
 
+test("missing reasoning controls open issues without deleting existing models or blocking valid ones", async () => {
+  const dir = await mkdtemp(path.join(import.meta.dirname, "../../../providers/.reasoning-sync-"));
+  const modelsDir = path.join(dir, "models");
+  const ids = ["anthropic/claude-fable-5.1", "anthropic/claude-fable-5-1"];
+  const file = path.join(modelsDir, `${ids[0]}.toml`);
+  const content = '# Keep authored controls\nbase_model = "anthropic/claude-fable-5-1"\nreasoning_options = [{ type = "effort", values = ["high"] }]\n';
+  await mkdir(path.dirname(file), { recursive: true });
+  await writeFile(file, content);
+  const issues = spyOn(missingIssues, "openMissingModelIssues").mockResolvedValue([]);
+  const provider = {
+    ...cloudflareAiGateway, modelsDir,
+    async fetchModels() {
+      return [...ids, "openai/gpt-4.1"].map((model_id) => ({
+        catalog: { model_id, task: "Text Generation", pricing: { "Input tokens (per 1M)": 1, "Output tokens (per 1M)": 2 } },
+      }));
+    },
+  };
+  try {
+    const result = await syncProvider(provider, { openIssues: true });
+    expect(result).toMatchObject({ created: 1, updated: 0, deleted: 0, unchanged: 1 });
+    expect(await readFile(file, "utf8")).toBe(content);
+    expect(await Bun.file(path.join(modelsDir, `${ids[1]}.toml`)).exists()).toBe(false);
+    expect(issues.mock.calls[0]?.[1]).toEqual(ids);
+    expect(issues.mock.calls[0]?.[2]?.reasons?.[ids[0]!]).toContain("reasoning_options");
+    await expect(syncProvider({ ...provider, async fetchModels() { throw new Error("fetch failed"); } })).rejects.toThrow("fetch failed");
+    expect(issues).toHaveBeenCalledTimes(1);
+  } finally {
+    issues.mockRestore();
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
 test("builds Cloudflare AI Gateway overrides from catalog metadata", () => {
   const model = buildCloudflareAiGatewayModel(
     {

From e628c5a472de83dcae5bc689eff877a865416451 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:24:56 +0000
Subject: [PATCH 0575/2012] chore(sync): update Kilo model catalog (#6214)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inclusionai/ling-3.0-flash-fin.toml       | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml
new file mode 100644
index 00000000000..0482b60a0bb
--- /dev/null
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml
@@ -0,0 +1,28 @@
+name = "Ling 3.0 Flash Fin"
+description = "Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for real-world investment..."
+family = "ling"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.06
+output = 0.18
+cache_read = 0.012
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text"]
+output = ["text"]

From b7695b2c04cf4610a6a162050e2f8e17083753de Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:25:11 +0000
Subject: [PATCH 0576/2012] chore(sync): update Charm Hyper model catalog
 (#6212)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml | 6 +++---
 providers/hyper/models/glm-5.toml              | 6 +++---
 providers/hyper/models/minimax-m2.7.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 7c758447185..c98d649e236 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.116
-output = 0.38
-cache_write = 0.058
+input = 0.106
+output = 0.368
+cache_write = 0.053
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index aaa2e1d6169..ac7c97dd31d 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.85
-output = 2.774
-cache_write = 0.425
+input = 0.93
+output = 2.878
+cache_write = 0.465
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 350561786a4..e1e2973ecc9 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.426
-output = 1.62
-cache_write = 0.213
+input = 0.418
+output = 1.588
+cache_write = 0.209
 
 [limit]
 context = 262_100

From d79171f8e9e45a6fe21d1b405672dc30b1d7e011 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:25:30 +0000
Subject: [PATCH 0577/2012] chore(sync): update OpenRouter model catalog
 (#6211)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../inclusionai/ling-3.0-flash-fin.toml       | 29 +++++++++++++++++++
 .../openrouter/models/qwen/qwen3.8-27b.toml   |  5 ++--
 4 files changed, 37 insertions(+), 9 deletions(-)
 create mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-fin.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 97904bfc0b8..b239680e290 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.035996
-output = 2.071992
-cache_read = 0.086333
+input = 1.034256
+output = 2.068512
+cache_read = 0.086188
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-fin.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-fin.toml
new file mode 100644
index 00000000000..85c929a5d94
--- /dev/null
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-fin.toml
@@ -0,0 +1,29 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ling 3.0 Flash Fin"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "ling"
+release_date = "2026-08-27"
+last_updated = "2026-08-27"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.06
+output = 0.18
+cache_read = 0.012
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index d6e1375e402..d1ddcffa2ed 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,10 +11,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.425
-output = 2.55
+input = 0.42
+output = 3
 cache_read = 0.085
-cache_write = 0.53125
 
 [limit]
 context = 1_000_000

From 0697196befc273c9a83fb88034304bbb6987d6cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:25:34 +0000
Subject: [PATCH 0578/2012] chore(sync): update Merge Gateway model catalog
 (#6213)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 0a7a2254ee8..b4bdd294444 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 3daf906b1190e761f44c65248e0505e9b248498b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 17:25:40 +0000
Subject: [PATCH 0579/2012] chore(sync): update NanoGPT model catalog (#6215)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/anthropic/claude-fable-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/nano-gpt/models/anthropic/claude-fable-latest.toml b/providers/nano-gpt/models/anthropic/claude-fable-latest.toml
index b0d0cc6cd6d..48f22c01e7f 100644
--- a/providers/nano-gpt/models/anthropic/claude-fable-latest.toml
+++ b/providers/nano-gpt/models/anthropic/claude-fable-latest.toml
@@ -16,7 +16,7 @@ values = ["low", "medium", "high", "xhigh", "max"]
 [cost]
 input = 10
 output = 50
-cache_read = 1
+cache_read = 0.25
 cache_write = 12.5
 
 [limit]

From 78d2606cd0f5daff6405d00979d87c86a83c6137 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 18:32:04 +0000
Subject: [PATCH 0580/2012] chore(sync): update LLM Gateway model catalog
 (#6220)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../muse-spark-1.3-contributor.toml           | 28 +++++++++++++++++++
 .../models/meta/muse-spark-1.3.toml           | 14 ++++++++++
 2 files changed, 42 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/meta-contributor/muse-spark-1.3-contributor.toml
 create mode 100644 providers/llmgateway-providers/models/meta/muse-spark-1.3.toml

diff --git a/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.3-contributor.toml b/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..63670f90238
--- /dev/null
+++ b/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.3-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3 Contributor (Meta Contributor)"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
new file mode 100644
index 00000000000..37217f9c0c2
--- /dev/null
+++ b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
@@ -0,0 +1,14 @@
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 (Meta)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[limit]
+output = 1_048_576

From 060d7e992a58f64c2033a3593618193cf63ed84d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 18:32:23 +0000
Subject: [PATCH 0581/2012] chore(sync): update Kilo model catalog (#6221)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index de31079e2dd..6481e4e1993 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.5
+output = 14
+cache_read = 0.29
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7e40b3c4374..62973e4f574 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.134
-output = 3.564
-cache_read = 0.1863
+input = 1.148
+output = 3.608
+cache_read = 0.1886
 
 [limit]
 context = 1_048_576

From b8bdf575119485c2d9bed144f1b9e5d59682e8f3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 18:32:28 +0000
Subject: [PATCH 0582/2012] chore(sync): update Merge Gateway model catalog
 (#6222)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 16d15cbb411..c680bd0f847 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +13,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index faa70b7c1e0..3ef83d21d9d 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -12,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From e11de97c90f9e56bce05a398bd8a2a2fff0f3334 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 18:32:31 +0000
Subject: [PATCH 0583/2012] chore(sync): update OpenRouter model catalog
 (#6223)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 8 +++++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 6 +++---
 3 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 7912c0ac3a3..052fb05abd3 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 name = "MoonshotAI Kimi Latest"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 family = "kimi"
@@ -18,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.5
+output = 14
+cache_read = 0.29
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 4bdfd2001cf..639aea7c111 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.134
-output = 3.564
-cache_read = 0.1863
+input = 1.148
+output = 3.608
+cache_read = 0.1886
 
 [limit]
 context = 1_310_720

From 7643d3456b944e01bb28040720cc3735c2a3f98e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 18:32:37 +0000
Subject: [PATCH 0584/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6219)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/muse-spark-1.3-contributor.toml    | 28 +++++++++++++++++++
 .../llmgateway/models/muse-spark-1.3.toml     | 10 +++++++
 2 files changed, 38 insertions(+)
 create mode 100644 providers/llmgateway/models/muse-spark-1.3-contributor.toml
 create mode 100644 providers/llmgateway/models/muse-spark-1.3.toml

diff --git a/providers/llmgateway/models/muse-spark-1.3-contributor.toml b/providers/llmgateway/models/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..e27694885d8
--- /dev/null
+++ b/providers/llmgateway/models/muse-spark-1.3-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.3 Contributor"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "muse"
+release_date = "2026-09-02"
+last_updated = "2026-09-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/llmgateway/models/muse-spark-1.3.toml b/providers/llmgateway/models/muse-spark-1.3.toml
new file mode 100644
index 00000000000..cba766d6589
--- /dev/null
+++ b/providers/llmgateway/models/muse-spark-1.3.toml
@@ -0,0 +1,10 @@
+base_model = "meta/muse-spark-1.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15

From cc4a3e5d48081f9fc4419799d29e40a47350e738 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 19:21:35 +0000
Subject: [PATCH 0585/2012] chore(sync): update Kilo model catalog (#6224)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat-v3.1.toml | 4 ++--
 providers/kilo/models/~z-ai/glm-flash-latest.toml      | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
index 4269d8ff1fa..7717353dce8 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
@@ -20,8 +20,8 @@ output = 1
 cache_read = 0.135
 
 [limit]
-context = 163_840
-output = 32_768
+context = 161_000
+output = 144_900
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..7b848afc9df 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From a52a1b7a6af20078b42715f7f49de135753853f4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 19:22:06 +0000
Subject: [PATCH 0586/2012] chore(sync): update OpenRouter model catalog
 (#6225)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-chat-v3.1.toml    | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
index 24302ed5e58..d60c40e2af2 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
@@ -17,13 +17,13 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.25
-output = 0.95
-cache_read = 0.13
+input = 0.55
+output = 1.65
+cache_read = 0.55
 
 [limit]
 context = 163_840
-output = 32_768
+output = 144_900
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b239680e290..07235ccdb08 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.034256
-output = 2.068512
-cache_read = 0.086188
+input = 1.030776
+output = 2.061552
+cache_read = 0.085898
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..8119fb87874 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 38ac4c4d67e19bbf297996eb80f4f51af72413d9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 21:24:07 +0000
Subject: [PATCH 0587/2012] chore(sync): update NanoGPT model catalog (#6227)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../MS3.2-24B-Magnum-Diamond.toml             |  4 +--
 .../EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0.toml  |  4 +--
 .../EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1.toml  |  4 +--
 .../models/Gryphe/MythoMax-L2-13b.toml        |  6 ++---
 .../Wayfarer-Large-70B-Llama-3.3.toml         |  4 +--
 .../models/NousResearch/hermes-4-70b.toml     | 25 -----------------
 ...2-The-Omega-Directive-24B-Unslop-v2.0.toml |  4 +--
 .../models/Sao10K/L3.1-70B-Hanami-x1.toml     |  4 +--
 .../models/Steelskull/L3.3-Cu-Mai-R1-70b.toml |  4 +--
 .../Steelskull/L3.3-Electra-R1-70b.toml       |  4 +--
 .../Steelskull/L3.3-MS-Nevoria-70b.toml       |  4 +--
 .../Steelskull/L3.3-Nevoria-R1-70b.toml       |  4 +--
 .../models/TEE/deepseek-v4-flash.toml         |  2 +-
 providers/nano-gpt/models/TEE/kimi-k3.toml    |  2 +-
 .../models/TheDrummer/Cydonia-24B-v2.toml     |  4 +--
 .../models/TheDrummer/Cydonia-24B-v4.1.toml   |  2 +-
 .../models/TheDrummer/Cydonia-24B-v4.toml     |  4 +--
 .../TheDrummer/UnslopNemo-12B-v4.1.toml       |  6 ++---
 .../models/TheDrummer/skyfall-36b-v2.toml     |  6 ++---
 .../models/baseten/Kimi-K2-Instruct-FP4.toml  |  4 +--
 .../models/bytedance-seed/seed-2-1-turbo.toml |  2 +-
 .../Mistral-Small-3.2-24B-Instruct-2506.toml  |  2 +-
 .../models/deepseek-ai/DeepSeek-R1-0528.toml  |  6 ++---
 .../dots-studio/dots-3-note-preview.toml      | 27 -------------------
 providers/nano-gpt/models/glm-z1-air.toml     | 24 -----------------
 providers/nano-gpt/models/holo3-35b-a3b.toml  |  2 +-
 .../models/holo3-35b-a3b:thinking.toml        |  2 +-
 .../Llama-3.3-70B-Instruct-abliterated.toml   |  4 +--
 .../models/ibm-granite/granite-4.1-8b.toml    |  2 +-
 .../models/meganova-ai/manta-pro-1.0.toml     |  4 +--
 .../models/meta/muse-glimmer-30b.toml         |  1 +
 .../models/mistral-small-31-24b-instruct.toml |  2 +-
 .../models/mistralai/mistral-large.toml       |  2 +-
 .../models/mistralai/mistral-saba.toml        |  6 ++---
 .../mixtral-8x22b-instruct-v0.1.toml          |  2 +-
 .../moonshotai/Kimi-K2-Instruct-0905.toml     |  6 ++---
 .../nano-gpt/models/moonshotai/kimi-k3.toml   |  2 +-
 .../models/moonshotai/kimi-latest.toml        |  2 +-
 .../nano-gpt/models/nex-agi/nex-n2-mini.toml  |  2 +-
 .../nano-gpt/models/nex-agi/nex-n2-pro.toml   |  2 +-
 .../L3.1-70B-Celeste-V0.1-BF16.toml           |  4 +--
 .../nvidia/nemotron-3-nano-30b-a3b.toml       |  4 +--
 .../pamanseau/OpenReasoning-Nemotron-32B.toml |  4 +--
 .../perplexity-academic-researcher.toml       |  6 ++---
 .../qwen/Qwen3-235B-A22B-Instruct-2507.toml   |  5 ++--
 .../qwen/Qwen3-235B-A22B-Thinking-2507.toml   |  6 ++---
 .../qwen/Qwen3-Next-80B-A3B-Instruct.toml     |  6 ++---
 .../qwen/Qwen3-VL-235B-A22B-Instruct.toml     |  4 +--
 .../models/qwen25-vl-72b-instruct.toml        |  6 ++---
 .../nano-gpt/models/sonar-deep-research.toml  |  5 ++--
 .../nano-gpt/models/sonar-reasoning-pro.toml  |  5 ++--
 providers/nano-gpt/models/sonar.toml          |  6 ++---
 .../models/soob3123/GrayLine-Qwen3-8B.toml    |  4 +--
 .../stepfun-ai/step-3.5-flash-2603.toml       |  5 ++++
 .../models/stepfun-ai/step-3.5-flash.toml     |  5 ++++
 providers/nano-gpt/models/tencent/hy3.toml    |  1 -
 .../models/unsloth/gemma-3-12b-it.toml        |  6 ++---
 .../nano-gpt/models/upstage/solar-pro-3.toml  |  6 ++---
 providers/nano-gpt/models/x-ai/grok-4.3.toml  |  2 +-
 providers/nano-gpt/models/x-ai/grok-4.5.toml  |  1 +
 providers/nano-gpt/models/x-ai/grok-4.6.toml  |  1 +
 .../nano-gpt/models/x-ai/grok-build-0.1.toml  |  1 +
 .../nano-gpt/models/x-ai/grok-latest.toml     |  2 +-
 .../nano-gpt/models/z-ai/GLM-4.6-turbo.toml   |  6 ++---
 .../models/z-ai/GLM-4.6-turbo:thinking.toml   |  6 ++---
 providers/nano-gpt/models/z-ai/glm-4.5v.toml  |  4 +--
 .../models/z-ai/glm-4.5v:thinking.toml        |  4 +--
 67 files changed, 125 insertions(+), 193 deletions(-)
 delete mode 100644 providers/nano-gpt/models/NousResearch/hermes-4-70b.toml
 delete mode 100644 providers/nano-gpt/models/dots-studio/dots-3-note-preview.toml
 delete mode 100644 providers/nano-gpt/models/glm-z1-air.toml

diff --git a/providers/nano-gpt/models/Doctor-Shotgun/MS3.2-24B-Magnum-Diamond.toml b/providers/nano-gpt/models/Doctor-Shotgun/MS3.2-24B-Magnum-Diamond.toml
index e6f30b6dafe..1b216b2d0aa 100644
--- a/providers/nano-gpt/models/Doctor-Shotgun/MS3.2-24B-Magnum-Diamond.toml
+++ b/providers/nano-gpt/models/Doctor-Shotgun/MS3.2-24B-Magnum-Diamond.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0.toml b/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0.toml
index 539fa619452..698aeb18841 100644
--- a/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0.toml
+++ b/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0.toml
@@ -15,8 +15,8 @@ output = 2.006
 cache_read = 1.003
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1.toml b/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1.toml
index a776453d756..3de002b8cf5 100644
--- a/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1.toml
+++ b/providers/nano-gpt/models/EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1.toml
@@ -15,8 +15,8 @@ output = 2.006
 cache_read = 1.003
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/Gryphe/MythoMax-L2-13b.toml b/providers/nano-gpt/models/Gryphe/MythoMax-L2-13b.toml
index 7ef3547c0ea..5345194c389 100644
--- a/providers/nano-gpt/models/Gryphe/MythoMax-L2-13b.toml
+++ b/providers/nano-gpt/models/Gryphe/MythoMax-L2-13b.toml
@@ -15,9 +15,9 @@ output = 0.1003
 cache_read = 0.05015
 
 [limit]
-context = 4_000
-input = 4_000
-output = 4_096
+context = 4_096
+input = 4_096
+output = 3_686
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/LatitudeGames/Wayfarer-Large-70B-Llama-3.3.toml b/providers/nano-gpt/models/LatitudeGames/Wayfarer-Large-70B-Llama-3.3.toml
index d1c7072ed87..7ae5f729bb8 100644
--- a/providers/nano-gpt/models/LatitudeGames/Wayfarer-Large-70B-Llama-3.3.toml
+++ b/providers/nano-gpt/models/LatitudeGames/Wayfarer-Large-70B-Llama-3.3.toml
@@ -15,8 +15,8 @@ output = 0.7
 cache_read = 0.35
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/NousResearch/hermes-4-70b.toml b/providers/nano-gpt/models/NousResearch/hermes-4-70b.toml
deleted file mode 100644
index 96dd6842d7f..00000000000
--- a/providers/nano-gpt/models/NousResearch/hermes-4-70b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Included in subscription
-name = "Hermes 4 Medium"
-description = "General-purpose chat model for instruction following, writing, and analysis"
-family = "nousresearch"
-release_date = "2025-07-03"
-last_updated = "2025-07-03"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.2006
-output = 0.3995
-cache_read = 0.1003
-
-[limit]
-context = 128_000
-input = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/ReadyArt/MS3.2-The-Omega-Directive-24B-Unslop-v2.0.toml b/providers/nano-gpt/models/ReadyArt/MS3.2-The-Omega-Directive-24B-Unslop-v2.0.toml
index b465cf58bb5..b492b8276d3 100644
--- a/providers/nano-gpt/models/ReadyArt/MS3.2-The-Omega-Directive-24B-Unslop-v2.0.toml
+++ b/providers/nano-gpt/models/ReadyArt/MS3.2-The-Omega-Directive-24B-Unslop-v2.0.toml
@@ -15,8 +15,8 @@ output = 0.5
 cache_read = 0.25
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/Sao10K/L3.1-70B-Hanami-x1.toml b/providers/nano-gpt/models/Sao10K/L3.1-70B-Hanami-x1.toml
index ca81d94e690..6aebe0fe20b 100644
--- a/providers/nano-gpt/models/Sao10K/L3.1-70B-Hanami-x1.toml
+++ b/providers/nano-gpt/models/Sao10K/L3.1-70B-Hanami-x1.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/Steelskull/L3.3-Cu-Mai-R1-70b.toml b/providers/nano-gpt/models/Steelskull/L3.3-Cu-Mai-R1-70b.toml
index 449b394b5b4..1b9b19b5e64 100644
--- a/providers/nano-gpt/models/Steelskull/L3.3-Cu-Mai-R1-70b.toml
+++ b/providers/nano-gpt/models/Steelskull/L3.3-Cu-Mai-R1-70b.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/Steelskull/L3.3-Electra-R1-70b.toml b/providers/nano-gpt/models/Steelskull/L3.3-Electra-R1-70b.toml
index a8966b4cb59..ecd8bdf659d 100644
--- a/providers/nano-gpt/models/Steelskull/L3.3-Electra-R1-70b.toml
+++ b/providers/nano-gpt/models/Steelskull/L3.3-Electra-R1-70b.toml
@@ -15,8 +15,8 @@ output = 0.69989
 cache_read = 0.349945
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/Steelskull/L3.3-MS-Nevoria-70b.toml b/providers/nano-gpt/models/Steelskull/L3.3-MS-Nevoria-70b.toml
index 416ea29ff99..bfaee82a467 100644
--- a/providers/nano-gpt/models/Steelskull/L3.3-MS-Nevoria-70b.toml
+++ b/providers/nano-gpt/models/Steelskull/L3.3-MS-Nevoria-70b.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/Steelskull/L3.3-Nevoria-R1-70b.toml b/providers/nano-gpt/models/Steelskull/L3.3-Nevoria-R1-70b.toml
index 596436d3fca..cdd4ff13a24 100644
--- a/providers/nano-gpt/models/Steelskull/L3.3-Nevoria-R1-70b.toml
+++ b/providers/nano-gpt/models/Steelskull/L3.3-Nevoria-R1-70b.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml b/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml
index da0ae193f80..e9d130676e5 100644
--- a/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml
+++ b/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml
@@ -10,4 +10,4 @@ cache_read = 0.04
 [limit]
 context = 1_048_576
 input = 1_048_576
-output = 1_048_576
+output = 393_216
diff --git a/providers/nano-gpt/models/TEE/kimi-k3.toml b/providers/nano-gpt/models/TEE/kimi-k3.toml
index d8557392864..d4a33300b22 100644
--- a/providers/nano-gpt/models/TEE/kimi-k3.toml
+++ b/providers/nano-gpt/models/TEE/kimi-k3.toml
@@ -12,7 +12,7 @@ cache_read = 1.5
 
 [limit]
 input = 1_048_576
-output = 1_048_576
+output = 65_535
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v2.toml b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v2.toml
index 5f84c90905a..cf112ef6ace 100644
--- a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v2.toml
+++ b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v2.toml
@@ -15,8 +15,8 @@ output = 0.1207
 cache_read = 0.05015
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
index 51f9accdebc..356a664e624 100644
--- a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
+++ b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
@@ -17,7 +17,7 @@ cache_read = 0.16
 [limit]
 context = 131_072
 input = 131_072
-output = 131_072
+output = 117_964
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.toml b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.toml
index e1d5cba4614..c623200bfe3 100644
--- a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.toml
+++ b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.toml
@@ -15,8 +15,8 @@ output = 0.2414
 cache_read = 0.1003
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml b/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
index c0ddb13b127..07e3e20f35d 100644
--- a/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
+++ b/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
@@ -15,9 +15,9 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 8_192
-input = 8_192
-output = 8_192
+context = 32_768
+input = 32_768
+output = 26_214
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/nano-gpt/models/TheDrummer/skyfall-36b-v2.toml b/providers/nano-gpt/models/TheDrummer/skyfall-36b-v2.toml
index 710156ef1b7..13f7c7eff9a 100644
--- a/providers/nano-gpt/models/TheDrummer/skyfall-36b-v2.toml
+++ b/providers/nano-gpt/models/TheDrummer/skyfall-36b-v2.toml
@@ -15,9 +15,9 @@ output = 0.8
 cache_read = 0.25
 
 [limit]
-context = 32_000
-input = 32_000
-output = 32_768
+context = 32_768
+input = 32_768
+output = 29_491
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml b/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
index bd61bb32c3a..c68686084a4 100644
--- a/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
+++ b/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
@@ -15,8 +15,8 @@ output = 1.8
 cache_read = 0.2
 
 [limit]
-context = 128_000
-input = 128_000
+context = 131_072
+input = 131_072
 output = 131_072
 
 [modalities]
diff --git a/providers/nano-gpt/models/bytedance-seed/seed-2-1-turbo.toml b/providers/nano-gpt/models/bytedance-seed/seed-2-1-turbo.toml
index 4ca66a0f23f..9c809fe5ded 100644
--- a/providers/nano-gpt/models/bytedance-seed/seed-2-1-turbo.toml
+++ b/providers/nano-gpt/models/bytedance-seed/seed-2-1-turbo.toml
@@ -18,7 +18,7 @@ cache_read = 0.25
 [limit]
 context = 262_144
 input = 262_144
-output = 262_144
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml b/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml
index ca9047a7d38..b4ab0f2f7a0 100644
--- a/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml
+++ b/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml
@@ -17,7 +17,7 @@ cache_read = 0.1
 [limit]
 context = 128_000
 input = 128_000
-output = 131_072
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/deepseek-ai/DeepSeek-R1-0528.toml b/providers/nano-gpt/models/deepseek-ai/DeepSeek-R1-0528.toml
index 54d5251a668..d527629cafb 100644
--- a/providers/nano-gpt/models/deepseek-ai/DeepSeek-R1-0528.toml
+++ b/providers/nano-gpt/models/deepseek-ai/DeepSeek-R1-0528.toml
@@ -16,9 +16,9 @@ output = 1.7
 cache_read = 0.2
 
 [limit]
-context = 128_000
-input = 128_000
-output = 163_840
+context = 163_840
+input = 163_840
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/dots-studio/dots-3-note-preview.toml b/providers/nano-gpt/models/dots-studio/dots-3-note-preview.toml
deleted file mode 100644
index 6f3af28a475..00000000000
--- a/providers/nano-gpt/models/dots-studio/dots-3-note-preview.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Dots3-Note Preview"
-description = "Dots Studio's open-weight multimodal Mixture-of-Experts model activates 16B of 280B parameters for long-context reasoning, coding, visual and document understanding, tool use, and long-horizon agent workflows. Prompts and completions of this model may be logged."
-release_date = "2026-08-15"
-last_updated = "2026-08-15"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 393_216
-input = 393_216
-output = 393_216
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/glm-z1-air.toml b/providers/nano-gpt/models/glm-z1-air.toml
deleted file mode 100644
index b5111e5aa66..00000000000
--- a/providers/nano-gpt/models/glm-z1-air.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GLM Z1 Air"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
-release_date = "2025-04-15"
-last_updated = "2025-04-15"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.07
-output = 0.07
-cache_read = 0.035
-
-[limit]
-context = 32_000
-input = 32_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/holo3-35b-a3b.toml b/providers/nano-gpt/models/holo3-35b-a3b.toml
index 33930849ca6..b4db0baf2a2 100644
--- a/providers/nano-gpt/models/holo3-35b-a3b.toml
+++ b/providers/nano-gpt/models/holo3-35b-a3b.toml
@@ -20,7 +20,7 @@ cache_read = 0.125
 [limit]
 context = 65_536
 input = 65_536
-output = 65_536
+output = 8_192
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/holo3-35b-a3b:thinking.toml b/providers/nano-gpt/models/holo3-35b-a3b:thinking.toml
index dfbfcfaedaa..d9afd26c8f6 100644
--- a/providers/nano-gpt/models/holo3-35b-a3b:thinking.toml
+++ b/providers/nano-gpt/models/holo3-35b-a3b:thinking.toml
@@ -18,7 +18,7 @@ cache_read = 0.125
 [limit]
 context = 65_536
 input = 65_536
-output = 65_536
+output = 8_192
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/huihui-ai/Llama-3.3-70B-Instruct-abliterated.toml b/providers/nano-gpt/models/huihui-ai/Llama-3.3-70B-Instruct-abliterated.toml
index 6451f7718af..e8d090b4d21 100644
--- a/providers/nano-gpt/models/huihui-ai/Llama-3.3-70B-Instruct-abliterated.toml
+++ b/providers/nano-gpt/models/huihui-ai/Llama-3.3-70B-Instruct-abliterated.toml
@@ -15,8 +15,8 @@ output = 0.7
 cache_read = 0.35
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml b/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
index 4cef656c271..c17c8b75739 100644
--- a/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
+++ b/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
@@ -18,7 +18,7 @@ cache_read = 0.05
 [limit]
 context = 131_072
 input = 131_072
-output = 131_072
+output = 117_964
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/meganova-ai/manta-pro-1.0.toml b/providers/nano-gpt/models/meganova-ai/manta-pro-1.0.toml
index 6c2c9eb1e8a..94d60025005 100644
--- a/providers/nano-gpt/models/meganova-ai/manta-pro-1.0.toml
+++ b/providers/nano-gpt/models/meganova-ai/manta-pro-1.0.toml
@@ -15,8 +15,8 @@ output = 0.5
 cache_read = 0.03
 
 [limit]
-context = 32_768
-input = 32_768
+context = 65_536
+input = 65_536
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/meta/muse-glimmer-30b.toml b/providers/nano-gpt/models/meta/muse-glimmer-30b.toml
index 61a7bd6b3bf..f2db2ba5f63 100644
--- a/providers/nano-gpt/models/meta/muse-glimmer-30b.toml
+++ b/providers/nano-gpt/models/meta/muse-glimmer-30b.toml
@@ -11,3 +11,4 @@ cache_read = 0.04
 
 [limit]
 input = 131_072
+output = 117_964
diff --git a/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml b/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
index 027bb7eb4a4..e7f724c55a3 100644
--- a/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
+++ b/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
@@ -17,7 +17,7 @@ cache_read = 0.05
 [limit]
 context = 128_000
 input = 128_000
-output = 131_072
+output = 102_400
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/mistralai/mistral-large.toml b/providers/nano-gpt/models/mistralai/mistral-large.toml
index 57daad375cb..e2b013f694f 100644
--- a/providers/nano-gpt/models/mistralai/mistral-large.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-large.toml
@@ -17,7 +17,7 @@ cache_read = 0.2
 [limit]
 context = 128_000
 input = 128_000
-output = 256_000
+output = 102_400
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/mistralai/mistral-saba.toml b/providers/nano-gpt/models/mistralai/mistral-saba.toml
index e7fa089e08a..f734da7f7ef 100644
--- a/providers/nano-gpt/models/mistralai/mistral-saba.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-saba.toml
@@ -15,9 +15,9 @@ output = 0.595
 cache_read = 0.09945
 
 [limit]
-context = 32_000
-input = 32_000
-output = 32_768
+context = 32_768
+input = 32_768
+output = 26_214
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/mistralai/mixtral-8x22b-instruct-v0.1.toml b/providers/nano-gpt/models/mistralai/mixtral-8x22b-instruct-v0.1.toml
index 4046941d9fb..86ba96a048d 100644
--- a/providers/nano-gpt/models/mistralai/mixtral-8x22b-instruct-v0.1.toml
+++ b/providers/nano-gpt/models/mistralai/mixtral-8x22b-instruct-v0.1.toml
@@ -17,7 +17,7 @@ cache_read = 0.2
 [limit]
 context = 65_536
 input = 65_536
-output = 65_536
+output = 52_428
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/moonshotai/Kimi-K2-Instruct-0905.toml b/providers/nano-gpt/models/moonshotai/Kimi-K2-Instruct-0905.toml
index 4f6b5e80cee..361a76e0f44 100644
--- a/providers/nano-gpt/models/moonshotai/Kimi-K2-Instruct-0905.toml
+++ b/providers/nano-gpt/models/moonshotai/Kimi-K2-Instruct-0905.toml
@@ -15,9 +15,9 @@ output = 1.8
 cache_read = 0.2
 
 [limit]
-context = 256_000
-input = 256_000
-output = 262_144
+context = 262_144
+input = 262_144
+output = 100_352
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/moonshotai/kimi-k3.toml b/providers/nano-gpt/models/moonshotai/kimi-k3.toml
index fb933a243d2..35abf18bf41 100644
--- a/providers/nano-gpt/models/moonshotai/kimi-k3.toml
+++ b/providers/nano-gpt/models/moonshotai/kimi-k3.toml
@@ -11,4 +11,4 @@ cache_read = 0.2
 
 [limit]
 input = 1_048_576
-output = 1_048_576
+output = 943_718
diff --git a/providers/nano-gpt/models/moonshotai/kimi-latest.toml b/providers/nano-gpt/models/moonshotai/kimi-latest.toml
index d4fb34a816a..fd837ffa80f 100644
--- a/providers/nano-gpt/models/moonshotai/kimi-latest.toml
+++ b/providers/nano-gpt/models/moonshotai/kimi-latest.toml
@@ -22,7 +22,7 @@ cache_read = 0.2
 [limit]
 context = 1_048_576
 input = 1_048_576
-output = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml b/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
index f968a2c4a61..0f217c50f68 100644
--- a/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
+++ b/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
@@ -21,7 +21,7 @@ cache_read = 0.0025
 [limit]
 context = 262_144
 input = 262_144
-output = 262_144
+output = 235_929
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml b/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
index 3b5671f942f..4cb8ac34370 100644
--- a/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
+++ b/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
@@ -21,7 +21,7 @@ cache_read = 0.25
 [limit]
 context = 262_144
 input = 262_144
-output = 262_144
+output = 235_929
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/nothingiisreal/L3.1-70B-Celeste-V0.1-BF16.toml b/providers/nano-gpt/models/nothingiisreal/L3.1-70B-Celeste-V0.1-BF16.toml
index 186b9deba87..53ac8ca2e8d 100644
--- a/providers/nano-gpt/models/nothingiisreal/L3.1-70B-Celeste-V0.1-BF16.toml
+++ b/providers/nano-gpt/models/nothingiisreal/L3.1-70B-Celeste-V0.1-BF16.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 16_384
 
 [modalities]
diff --git a/providers/nano-gpt/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/nano-gpt/models/nvidia/nemotron-3-nano-30b-a3b.toml
index 03ce1728d5a..63920b4b3c2 100644
--- a/providers/nano-gpt/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/nano-gpt/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -10,5 +10,5 @@ output = 0.68
 cache_read = 0.085
 
 [limit]
-context = 256_000
-input = 256_000
+input = 262_144
+output = 235_929
diff --git a/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml b/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
index 8f2ce8e8aeb..63690302680 100644
--- a/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
+++ b/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
@@ -16,8 +16,8 @@ output = 0.4
 cache_read = 0.05
 
 [limit]
-context = 32_768
-input = 32_768
+context = 131_072
+input = 131_072
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/perplexity-academic-researcher.toml b/providers/nano-gpt/models/perplexity-academic-researcher.toml
index 44b93b1f04e..3cb12eae9f3 100644
--- a/providers/nano-gpt/models/perplexity-academic-researcher.toml
+++ b/providers/nano-gpt/models/perplexity-academic-researcher.toml
@@ -15,9 +15,9 @@ output = 8
 cache_read = 1
 
 [limit]
-context = 127_000
-input = 127_000
-output = 128_000
+context = 128_000
+input = 128_000
+output = 115_200
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml b/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml
index 66e71cbc4e5..ba1df7fad61 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml
+++ b/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml
@@ -9,6 +9,5 @@ output = 0.5
 cache_read = 0.065
 
 [limit]
-context = 256_000
-input = 256_000
-output = 262_144
+input = 262_144
+output = 235_929
diff --git a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml b/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
index 413764df999..e486bde9b43 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
+++ b/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
@@ -16,9 +16,9 @@ output = 0.5
 cache_read = 0.15
 
 [limit]
-context = 256_000
-input = 256_000
-output = 262_144
+context = 131_072
+input = 131_072
+output = 117_964
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml b/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml
index 7966eaf4786..86078c83dfc 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml
+++ b/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml
@@ -9,6 +9,6 @@ output = 0.65
 cache_read = 0.075
 
 [limit]
-context = 256_000
-input = 256_000
-output = 262_144
+context = 262_144
+input = 262_144
+output = 235_929
diff --git a/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml b/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml
index ac6a9dc7839..cfdc9a4897d 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml
+++ b/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml
@@ -9,6 +9,4 @@ output = 1.2
 cache_read = 0.15
 
 [limit]
-context = 128_000
-input = 128_000
-output = 262_144
+input = 131_072
diff --git a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml b/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
index 8648d0a8ccd..a6d75477904 100644
--- a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
+++ b/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
@@ -14,9 +14,9 @@ output = 0.69989
 cache_read = 0.349945
 
 [limit]
-context = 32_000
-input = 32_000
-output = 32_768
+context = 128_000
+input = 128_000
+output = 115_200
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/sonar-deep-research.toml b/providers/nano-gpt/models/sonar-deep-research.toml
index e24af61da6b..a14277752d8 100644
--- a/providers/nano-gpt/models/sonar-deep-research.toml
+++ b/providers/nano-gpt/models/sonar-deep-research.toml
@@ -9,6 +9,5 @@ output = 13.6
 cache_read = 1.7
 
 [limit]
-context = 60_000
-input = 60_000
-output = 128_000
+input = 128_000
+output = 115_200
diff --git a/providers/nano-gpt/models/sonar-reasoning-pro.toml b/providers/nano-gpt/models/sonar-reasoning-pro.toml
index aaff135ee6b..99d46d6010b 100644
--- a/providers/nano-gpt/models/sonar-reasoning-pro.toml
+++ b/providers/nano-gpt/models/sonar-reasoning-pro.toml
@@ -10,9 +10,8 @@ output = 8
 cache_read = 1
 
 [limit]
-context = 127_000
-input = 127_000
-output = 128_000
+input = 128_000
+output = 115_200
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/sonar.toml b/providers/nano-gpt/models/sonar.toml
index c2e65f7ee1f..015df323ea3 100644
--- a/providers/nano-gpt/models/sonar.toml
+++ b/providers/nano-gpt/models/sonar.toml
@@ -8,6 +8,6 @@ output = 1
 cache_read = 0.5
 
 [limit]
-context = 127_000
-input = 127_000
-output = 128_000
+context = 127_072
+input = 127_072
+output = 114_364
diff --git a/providers/nano-gpt/models/soob3123/GrayLine-Qwen3-8B.toml b/providers/nano-gpt/models/soob3123/GrayLine-Qwen3-8B.toml
index 2feaa84c2b0..1190cb8c632 100644
--- a/providers/nano-gpt/models/soob3123/GrayLine-Qwen3-8B.toml
+++ b/providers/nano-gpt/models/soob3123/GrayLine-Qwen3-8B.toml
@@ -15,8 +15,8 @@ output = 0.3
 cache_read = 0.15
 
 [limit]
-context = 16_384
-input = 16_384
+context = 32_768
+input = 32_768
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/stepfun-ai/step-3.5-flash-2603.toml b/providers/nano-gpt/models/stepfun-ai/step-3.5-flash-2603.toml
index 581f487aac5..c2a46b10c85 100644
--- a/providers/nano-gpt/models/stepfun-ai/step-3.5-flash-2603.toml
+++ b/providers/nano-gpt/models/stepfun-ai/step-3.5-flash-2603.toml
@@ -11,3 +11,8 @@ values = ["low", "medium", "high"]
 input = 0.1
 output = 0.3
 cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 65_536
diff --git a/providers/nano-gpt/models/stepfun-ai/step-3.5-flash.toml b/providers/nano-gpt/models/stepfun-ai/step-3.5-flash.toml
index d45fbad1853..c7f7d5d3510 100644
--- a/providers/nano-gpt/models/stepfun-ai/step-3.5-flash.toml
+++ b/providers/nano-gpt/models/stepfun-ai/step-3.5-flash.toml
@@ -10,3 +10,8 @@ values = ["low", "medium", "high"]
 input = 0.1
 output = 0.3
 cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 65_536
diff --git a/providers/nano-gpt/models/tencent/hy3.toml b/providers/nano-gpt/models/tencent/hy3.toml
index c8d5b00fac1..2e87496bb3a 100644
--- a/providers/nano-gpt/models/tencent/hy3.toml
+++ b/providers/nano-gpt/models/tencent/hy3.toml
@@ -14,4 +14,3 @@ cache_read = 0.029
 [limit]
 context = 262_144
 input = 262_144
-output = 262_144
diff --git a/providers/nano-gpt/models/unsloth/gemma-3-12b-it.toml b/providers/nano-gpt/models/unsloth/gemma-3-12b-it.toml
index c8a69735228..fc3ddaae4e8 100644
--- a/providers/nano-gpt/models/unsloth/gemma-3-12b-it.toml
+++ b/providers/nano-gpt/models/unsloth/gemma-3-12b-it.toml
@@ -15,9 +15,9 @@ output = 0.272
 cache_read = 0.136
 
 [limit]
-context = 128_000
-input = 128_000
-output = 131_072
+context = 131_072
+input = 131_072
+output = 16_384
 
 [modalities]
 input = ["text", "pdf"]
diff --git a/providers/nano-gpt/models/upstage/solar-pro-3.toml b/providers/nano-gpt/models/upstage/solar-pro-3.toml
index bf315c2409d..b007f270ad0 100644
--- a/providers/nano-gpt/models/upstage/solar-pro-3.toml
+++ b/providers/nano-gpt/models/upstage/solar-pro-3.toml
@@ -16,9 +16,9 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 128_000
-input = 128_000
-output = 128_000
+context = 131_072
+input = 131_072
+output = 117_964
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/x-ai/grok-4.3.toml b/providers/nano-gpt/models/x-ai/grok-4.3.toml
index 62ed7bd7f23..8ffee4ccf4d 100644
--- a/providers/nano-gpt/models/x-ai/grok-4.3.toml
+++ b/providers/nano-gpt/models/x-ai/grok-4.3.toml
@@ -12,7 +12,7 @@ cache_read = 0.2
 
 [limit]
 input = 1_000_000
-output = 1_000_000
+output = 900_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/x-ai/grok-4.5.toml b/providers/nano-gpt/models/x-ai/grok-4.5.toml
index a77fe23ecfd..fa9d5d73dbe 100644
--- a/providers/nano-gpt/models/x-ai/grok-4.5.toml
+++ b/providers/nano-gpt/models/x-ai/grok-4.5.toml
@@ -11,3 +11,4 @@ cache_read = 0.5
 
 [limit]
 input = 500_000
+output = 450_000
diff --git a/providers/nano-gpt/models/x-ai/grok-4.6.toml b/providers/nano-gpt/models/x-ai/grok-4.6.toml
index 44110f62ce5..f40e0799ed7 100644
--- a/providers/nano-gpt/models/x-ai/grok-4.6.toml
+++ b/providers/nano-gpt/models/x-ai/grok-4.6.toml
@@ -11,3 +11,4 @@ cache_read = 0.5
 
 [limit]
 input = 500_000
+output = 450_000
diff --git a/providers/nano-gpt/models/x-ai/grok-build-0.1.toml b/providers/nano-gpt/models/x-ai/grok-build-0.1.toml
index 53eb6a95c3d..820097adddf 100644
--- a/providers/nano-gpt/models/x-ai/grok-build-0.1.toml
+++ b/providers/nano-gpt/models/x-ai/grok-build-0.1.toml
@@ -12,6 +12,7 @@ cache_read = 0.2
 
 [limit]
 input = 256_000
+output = 230_400
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/x-ai/grok-latest.toml b/providers/nano-gpt/models/x-ai/grok-latest.toml
index c747136ab59..ca0fe338dd6 100644
--- a/providers/nano-gpt/models/x-ai/grok-latest.toml
+++ b/providers/nano-gpt/models/x-ai/grok-latest.toml
@@ -22,7 +22,7 @@ cache_read = 0.5
 [limit]
 context = 500_000
 input = 500_000
-output = 500_000
+output = 450_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
index 66af8b5bcc7..4969196b493 100644
--- a/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo.toml
@@ -15,9 +15,9 @@ output = 3
 cache_read = 0.5
 
 [limit]
-context = 200_000
-input = 200_000
-output = 204_800
+context = 204_800
+input = 204_800
+output = 131_072
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
index 8151551a0e2..6a53fc7eeff 100644
--- a/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/GLM-4.6-turbo:thinking.toml
@@ -16,9 +16,9 @@ output = 3
 cache_read = 0.5
 
 [limit]
-context = 200_000
-input = 200_000
-output = 204_800
+context = 204_800
+input = 204_800
+output = 131_072
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/glm-4.5v.toml b/providers/nano-gpt/models/z-ai/glm-4.5v.toml
index fa0c7912302..1d268f0dde8 100644
--- a/providers/nano-gpt/models/z-ai/glm-4.5v.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.5v.toml
@@ -12,8 +12,8 @@ output = 1.8
 cache_read = 0.3
 
 [limit]
-input = 64_000
-output = 96_000
+context = 65_536
+input = 65_536
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nano-gpt/models/z-ai/glm-4.5v:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.5v:thinking.toml
index df038ad9ff1..7419ab22313 100644
--- a/providers/nano-gpt/models/z-ai/glm-4.5v:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.5v:thinking.toml
@@ -10,8 +10,8 @@ output = 1.8
 cache_read = 0.3
 
 [limit]
-input = 64_000
-output = 96_000
+context = 65_536
+input = 65_536
 
 [modalities]
 input = ["text", "image"]

From ca8640f9828351b9380b2d5e79ae1198eb5f5ba4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 21:25:18 +0000
Subject: [PATCH 0588/2012] chore(sync): update OpenRouter model catalog
 (#6228)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 07235ccdb08..92d05b69dd5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.030776
-output = 2.061552
-cache_read = 0.085898
+input = 1.02747
+output = 2.05494
+cache_read = 0.085623
 
 [limit]
 context = 1_048_576

From 362ee5cf5a954261da015d3d273832affef4be03 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 22:24:52 +0000
Subject: [PATCH 0589/2012] chore(sync): update OpenRouter model catalog
 (#6231)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 92d05b69dd5..e9c8db3462a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02747
-output = 2.05494
-cache_read = 0.085623
+input = 1.024164
+output = 2.048328
+cache_read = 0.085347
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8119fb87874..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 639aea7c111..51793fd9815 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.148
-output = 3.608
-cache_read = 0.1886
+input = 1.15
+output = 3.5
+cache_read = 0.1
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From ab130b6467fceb9bd5d18c0bd05111980698fad8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 22:25:42 +0000
Subject: [PATCH 0590/2012] chore(sync): update Kilo model catalog (#6230)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-flash-latest.toml |  8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml       | 10 +++++-----
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 7b848afc9df..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 62973e4f574..82fdc44362b 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.148
-output = 3.608
-cache_read = 0.1886
+input = 1.15
+output = 3.5
+cache_read = 0.1
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From df4239845062a579d30b283773fa81233e325e93 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 23:22:17 +0000
Subject: [PATCH 0591/2012] chore(sync): update OpenRouter model catalog
 (#6232)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e9c8db3462a..42833ea0214 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.024164
-output = 2.048328
-cache_read = 0.085347
+input = 1.022424
+output = 2.044848
+cache_read = 0.085202
 
 [limit]
 context = 1_048_576

From 43b38947e53c7ad56b19eb1881d3a4fb442f346d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 00:58:13 +0000
Subject: [PATCH 0592/2012] chore(sync): update OpenRouter model catalog
 (#6234)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml        |  6 +++---
 .../models/deepseek/deepseek-v4-pro-0813.toml     |  6 +++---
 .../models/deepseek/deepseek-v4-pro.toml          |  6 +++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml |  8 ++++----
 .../nvidia/nemotron-3.5-content-safety.toml       | 15 +++++++++++++++
 providers/openrouter/models/tencent/hy3.toml      |  6 +++---
 .../models/undi95/remm-slerp-l2-13b.toml          |  4 ++--
 7 files changed, 33 insertions(+), 18 deletions(-)
 create mode 100644 providers/openrouter/models/nvidia/nemotron-3.5-content-safety.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 42833ea0214..66443ade4c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.022424
-output = 2.044848
-cache_read = 0.085202
+input = 1.04226
+output = 2.08452
+cache_read = 0.086855
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 5dcd117448e..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 182_520
+output = 32_768
diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-content-safety.toml b/providers/openrouter/models/nvidia/nemotron-3.5-content-safety.toml
new file mode 100644
index 00000000000..c5cb87fdd33
--- /dev/null
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-content-safety.toml
@@ -0,0 +1,15 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "nvidia/nemotron-3.5-content-safety"
+structured_output = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.2
+output = 0.2
+
+[limit]
+context = 131_072
+output = 117_964
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/undi95/remm-slerp-l2-13b.toml b/providers/openrouter/models/undi95/remm-slerp-l2-13b.toml
index 1f9dedc63f2..abddf28ecbc 100644
--- a/providers/openrouter/models/undi95/remm-slerp-l2-13b.toml
+++ b/providers/openrouter/models/undi95/remm-slerp-l2-13b.toml
@@ -11,12 +11,12 @@ knowledge = "2023-06-30"
 open_weights = true
 
 [cost]
-input = 0.45
+input = 0.35
 output = 0.65
 
 [limit]
 context = 6_144
-output = 4_096
+output = 5_529
 
 [modalities]
 input = ["text"]

From a37eca84b8b387d9e44393fc33b7d8e592bb1db9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 00:58:17 +0000
Subject: [PATCH 0593/2012] chore(sync): update Kilo model catalog (#6233)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml |  4 ++--
 .../nvidia/nemotron-3.5-content-safety.toml       | 15 +++++++++++++++
 providers/kilo/models/tencent/hy3.toml            |  6 +++---
 .../kilo/models/undi95/remm-slerp-l2-13b.toml     |  2 +-
 4 files changed, 21 insertions(+), 6 deletions(-)
 create mode 100644 providers/kilo/models/nvidia/nemotron-3.5-content-safety.toml

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 25c72a0b0fb..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 202_800
-output = 182_520
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/nvidia/nemotron-3.5-content-safety.toml b/providers/kilo/models/nvidia/nemotron-3.5-content-safety.toml
new file mode 100644
index 00000000000..9dd8f8ac54c
--- /dev/null
+++ b/providers/kilo/models/nvidia/nemotron-3.5-content-safety.toml
@@ -0,0 +1,15 @@
+base_model = "nvidia/nemotron-3.5-content-safety"
+description = "NVIDIA Nemotron 3.5 Content Safety is a compact 4B-parameter multimodal guardrail model from NVIDIA, fine-tuned from Google Gemma-3-4B. It moderates both inputs to and responses from LLMs and VLMs, accepting..."
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.2
+output = 0.2
+
+[limit]
+context = 131_072
+output = 117_964
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/undi95/remm-slerp-l2-13b.toml b/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
index 96a93a9047f..40ab1560284 100644
--- a/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
+++ b/providers/kilo/models/undi95/remm-slerp-l2-13b.toml
@@ -15,7 +15,7 @@ output = 0.65
 
 [limit]
 context = 6_144
-output = 4_096
+output = 5_529
 
 [modalities]
 input = ["text"]

From afc51ee6bf3602a05fec27368b95169dc5615931 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 01:32:08 +0000
Subject: [PATCH 0594/2012] chore(sync): update OpenRouter model catalog
 (#6235)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 6e50c8a6bcb76d0af32535336484b8d6b2940964 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 01:32:24 +0000
Subject: [PATCH 0595/2012] chore(sync): update NanoGPT model catalog (#6236)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From d7cd229d133df5bb547c4c4d4f1095e63e04f0ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 03:29:53 +0000
Subject: [PATCH 0596/2012] chore(sync): update CrossModel model catalog
 (#6238)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/gemini/gemini-3.8-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/crossmodel/models/gemini/gemini-3.8-flash.toml

diff --git a/providers/crossmodel/models/gemini/gemini-3.8-flash.toml b/providers/crossmodel/models/gemini/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..88a1cb041bf
--- /dev/null
+++ b/providers/crossmodel/models/gemini/gemini-3.8-flash.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.75
+
+[modalities]
+input = ["text", "image", "audio", "video"]

From 736401aa1f5032643f578196f4f0ea2f9e31fdbc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 04:29:51 +0000
Subject: [PATCH 0597/2012] chore(sync): update NanoGPT model catalog (#6239)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From f4311f3dc1e574d558b184043b446776fd85ad31 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 04:29:58 +0000
Subject: [PATCH 0598/2012] chore(sync): update OpenRouter model catalog
 (#6240)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 92d960a94397043092ae17b39dd3bae9274be0a2 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 4 Sep 2026 13:14:39 +0800
Subject: [PATCH 0599/2012] feat(opencode-go): add Omen Alpha

---
 providers/opencode-go/models/omen-alpha.toml | 27 ++++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/opencode-go/models/omen-alpha.toml

diff --git a/providers/opencode-go/models/omen-alpha.toml b/providers/opencode-go/models/omen-alpha.toml
new file mode 100644
index 00000000000..fb6b531bf56
--- /dev/null
+++ b/providers/opencode-go/models/omen-alpha.toml
@@ -0,0 +1,27 @@
+name = "Omen Alpha"
+description = "Stealth multimodal reasoning model for coding, agentic tasks, and tool use"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.2
+output = 0.66
+cache_read = 0.04
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]

From 7257e128ccf26c5713185bec56f4a3c9d95a380d Mon Sep 17 00:00:00 2001
From: Clayton Ford <103367913+claybford@users.noreply.github.com>
Date: Fri, 4 Sep 2026 01:15:57 -0400
Subject: [PATCH 0600/2012] fireworks-ai: add low effort tier to glm-5p3 and
 glm-5p3-flash (#6237)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Both models accept reasoning_effort=low as a real, distinct tier on the
Fireworks API — verified empirically (2026-09-04): the same prompt yields
~10/25/198 reasoning tokens for low/high/max on glm-5p3 and ~9/17/559 on
glm-5p3-flash. The previous comment claimed the GLM 5.2 two-tier collapse
(low/medium -> high) applied; it does not for the 5.3 family. This matches
the upstream z.ai entries for glm-5.3 / glm-5.3-flash, which already list
low/high/max.

Co-authored-by: claybford <claybford@users.noreply.github.com>
---
 .../accounts/fireworks/models/glm-5p3-flash.toml     | 11 +++++++----
 .../models/accounts/fireworks/models/glm-5p3.toml    | 12 +++++++-----
 2 files changed, 14 insertions(+), 9 deletions(-)

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
index 7e8d208090a..8a1203370b8 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
@@ -1,14 +1,17 @@
 # Pricing: https://fireworks.ai/models/fireworks/glm-5p3-flash (accessed 2026-09-02)
 # Released on Fireworks Serverless API 2026-08-26.
-# Thinking-only model: reasoning cannot be disabled. Use the same two effective
-# tiers as Fireworks GLM 5.3: low/medium collapse to high; max/xhigh select max.
-# https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-08-29)
+# Thinking-only model: reasoning cannot be disabled. Three distinct effort
+# tiers like Fireworks GLM 5.3 (NOT the two-tier GLM 5.2 collapse): verified
+# empirically 2026-09-04 — same prompt yields ~9/17/559 reasoning tokens for
+# low/high/max. max/xhigh select max.
+# https://docs.fireworks.ai/api-reference/post-chatcompletions
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
+last_updated = "2026-09-04"
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
index 8fb8037dd97..fa2c2de0072 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
@@ -1,16 +1,18 @@
 # Pricing: https://docs.fireworks.ai/serverless/pricing (accessed 2026-08-29)
 # Released on Fireworks Serverless API 2026-08-28.
 # Thinking-only model: reasoning cannot be disabled (reasoning_effort="none"
-# and thinking.type="disabled" both return 400). Two effective tiers like
-# GLM 5.2: low/medium collapse to high, max/xhigh select max.
-# https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-08-29)
+# and thinking.type="disabled" both return 400). Unlike GLM 5.2 (two tiers,
+# low/medium collapse to high), GLM 5.3 has three distinct effort tiers:
+# verified empirically 2026-09-04 — same prompt yields ~10/25/198 reasoning
+# tokens for low/high/max. max/xhigh select max.
+# https://docs.fireworks.ai/api-reference/post-chatcompletions
 base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
-last_updated = "2026-08-28"
+last_updated = "2026-09-04"
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"

From f476b2d34eec46afc7bdb5f735943312fe29ed45 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 00:16:10 -0500
Subject: [PATCH 0601/2012] fix: [missing-model] github-copilot:
 gemini-3.8-flash (#6229)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/gemini-3.8-flash.toml              | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/github-copilot/models/gemini-3.8-flash.toml

diff --git a/providers/github-copilot/models/gemini-3.8-flash.toml b/providers/github-copilot/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..74432656991
--- /dev/null
+++ b/providers/github-copilot/models/gemini-3.8-flash.toml
@@ -0,0 +1,21 @@
+# Sources:
+# - https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# - https://docs.github.com/en/copilot/reference/ai-models/supported-models
+# - https://api.githubcopilot.com/models
+# Pricing: Gemini 3.8 Flash promo $0.75/$0.075/$3.75 per 1M tokens through 2026-12-31
+# Reasoning: effort low|medium|high (matches lab + github-copilot gemini-3.7-flash peer)
+base_model = "google/gemini-3.8-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+
+[limit]
+context = 1_000_000
+input = 936_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From e23e43b35fb484aced7cb10e4ac2d476e7cf636c Mon Sep 17 00:00:00 2001
From: m3 <Marvae@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:19:12 +0800
Subject: [PATCH 0602/2012] chore(github-copilot): refresh model catalog
 (#6218)

* chore(github-copilot): refresh model catalog

Add Claude Fable 5.1 and drop models GitHub has retired from Copilot.

* feat(github-copilot): add Gemini 3.8 Flash

---------

Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
---
 .../models/claude-fable-5.1.toml               | 11 +++++++++++
 .../github-copilot/models/claude-opus-4.5.toml | 12 ------------
 .../github-copilot/models/claude-opus-4.6.toml | 17 -----------------
 .../models/claude-sonnet-4.5.toml              | 12 ------------
 .../github-copilot/models/claude-sonnet-4.toml | 13 -------------
 .../models/gemini-3.1-pro-preview.toml         | 18 ------------------
 providers/github-copilot/models/gpt-4.1.toml   | 11 -----------
 .../github-copilot/models/gpt-5.2-codex.toml   |  7 -------
 providers/github-copilot/models/gpt-5.2.toml   |  7 -------
 9 files changed, 11 insertions(+), 97 deletions(-)
 create mode 100644 providers/github-copilot/models/claude-fable-5.1.toml
 delete mode 100644 providers/github-copilot/models/claude-opus-4.5.toml
 delete mode 100644 providers/github-copilot/models/claude-opus-4.6.toml
 delete mode 100644 providers/github-copilot/models/claude-sonnet-4.5.toml
 delete mode 100644 providers/github-copilot/models/claude-sonnet-4.toml
 delete mode 100644 providers/github-copilot/models/gemini-3.1-pro-preview.toml
 delete mode 100644 providers/github-copilot/models/gpt-4.1.toml
 delete mode 100644 providers/github-copilot/models/gpt-5.2-codex.toml
 delete mode 100644 providers/github-copilot/models/gpt-5.2.toml

diff --git a/providers/github-copilot/models/claude-fable-5.1.toml b/providers/github-copilot/models/claude-fable-5.1.toml
new file mode 100644
index 00000000000..8cc14c5ec61
--- /dev/null
+++ b/providers/github-copilot/models/claude-fable-5.1.toml
@@ -0,0 +1,11 @@
+# Sources:
+# - https://github.blog/changelog/2026-09-01-claude-fable-5-1-generally-available-in-github-copilot
+# - https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/github-copilot/models/claude-opus-4.5.toml b/providers/github-copilot/models/claude-opus-4.5.toml
deleted file mode 100644
index 7d1d43dd490..00000000000
--- a/providers/github-copilot/models/claude-opus-4.5.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 32_000 }]
-
-[cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
-
-[limit]
-input = 168_000
-output = 32_000
diff --git a/providers/github-copilot/models/claude-opus-4.6.toml b/providers/github-copilot/models/claude-opus-4.6.toml
deleted file mode 100644
index fbbf9fd23e2..00000000000
--- a/providers/github-copilot/models/claude-opus-4.6.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }]
-
-[cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
-
-[limit]
-context = 200_000
-input = 168_000
-output = 32_000
-
-[experimental.modes.fast]
-cost = { input = 30, output = 150, cache_read = 3, cache_write = 37.5 }
-provider = { body = { speed = "fast" }, headers = { anthropic-beta = "fast-mode-2026-02-01" } }
diff --git a/providers/github-copilot/models/claude-sonnet-4.5.toml b/providers/github-copilot/models/claude-sonnet-4.5.toml
deleted file mode 100644
index 096431de139..00000000000
--- a/providers/github-copilot/models/claude-sonnet-4.5.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 32_000 }]
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
-
-[limit]
-input = 168_000
-output = 32_000
diff --git a/providers/github-copilot/models/claude-sonnet-4.toml b/providers/github-copilot/models/claude-sonnet-4.toml
deleted file mode 100644
index 9e9ade55dea..00000000000
--- a/providers/github-copilot/models/claude-sonnet-4.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
-
-[limit]
-context = 216_000
-input = 128_000
-output = 16_000
diff --git a/providers/github-copilot/models/gemini-3.1-pro-preview.toml b/providers/github-copilot/models/gemini-3.1-pro-preview.toml
deleted file mode 100644
index 01991d61de1..00000000000
--- a/providers/github-copilot/models/gemini-3.1-pro-preview.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "google/gemini-3.1-pro-preview"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 256, max = 32_000 }]
-
-[cost]
-input = 2
-output = 12
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 18
-cache_read = 0.4
-
-[limit]
-context = 1_000_000
-input = 936_000
-output = 64_000
diff --git a/providers/github-copilot/models/gpt-4.1.toml b/providers/github-copilot/models/gpt-4.1.toml
deleted file mode 100644
index e8b167e7baa..00000000000
--- a/providers/github-copilot/models/gpt-4.1.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "openai/gpt-4.1"
-
-[cost]
-input = 2
-output = 8
-cache_read = 0.5
-
-[limit]
-context = 128_000
-input = 128_000
-output = 16_384
diff --git a/providers/github-copilot/models/gpt-5.2-codex.toml b/providers/github-copilot/models/gpt-5.2-codex.toml
deleted file mode 100644
index a03382c9d90..00000000000
--- a/providers/github-copilot/models/gpt-5.2-codex.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "openai/gpt-5.2-codex"
-reasoning_options = []
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175
diff --git a/providers/github-copilot/models/gpt-5.2.toml b/providers/github-copilot/models/gpt-5.2.toml
deleted file mode 100644
index 4f38ef3d4dc..00000000000
--- a/providers/github-copilot/models/gpt-5.2.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "openai/gpt-5.2"
-reasoning_options = []
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175

From 3076b42ec119e0d2d506bf871bc51e840b2880ff Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 00:19:21 -0500
Subject: [PATCH 0603/2012] fix: [missing-model] cloudflare-ai-gateway:
 anthropic/claude-fable-5.1 (#6217)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/cloudflare-ai-gateway/curation.toml | 12 ++++++++++
 .../models/anthropic/claude-fable-5.1.toml    | 23 +++++++++++++++++++
 2 files changed, 35 insertions(+)
 create mode 100644 providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml

diff --git a/providers/cloudflare-ai-gateway/curation.toml b/providers/cloudflare-ai-gateway/curation.toml
index 8772ea87336..697217a5ffb 100644
--- a/providers/cloudflare-ai-gateway/curation.toml
+++ b/providers/cloudflare-ai-gateway/curation.toml
@@ -39,6 +39,18 @@ skip = [
 structured_output = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
+# Claude Fable 5.1: adaptive thinking always on; caller control is effort only (no thinking toggle).
+# Wire: output_config.effort = low|medium|high|xhigh|max on the Anthropic Messages path.
+# https://platform.claude.com/docs/en/build-with-claude/effort
+# https://platform.claude.com/docs/en/about-claude/pricing
+[models."anthropic/claude-fable-5.1"]
+structured_output = true
+note = [
+  "Effort: output_config.effort = low|medium|high|xhigh|max (Anthropic Messages API; adaptive thinking always on).",
+  "Sources: https://platform.claude.com/docs/en/build-with-claude/effort ; https://platform.claude.com/docs/en/about-claude/pricing",
+]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
 [models."anthropic/claude-haiku-4.5"]
 structured_output = true
 reasoning_options = [{ type = "budget_tokens", min = 1024 }]
diff --git a/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml b/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml
new file mode 100644
index 00000000000..56d384093a6
--- /dev/null
+++ b/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml
@@ -0,0 +1,23 @@
+# Effort: output_config.effort = low|medium|high|xhigh|max (Anthropic Messages API; adaptive thinking always on).
+# Sources:
+# - https://platform.claude.com/docs/en/build-with-claude/effort
+# - https://platform.claude.com/docs/en/about-claude/pricing
+# - https://platform.claude.com/docs/en/models/overview
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[limit]
+context = 1_000_000
+
+[provider]
+npm = "@ai-sdk/anthropic"

From 2577be9d52a764d5a51d50bfaaf7f8534b856fb4 Mon Sep 17 00:00:00 2001
From: MicroHEROX <140800253+MicroHEROX@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:19:43 +0800
Subject: [PATCH 0604/2012] fix(amd): Qwen3.8-Flash-Next accepts image input on
 the AMD host (#6200)

---
 providers/amd/models/Qwen3.8-Flash-Next.toml | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/providers/amd/models/Qwen3.8-Flash-Next.toml b/providers/amd/models/Qwen3.8-Flash-Next.toml
index 290a819bfdd..3f2fb729032 100644
--- a/providers/amd/models/Qwen3.8-Flash-Next.toml
+++ b/providers/amd/models/Qwen3.8-Flash-Next.toml
@@ -1,16 +1,17 @@
-# Sources (accessed 2026-08-28):
+# Sources (accessed 2026-09-03):
 # https://developer.amd.com.cn/radeon/tokenfactory
 # Pricing from the Token Factory models API: input $0.15, output $0.47,
 # cache read $0.016 per MTok.
-# AMD hosts a text-only vllm-router deployment (input/output modalities are
-# text only), so the multimodal base entry is overridden below.
+# The AMD deployment accepts image input (GET /v1/models reports text+image
+# modalities) and image understanding was verified live: a rendered-digit
+# test image was read back correctly. The multimodal base entry is therefore
+# overridden to text+image (no video on this host).
 # Reasoning control verified live against POST /v1/chat/completions:
 # reasoning_effort is a validated enum low|medium|xhigh with xhigh the default
 # (values outside the enum return 400; no none, no toggle). No
 # reasoning_content is ever returned (stream + non-stream), so there is no
 # toggle and no interleaved channel on this host.
 base_model = "alibaba/qwen3.8-flash-next"
-attachment = false
 reasoning_options = [
   { type = "effort", values = ["low", "medium", "xhigh"] },
 ]
@@ -21,5 +22,5 @@ output = 0.47
 cache_read = 0.016
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]

From ce1b6f35056d17d9f9c3225fd81a0dcd83e7366b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E8=BD=BB=E5=B0=98?=
 <53001305+SlightDust@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:20:15 +0800
Subject: [PATCH 0605/2012] feat: add GLM-5.3, GLM-5.3-Flash, Qwen3.8-Flash to
 SCNet Token Plan (#6168)

---
 .../scnet-token-plan/models/GLM-5.3-Flash.toml      | 10 ++++++++++
 providers/scnet-token-plan/models/GLM-5.3.toml      | 10 ++++++++++
 .../scnet-token-plan/models/Qwen3.8-Flash.toml      | 13 +++++++++++++
 3 files changed, 33 insertions(+)
 create mode 100644 providers/scnet-token-plan/models/GLM-5.3-Flash.toml
 create mode 100644 providers/scnet-token-plan/models/GLM-5.3.toml
 create mode 100644 providers/scnet-token-plan/models/Qwen3.8-Flash.toml

diff --git a/providers/scnet-token-plan/models/GLM-5.3-Flash.toml b/providers/scnet-token-plan/models/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..3267da0b1b4
--- /dev/null
+++ b/providers/scnet-token-plan/models/GLM-5.3-Flash.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/scnet-token-plan/models/GLM-5.3.toml b/providers/scnet-token-plan/models/GLM-5.3.toml
new file mode 100644
index 00000000000..0c44ce127bf
--- /dev/null
+++ b/providers/scnet-token-plan/models/GLM-5.3.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/scnet-token-plan/models/Qwen3.8-Flash.toml b/providers/scnet-token-plan/models/Qwen3.8-Flash.toml
new file mode 100644
index 00000000000..cdc9432554b
--- /dev/null
+++ b/providers/scnet-token-plan/models/Qwen3.8-Flash.toml
@@ -0,0 +1,13 @@
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort only applies to DeepSeek-V4 series (high|max); not documented for Qwen3 series.
+# https://www.scnet.cn/ac/openapi/doc/2.0/moduleapi/api/chat.html (accessed 2026-09-01)
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0

From 95851b92c1e8010feba98351d3b37ecc2d5380e2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 00:23:22 -0500
Subject: [PATCH 0606/2012] chore(sync): update Baseten model catalog (#6139)

* chore(sync): update Baseten model catalog

* fix(baseten): add GLM 5.3 Fast reasoning effort options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../baseten/models/zai-org/GLM-5.3-Fast.toml  | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/baseten/models/zai-org/GLM-5.3-Fast.toml

diff --git a/providers/baseten/models/zai-org/GLM-5.3-Fast.toml b/providers/baseten/models/zai-org/GLM-5.3-Fast.toml
new file mode 100644
index 00000000000..bbee46d8d36
--- /dev/null
+++ b/providers/baseten/models/zai-org/GLM-5.3-Fast.toml
@@ -0,0 +1,20 @@
+# Effort: top-level reasoning_effort = none|low|high|max (default high).
+# none requests a direct answer but may still return short reasoning_content;
+# chat_template_args.enable_thinking=false is rejected, so no separate toggle.
+# https://docs.baseten.co/inference/model-apis/reasoning
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 Fast"
+description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+attachment = true
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
+
+[cost]
+input = 2.1
+output = 6.6
+
+[limit]
+context = 1_048_576
+output = 262_144
+
+[modalities]
+input = ["text", "image"]

From f1f0c1398f2d7511dd85e3dc5d8e621c5843d7bc Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Fri, 4 Sep 2026 01:24:54 -0400
Subject: [PATCH 0607/2012] update zen model

---
 providers/opencode-go/models/omen-alpha.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/opencode-go/models/omen-alpha.toml b/providers/opencode-go/models/omen-alpha.toml
index fb6b531bf56..dd02fb4d829 100644
--- a/providers/opencode-go/models/omen-alpha.toml
+++ b/providers/opencode-go/models/omen-alpha.toml
@@ -1,10 +1,10 @@
 name = "Omen Alpha"
-description = "Stealth multimodal reasoning model for coding, agentic tasks, and tool use"
+description = "oH man anothEr aLPha ModEl"
 release_date = "2026-09-04"
 last_updated = "2026-09-04"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["low", "high"] }]
 temperature = true
 tool_call = true
 structured_output = true
@@ -19,9 +19,9 @@ output = 0.66
 cache_read = 0.04
 
 [limit]
-context = 1_000_000
-output = 131_072
+context = 500_000
+output = 128_000
 
 [modalities]
-input = ["text", "image", "video", "pdf"]
+input = ["text", "image"]
 output = ["text"]

From 2756cc7dd54b930aad20bdf5eb06392a8628bdd0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 05:26:22 +0000
Subject: [PATCH 0608/2012] chore(sync): update Eden AI model catalog (#6247)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../accounts/fireworks/models/inkling.toml         | 14 ++++++++++++++
 .../edenai/models/qwen/deepseek-v4-flash-0731.toml |  4 ++--
 .../edenai/models/qwen/deepseek-v4-pro-0813.toml   |  4 ++--
 3 files changed, 18 insertions(+), 4 deletions(-)
 create mode 100644 providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml

diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml
new file mode 100644
index 00000000000..add2a32eee5
--- /dev/null
+++ b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml
@@ -0,0 +1,14 @@
+base_model = "thinkingmachines/inkling"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "max"]
+
+[cost]
+input = 1
+output = 4.05
+cache_read = 0.17
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 1f11f70d4ba..d5d8773c4b6 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 7e31b605e19..05bf93bee1e 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From 1f34cc08d3e9ab4e0c05380f0069eb8d673f25a0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 05:26:33 +0000
Subject: [PATCH 0609/2012] chore(sync): update Cloudflare AI Gateway model
 catalog (#6248)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-fable-5.1.toml                   | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml b/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml
index 56d384093a6..4ff3a9fffa1 100644
--- a/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml
+++ b/providers/cloudflare-ai-gateway/models/anthropic/claude-fable-5.1.toml
@@ -1,8 +1,5 @@
 # Effort: output_config.effort = low|medium|high|xhigh|max (Anthropic Messages API; adaptive thinking always on).
-# Sources:
-# - https://platform.claude.com/docs/en/build-with-claude/effort
-# - https://platform.claude.com/docs/en/about-claude/pricing
-# - https://platform.claude.com/docs/en/models/overview
+# Sources: https://platform.claude.com/docs/en/build-with-claude/effort ; https://platform.claude.com/docs/en/about-claude/pricing
 base_model = "anthropic/claude-fable-5-1"
 structured_output = true
 

From 4945104dd74bf70ffd3648de697c1a80d2c3933f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 06:39:57 +0000
Subject: [PATCH 0610/2012] chore(sync): update NanoGPT model catalog (#6257)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/qwen/qwen3-235b-a22b.toml         | 5 ++---
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/qwen/qwen3-235b-a22b.toml b/providers/nano-gpt/models/qwen/qwen3-235b-a22b.toml
index 4438be6a0c3..dab4411afa2 100644
--- a/providers/nano-gpt/models/qwen/qwen3-235b-a22b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-235b-a22b.toml
@@ -11,9 +11,8 @@ output = 0.5
 cache_read = 0.15
 
 [limit]
-context = 41_000
-input = 41_000
-output = 32_768
+context = 262_144
+input = 262_144
 
 [modalities]
 input = ["text", "pdf"]

From e22fbc314230e4283d2903aca3f420efeb4b7c98 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 06:40:05 +0000
Subject: [PATCH 0611/2012] chore(sync): update OpenRouter model catalog
 (#6256)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 37e5711e7a3edfb049e387d243719a498e9d2f63 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 06:40:51 +0000
Subject: [PATCH 0612/2012] chore(sync): update CrossModel model catalog
 (#6258)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/crossmodel/models/gemini/gemini-3.6-flash.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/crossmodel/models/gemini/gemini-3.6-flash.toml b/providers/crossmodel/models/gemini/gemini-3.6-flash.toml
index 94454c82587..4a8ddc3475d 100644
--- a/providers/crossmodel/models/gemini/gemini-3.6-flash.toml
+++ b/providers/crossmodel/models/gemini/gemini-3.6-flash.toml
@@ -8,10 +8,10 @@ type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
-cache_write = 1.5
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.75
 
 [modalities]
 input = ["text", "image", "audio", "video"]

From ba816a069564b9dbf7c61cc117a79301b3f9ecdc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 07:28:31 +0000
Subject: [PATCH 0613/2012] chore(sync): update NanoGPT model catalog (#6259)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../Steelskull/L3.3-MS-Evayale-70B.toml       | 24 -------------------
 .../TheDrummer/UnslopNemo-12B-v4.1.toml       |  4 ++--
 .../pamanseau/OpenReasoning-Nemotron-32B.toml |  4 ++--
 .../models/qwen25-vl-72b-instruct.toml        |  4 ++--
 4 files changed, 6 insertions(+), 30 deletions(-)
 delete mode 100644 providers/nano-gpt/models/Steelskull/L3.3-MS-Evayale-70B.toml

diff --git a/providers/nano-gpt/models/Steelskull/L3.3-MS-Evayale-70B.toml b/providers/nano-gpt/models/Steelskull/L3.3-MS-Evayale-70B.toml
deleted file mode 100644
index a446e987fbd..00000000000
--- a/providers/nano-gpt/models/Steelskull/L3.3-MS-Evayale-70B.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Evayale 70b "
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-01-01"
-last_updated = "2024-12-06"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.493
-output = 0.493
-cache_read = 0.2465
-
-[limit]
-context = 16_384
-input = 16_384
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml b/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
index 07e3e20f35d..6f130d0c2f8 100644
--- a/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
+++ b/providers/nano-gpt/models/TheDrummer/UnslopNemo-12B-v4.1.toml
@@ -15,8 +15,8 @@ output = 0.493
 cache_read = 0.2465
 
 [limit]
-context = 32_768
-input = 32_768
+context = 8_192
+input = 8_192
 output = 26_214
 
 [modalities]
diff --git a/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml b/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
index 63690302680..8f2ce8e8aeb 100644
--- a/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
+++ b/providers/nano-gpt/models/pamanseau/OpenReasoning-Nemotron-32B.toml
@@ -16,8 +16,8 @@ output = 0.4
 cache_read = 0.05
 
 [limit]
-context = 131_072
-input = 131_072
+context = 32_768
+input = 32_768
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml b/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
index a6d75477904..4815829fbbe 100644
--- a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
+++ b/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
@@ -14,8 +14,8 @@ output = 0.69989
 cache_read = 0.349945
 
 [limit]
-context = 128_000
-input = 128_000
+context = 32_000
+input = 32_000
 output = 115_200
 
 [modalities]

From 9ca56fa66a3ee9204fff2bac03c5c981569c4aea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 08:34:06 +0000
Subject: [PATCH 0614/2012] chore(sync): update NanoGPT model catalog (#6260)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/qwen3-8b.toml | 27 +++++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/nano-gpt/models/TEE/qwen3-8b.toml

diff --git a/providers/nano-gpt/models/TEE/qwen3-8b.toml b/providers/nano-gpt/models/TEE/qwen3-8b.toml
new file mode 100644
index 00000000000..cf35e5e6ed6
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/qwen3-8b.toml
@@ -0,0 +1,27 @@
+name = "Qwen3 8B TEE"
+description = "Qwen3 8B is an open-weight dense language model from Alibaba's Qwen team for efficient dialogue, reasoning, mathematics, coding, and tool use. Running inside a TEE (Trusted Execution Environment), with provider attestation support."
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.11
+output = 0.45
+cache_read = 0.11
+
+[limit]
+context = 40_960
+input = 40_960
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 3f1be67a217c63fd8604dc334a0e14ddcdee61fc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 08:34:35 +0000
Subject: [PATCH 0615/2012] chore(sync): update Cortecs model catalog (#6261)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/glm-4.7.toml         | 15 ------------
 providers/cortecs/models/hermes-4-70b.toml    | 23 -------------------
 .../models/qwen2.5-vl-72b-instruct.toml       |  4 ++--
 providers/cortecs/models/qwen3-32b.toml       |  8 +++----
 4 files changed, 6 insertions(+), 44 deletions(-)
 delete mode 100644 providers/cortecs/models/glm-4.7.toml
 delete mode 100644 providers/cortecs/models/hermes-4-70b.toml

diff --git a/providers/cortecs/models/glm-4.7.toml b/providers/cortecs/models/glm-4.7.toml
deleted file mode 100644
index e3b731338d6..00000000000
--- a/providers/cortecs/models/glm-4.7.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "zhipuai/glm-4.7"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-structured_output = true
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.78
-output = 2.785
-
-[limit]
-context = 202_752
-output = 202_752
diff --git a/providers/cortecs/models/hermes-4-70b.toml b/providers/cortecs/models/hermes-4-70b.toml
deleted file mode 100644
index e640f0f67d5..00000000000
--- a/providers/cortecs/models/hermes-4-70b.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Hermes 4 70B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2023-12"
-open_weights = true
-
-[cost]
-input = 0.129
-output = 0.399
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cortecs/models/qwen2.5-vl-72b-instruct.toml b/providers/cortecs/models/qwen2.5-vl-72b-instruct.toml
index 463981d6339..cff4ec1dda9 100644
--- a/providers/cortecs/models/qwen2.5-vl-72b-instruct.toml
+++ b/providers/cortecs/models/qwen2.5-vl-72b-instruct.toml
@@ -11,8 +11,8 @@ open_weights = false
 reasoning_options = []
 
 [cost]
-input = 0.25
-output = 0.747
+input = 1.014
+output = 1.014
 
 [limit]
 context = 32_000
diff --git a/providers/cortecs/models/qwen3-32b.toml b/providers/cortecs/models/qwen3-32b.toml
index 23c4b00ad58..d6ad27de61b 100644
--- a/providers/cortecs/models/qwen3-32b.toml
+++ b/providers/cortecs/models/qwen3-32b.toml
@@ -4,9 +4,9 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 0.099
-output = 0.299
+input = 0.089
+output = 0.312
 
 [limit]
-context = 40_000
-output = 40_000
+context = 32_000
+output = 32_000

From 2113d98c4fbf25eca1bf678179e5896e69f6aaeb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 09:28:43 +0000
Subject: [PATCH 0616/2012] chore(sync): update Eden AI model catalog (#6263)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/google/gemini-3.8-flash.toml | 12 ++++++------
 providers/edenai/models/vertex/gemini-3.8-flash.toml | 12 ++++++------
 .../edenai/models/vertex/gemini-3.8-flash@eu.toml    | 12 ++++++------
 .../edenai/models/vertex/gemini-3.8-flash@us.toml    | 12 ++++++------
 4 files changed, 24 insertions(+), 24 deletions(-)

diff --git a/providers/edenai/models/google/gemini-3.8-flash.toml b/providers/edenai/models/google/gemini-3.8-flash.toml
index edad81f8322..e1d0abd66e5 100644
--- a/providers/edenai/models/google/gemini-3.8-flash.toml
+++ b/providers/edenai/models/google/gemini-3.8-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash.toml b/providers/edenai/models/vertex/gemini-3.8-flash.toml
index edad81f8322..e1d0abd66e5 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
index 0937543bf13..3fbeb6b0ae9 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@us.toml b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
index 9e5f2a9b227..b0a19cbb5df 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 1.5
-output = 7.5
-reasoning = 7.5
-cache_read = 0.15
-cache_write = 0.083333
-input_audio = 1.5
+input = 0.75
+output = 3.75
+reasoning = 3.75
+cache_read = 0.075
+cache_write = 0.041667
+input_audio = 0.75

From 035c1828cffc61072845a195120684a57f81be5e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 10:27:01 +0000
Subject: [PATCH 0617/2012] chore(sync): update OpenRouter model catalog
 (#6265)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From cae2fbe616bb9530f5f684e8fca1fbdf9c489280 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 10:27:21 +0000
Subject: [PATCH 0618/2012] chore(sync): update NanoGPT model catalog (#6264)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 8c26bae778d4172f5efed8b3cc333b41960dfeb1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 12:36:44 +0000
Subject: [PATCH 0619/2012] chore(sync): update LLM Gateway model catalog
 (#6272)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../muse-spark-1.2-contributor.toml           | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/meta-contributor/muse-spark-1.2-contributor.toml

diff --git a/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.2-contributor.toml b/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.2-contributor.toml
new file mode 100644
index 00000000000..4573cc740e2
--- /dev/null
+++ b/providers/llmgateway-providers/models/meta-contributor/muse-spark-1.2-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.2 Contributor (Meta Contributor)"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "muse"
+release_date = "2026-08-06"
+last_updated = "2026-08-06"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From aa96facb4dcd131aaa36ea2ad20de012b8c55304 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 12:37:01 +0000
Subject: [PATCH 0620/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6271)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/muse-spark-1.2-contributor.toml    | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/llmgateway/models/muse-spark-1.2-contributor.toml

diff --git a/providers/llmgateway/models/muse-spark-1.2-contributor.toml b/providers/llmgateway/models/muse-spark-1.2-contributor.toml
new file mode 100644
index 00000000000..b26c9987b64
--- /dev/null
+++ b/providers/llmgateway/models/muse-spark-1.2-contributor.toml
@@ -0,0 +1,28 @@
+name = "Muse Spark 1.2 Contributor"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "muse"
+release_date = "2026-08-06"
+last_updated = "2026-08-06"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+cache_read = 0.002
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 5adf0141c74ddc052744a465f004db6bf6921638 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:26:20 +0000
Subject: [PATCH 0621/2012] chore(sync): update Charm Hyper model catalog
 (#6273)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.1.toml      | 6 +++---
 providers/hyper/models/glm-5.toml        | 6 +++---
 providers/hyper/models/kimi-k2.5.toml    | 6 +++---
 providers/hyper/models/minimax-m2.7.toml | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 4c86ffebd03..532a9b2b8f7 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.29
-output = 4.22
-cache_write = 0.645
+input = 1.262
+output = 4.132
+cache_write = 0.631
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index ac7c97dd31d..2be85ce1855 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.93
-output = 2.878
-cache_write = 0.465
+input = 0.86
+output = 2.784
+cache_write = 0.43
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index 3d00be969d8..6629d8c2afd 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5284
-output = 2.785
-cache_write = 0.2642
+input = 0.5144
+output = 2.755
+cache_write = 0.2572
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index e1e2973ecc9..4ca78e9dbd1 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.418
-output = 1.588
-cache_write = 0.209
+input = 0.484
+output = 1.852
+cache_write = 0.242
 
 [limit]
 context = 262_100

From 57b1ecd9c38e8215294ce865a957e5659fbf1120 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:26:45 +0000
Subject: [PATCH 0622/2012] chore(sync): update OpenRouter model catalog
 (#6274)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 66443ade4c4..4bf670bbd00 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.04226
-output = 2.08452
-cache_read = 0.086855
+input = 1.039302
+output = 2.078604
+cache_read = 0.086609
 
 [limit]
 context = 1_048_576

From dcd39e20a4291e8ced2b72b2e3b46cd45efe3d09 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 14:26:12 +0000
Subject: [PATCH 0623/2012] chore(sync): update OpenRouter model catalog
 (#6275)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..613ac412dfa 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08834
+output = 0.17668
+cache_read = 0.017668
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..09d98bc8130 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 1.12068
+output = 3.36204
+cache_read = 0.037356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4bf670bbd00..99f76b93913 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.039302
-output = 2.078604
-cache_read = 0.086609
+input = 1.032342
+output = 2.064684
+cache_read = 0.086029
 
 [limit]
 context = 1_048_576

From 0dff4695250fc02952e51a2b1716451fb17c60af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 15:26:12 +0000
Subject: [PATCH 0624/2012] chore(sync): update Kilo model catalog (#6277)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 80dc12b4ce8..5b7d7be3b0d 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04998
+output = 0.09996
+cache_read = 0.009996
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 131_072
 
 [modalities]
 input = ["text"]

From 5ae77a5fd800975d919e178f9b32a3032e126900 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 15:26:16 +0000
Subject: [PATCH 0625/2012] chore(sync): update GitHub Copilot model catalog
 (#6280)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/github-copilot/models/gpt-5.6-sol.toml | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/github-copilot/models/gpt-5.6-sol.toml b/providers/github-copilot/models/gpt-5.6-sol.toml
index f4370368104..3f18387bc22 100644
--- a/providers/github-copilot/models/gpt-5.6-sol.toml
+++ b/providers/github-copilot/models/gpt-5.6-sol.toml
@@ -6,14 +6,14 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
 input = 4
-output = 15
+output = 20
 cache_read = 0.4
 cache_write = 5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10

From d89ea60f2d874d92b01d7dd2b1523a29c959f78d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 15:26:22 +0000
Subject: [PATCH 0626/2012] chore(sync): update OpenRouter model catalog
 (#6278)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 613ac412dfa..ef9af884050 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08834
-output = 0.17668
-cache_read = 0.017668
+input = 0.08806
+output = 0.17612
+cache_read = 0.017612
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 09d98bc8130..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12068
-output = 3.36204
-cache_read = 0.037356
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 99f76b93913..261f50e9ca1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.032342
-output = 2.064684
-cache_read = 0.086029
+input = 1.02225
+output = 2.0445
+cache_read = 0.085188
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 27d5c88cb9e..4e3d7984114 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04998
+output = 0.09996
+cache_read = 0.009996
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 131_072
 
 [modalities]
 input = ["text"]

From efcc8d6d70c37268c6d868c4f78d10df1a8aea97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 15:26:26 +0000
Subject: [PATCH 0627/2012] chore(sync): update Eden AI model catalog (#6279)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 7 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index f41860fe6d8..7bb385c666d 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.754975
-output = 0.754975
+input = 0.75543
+output = 0.75543
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index e453aff36f2..0079b03c739 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -7,5 +7,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174225
-output = 0.754975
+input = 0.17433
+output = 0.75543
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index d5d8773c4b6..1f11f70d4ba 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 05bf93bee1e..7e31b605e19 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 98720232880..170b09c1bf1 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4646
-output = 0.9292
+input = 0.46488
+output = 0.92976
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index b3e11f0e545..3bc7ab9b2af 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.174225
-output = 0.6969
+input = 0.17433
+output = 0.69732
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 50d0c19815f..65b07937b29 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04535
-output = 1.04535
+input = 1.04598
+output = 1.04598

From 1cd20e61eded18b60e42b786dc4791587f170d26 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 11:08:39 -0500
Subject: [PATCH 0628/2012] fix: gpt 5.6 sol no longer has promotional pricing
 on github copilot (#6255)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/github-copilot/models/gpt-5.6-sol.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/github-copilot/models/gpt-5.6-sol.toml b/providers/github-copilot/models/gpt-5.6-sol.toml
index 3f18387bc22..b43e647a519 100644
--- a/providers/github-copilot/models/gpt-5.6-sol.toml
+++ b/providers/github-copilot/models/gpt-5.6-sol.toml
@@ -1,4 +1,5 @@
 # Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# GPT-5.6 Sol standard rates (promo ended): ≤272K $4/$0.40/$5/$20; >272K $8/$0.80/$10/$30 per 1M
 base_model = "openai/gpt-5.6-sol"
 
 [[reasoning_options]]

From 433ea82c1abec6faab6d7735d8060382f1f65f90 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 16:28:11 +0000
Subject: [PATCH 0629/2012] chore(sync): update OpenRouter model catalog
 (#6281)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++---
 .../models/ibm-granite/granite-4.1-8b.toml    | 24 -------------------
 .../models/qwen/qwen3.5-35b-a3b.toml          |  7 +++---
 .../openrouter/models/qwen/qwen3.6-27b.toml   |  9 +++----
 providers/openrouter/models/tencent/hy3.toml  |  6 ++---
 5 files changed, 12 insertions(+), 40 deletions(-)
 delete mode 100644 providers/openrouter/models/ibm-granite/granite-4.1-8b.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 261f50e9ca1..4e84faf4b10 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.02225
-output = 2.0445
-cache_read = 0.085188
+input = 1.011984
+output = 2.023968
+cache_read = 0.084332
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/ibm-granite/granite-4.1-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.1-8b.toml
deleted file mode 100644
index 3ea6f4851d5..00000000000
--- a/providers/openrouter/models/ibm-granite/granite-4.1-8b.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Granite 4.1 8B"
-description = "Open-weight instruction model for adaptable chat and self-hosted production workloads"
-family = "granite"
-release_date = "2026-04-30"
-last_updated = "2026-04-30"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.05
-output = 0.1
-cache_read = 0.05
-
-[limit]
-context = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index 9c78b7280fd..ed439da675c 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -6,12 +6,11 @@ base_model = "alibaba/qwen3.5-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.25
-output = 1.25
-cache_read = 0.25
+input = 0.08
+output = 0.75
 
 [limit]
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index d35b93066de..15ca4007021 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,12 +6,9 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.6
-output = 3.6
-cache_read = 0.12
-
-[limit]
-output = 235_929
+input = 0.3
+output = 2
+cache_read = 0.03
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 21c8c94a8a199734dc84beb1ebf338bbf709d999 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 16:28:55 +0000
Subject: [PATCH 0630/2012] chore(sync): update Kilo model catalog (#6282)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ibm-granite/granite-4.1-8b.toml    | 24 -------------------
 .../kilo/models/qwen/qwen3.5-35b-a3b.toml     |  2 +-
 providers/kilo/models/qwen/qwen3.6-27b.toml   |  3 ---
 providers/kilo/models/tencent/hy3.toml        |  6 ++---
 4 files changed, 4 insertions(+), 31 deletions(-)
 delete mode 100644 providers/kilo/models/ibm-granite/granite-4.1-8b.toml

diff --git a/providers/kilo/models/ibm-granite/granite-4.1-8b.toml b/providers/kilo/models/ibm-granite/granite-4.1-8b.toml
deleted file mode 100644
index f53347d702d..00000000000
--- a/providers/kilo/models/ibm-granite/granite-4.1-8b.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "IBM: Granite 4.1 8B"
-description = "Tool-capable chat model for instruction following and agentic application workflows"
-family = "granite"
-release_date = "2026-04-30"
-last_updated = "2026-04-30"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.05
-output = 0.1
-cache_read = 0.05
-
-[limit]
-context = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index 43f7cfc71e4..f9494ea15f2 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -10,7 +10,7 @@ input = 0.1625
 output = 1.3
 
 [limit]
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index 7aaf82ee885..61085517bb0 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -9,8 +9,5 @@ values = ["none", "high"]
 input = 0.45
 output = 2.7
 
-[limit]
-output = 235_929
-
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 2fe52a6efa63da62ee54c9b60dbc62a5304dd843 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 11:30:43 -0500
Subject: [PATCH 0631/2012] fix(copilot): ignore retired models during pricing
 sync (#6283)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/github-copilot.ts | 12 +++++-
 packages/core/test/github-copilot.test.ts     | 42 ++++++++++++++++---
 sync.md                                       |  1 +
 3 files changed, 48 insertions(+), 7 deletions(-)

diff --git a/packages/core/src/sync/providers/github-copilot.ts b/packages/core/src/sync/providers/github-copilot.ts
index 5cd810e0f1e..a75abd33d3a 100644
--- a/packages/core/src/sync/providers/github-copilot.ts
+++ b/packages/core/src/sync/providers/github-copilot.ts
@@ -38,7 +38,17 @@ const FILE_ALIASES: Record<string, string> = {
 const IGNORED_ROWS = new Set([
   // Goes in [experimental.modes.fast] under claude-opus-4.8
   "claude-opus-4.8-fast-mode-preview",
-  // No matching model in repo
+  // Retired models can remain in the pricing table; do not rediscover them.
+  // https://docs.github.com/en/copilot/reference/ai-models/supported-models#model-retirement-history
+  // Sonnet 4.6 is still available to annual-plan subscribers and stays eligible.
+  "claude-sonnet-4",
+  "claude-sonnet-4.5",
+  "claude-opus-4.5",
+  "claude-opus-4.6",
+  "gemini-3.1-pro",
+  "gpt-4.1",
+  "gpt-5.2",
+  "gpt-5.2-codex",
   "raptor-mini",
 ]);
 
diff --git a/packages/core/test/github-copilot.test.ts b/packages/core/test/github-copilot.test.ts
index d524665af3b..8111238e234 100644
--- a/packages/core/test/github-copilot.test.ts
+++ b/packages/core/test/github-copilot.test.ts
@@ -150,11 +150,11 @@ test("clears authored tiers and cache_write the table no longer lists", () => {
 
 test("resolves preview and alias filenames", () => {
   const preview = parseGitHubCopilotPricing([
-    row({ model: "Gemini 3.1 Pro", provider: "google", release_status: "Public preview" }),
+    row({ model: "Example Model", release_status: "Public preview" }),
   ]);
   expect(githubCopilot.translateModel(preview[0]!, translationContext({
-    "gemini-3.1-pro-preview": authoredTerra,
-  }))?.id).toBe("gemini-3.1-pro-preview");
+    "example-model-preview": authoredTerra,
+  }))?.id).toBe("example-model-preview");
 
   const alias = parseGitHubCopilotPricing([row({ model: "MAI-Code-1-Flash", provider: "microsoft" })]);
   expect(githubCopilot.translateModel(alias[0]!, translationContext({
@@ -163,9 +163,9 @@ test("resolves preview and alias filenames", () => {
 
   // An exact slug match wins over both fallbacks.
   expect(githubCopilot.translateModel(preview[0]!, translationContext({
-    "gemini-3.1-pro": authoredTerra,
-    "gemini-3.1-pro-preview": authoredTerra,
-  }))?.id).toBe("gemini-3.1-pro");
+    "example-model": authoredTerra,
+    "example-model-preview": authoredTerra,
+  }))?.id).toBe("example-model");
   expect(githubCopilot.translateModel(alias[0]!, translationContext({
     "mai-code-1-flash": authoredTerra,
     "mai-code-1-flash-picker": authoredTerra,
@@ -184,6 +184,36 @@ test("skips ignored rows silently and unmatched rows with an ID", () => {
   expect(githubCopilot.sourceID(unmatched!)).toBe("brand-new-model");
 });
 
+test.each([
+  "Claude Sonnet 4",
+  "Claude Sonnet 4.5",
+  "Claude Opus 4.5",
+  "Claude Opus 4.6",
+  "Gemini 3.1 Pro",
+  "GPT-4.1",
+  "GPT-5.2",
+  "GPT-5.2-Codex",
+  "Raptor mini",
+])("ignores retired %s for translation and missing-model discovery", (name) => {
+  const [model] = parseGitHubCopilotPricing([row({ model: name, release_status: "Public preview" })]);
+  expect(githubCopilot.sourceID(model!)).toBeUndefined();
+  expect(githubCopilot.translateModel(model!, translationContext({}))).toBeUndefined();
+  expect(githubCopilot.translateModel(model!, translationContext({
+    [model!.slug]: authoredTerra,
+  }))).toBeUndefined();
+  expect(githubCopilot.translateModel(model!, translationContext({
+    [`${model!.slug}-preview`]: authoredTerra,
+  }))).toBeUndefined();
+});
+
+test("keeps Sonnet 4.6 eligible for annual-plan subscribers", () => {
+  const [model] = parseGitHubCopilotPricing([row({ model: "Claude Sonnet 4.6", provider: "anthropic" })]);
+  expect(githubCopilot.sourceID(model!)).toBe("claude-sonnet-4.6");
+  expect(githubCopilot.translateModel(model!, translationContext({
+    "claude-sonnet-4.6": authoredTerra,
+  }))?.id).toBe("claude-sonnet-4.6");
+});
+
 const pricingYaml = `
 - model: 'GPT-5.6 Sol[^gpt-56-sol-promo]'
   provider: openai
diff --git a/sync.md b/sync.md
index 2acf38d148a..3e8b9f684ba 100644
--- a/sync.md
+++ b/sync.md
@@ -234,6 +234,7 @@ GitHub Copilot is implemented in `packages/core/src/sync/providers/github-copilo
 - The YML contains only token rates, so the sync only updates `[cost]`: `input`, `cached_input` (as `cache_read`), `cache_write`, `output`, and long-context rows as `cost.tiers`.
 - Display names are converted to file IDs, with minimal special case logic to match existing model entries.
 - Unmatched rows open missing-model issues, and local entries missing from the source are kept.
+- When removing a fully retired Copilot model, add its pricing-table slug to `IGNORED_ROWS` so stale pricing rows cannot trigger translation or missing-model issues. Models still served to some subscribers (such as Sonnet 4.6 on annual plans) remain eligible.
 
 ## xAI Notes
 

From 62915b6fa890985b3bcbb94e560a6fc9d3a5ad80 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 17:22:51 +0000
Subject: [PATCH 0632/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6285)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index 5a002179e53..035cb3a9470 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.13
-output = 0.4
-cache_read = 0.024
+input = 0.1
+output = 0.25
+cache_read = 0.02
 
 [limit]
 context = 1_048_576

From 76bfe9a2ddbdb749563e2c9115e06dbe612edf6b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 17:23:03 +0000
Subject: [PATCH 0633/2012] chore(sync): update LLM Gateway model catalog
 (#6286)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../consensusprotocol/deepseek-v4-flash.toml  |  4 ++--
 .../consensusprotocol/glm-5.3-flash.toml      | 19 +++++++++++++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
index e5c7e364851..e0bff4fb512 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -9,8 +9,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.13
-output = 0.27
+input = 0.08
+output = 0.16
 cache_read = 0.02
 
 [limit]
diff --git a/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
new file mode 100644
index 00000000000..bbc353ac4a5
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
@@ -0,0 +1,19 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (Consensus Protocol)"
+attachment = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.25
+cache_read = 0.02
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text"]

From aff9329adb32f1bc3afba99eb2872a881053ba5c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 17:23:09 +0000
Subject: [PATCH 0634/2012] chore(sync): update OpenRouter model catalog
 (#6288)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4e84faf4b10..a7b4e69f3db 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.011984
-output = 2.023968
-cache_read = 0.084332
+input = 1.001892
+output = 2.003784
+cache_read = 0.083491
 
 [limit]
 context = 1_048_576

From 463be73f01c5270c9556b66bbd175b9ca62f9a41 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:26:04 -0500
Subject: [PATCH 0635/2012] fix: add Muse Spark 1.3 max reasoning effort
 (#6291)

* fix(meta): add Muse Spark 1.3 max reasoning effort

* fix(providers): align Muse Spark 1.3 max effort support

---------

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../llmgateway-providers/models/meta/muse-spark-1.3.toml    | 6 +++++-
 providers/llmgateway/models/muse-spark-1.3.toml             | 6 +++++-
 providers/meta/models/muse-spark-1.3.toml                   | 5 +++--
 providers/vercel/models/meta/muse-spark-1.3.toml            | 2 +-
 4 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
index 37217f9c0c2..55785b2a519 100644
--- a/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
+++ b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
@@ -1,9 +1,13 @@
+# Effort: reasoning_effort or reasoning.effort is forwarded unchanged to Meta's Responses API.
+# https://docs.llmgateway.io/features/reasoning
+# https://dev.meta.ai/docs/reasoning.md (max is Standard-tier 1.3 only)
+# https://github.com/theopenco/llmgateway/blob/910a6632c47b83c74944bad231f25f17bd5acb7b/packages/actions/src/prepare-request-body.ts#L2231-L2241
 base_model = "meta/muse-spark-1.3"
 name = "Muse Spark 1.3 (Meta)"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.25
diff --git a/providers/llmgateway/models/muse-spark-1.3.toml b/providers/llmgateway/models/muse-spark-1.3.toml
index cba766d6589..72dcd9f04a4 100644
--- a/providers/llmgateway/models/muse-spark-1.3.toml
+++ b/providers/llmgateway/models/muse-spark-1.3.toml
@@ -1,8 +1,12 @@
+# Effort: reasoning_effort or reasoning.effort is forwarded unchanged to Meta's Responses API.
+# https://docs.llmgateway.io/features/reasoning
+# https://dev.meta.ai/docs/reasoning.md (max is Standard-tier 1.3 only)
+# https://github.com/theopenco/llmgateway/blob/910a6632c47b83c74944bad231f25f17bd5acb7b/packages/actions/src/prepare-request-body.ts#L2231-L2241
 base_model = "meta/muse-spark-1.3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.25
diff --git a/providers/meta/models/muse-spark-1.3.toml b/providers/meta/models/muse-spark-1.3.toml
index d2c1f6a4e45..fa41397fbca 100644
--- a/providers/meta/models/muse-spark-1.3.toml
+++ b/providers/meta/models/muse-spark-1.3.toml
@@ -3,11 +3,12 @@
 # https://dev.meta.ai/docs/pricing-rate-limits/
 # https://openrouter.ai/api/v1/models/meta/muse-spark-1.3/endpoints (Meta endpoint pricing; accessed 2026-09-02)
 # Standard-tier rates match Muse Spark 1.2; Axios reports 1.3 priced the same as predecessor.
-# Effort: reasoning_effort = minimal|low|medium|high|xhigh (max not yet live per Meta launch post)
+# https://dev.meta.ai/docs/reasoning.md (max is Standard-tier 1.3 only; verified 2026-09-04)
+# Effort: reasoning_effort (Chat Completions) or reasoning.effort (Responses) = minimal|low|medium|high|xhigh|max
 
 base_model = "meta/muse-spark-1.3"
 
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 1.25
diff --git a/providers/vercel/models/meta/muse-spark-1.3.toml b/providers/vercel/models/meta/muse-spark-1.3.toml
index 13867dbedd3..2fb6a5058fe 100644
--- a/providers/vercel/models/meta/muse-spark-1.3.toml
+++ b/providers/vercel/models/meta/muse-spark-1.3.toml
@@ -1,4 +1,4 @@
-# Effort levels match Meta's native model and established relay peers; max is not yet available.
+# Vercel's documented unified effort levels stop at xhigh; native Meta max passthrough is unverified.
 # https://research.meta.ai/blog/introducing-muse-spark-1-3
 # https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
 base_model = "meta/muse-spark-1.3"

From 83e76db3e88cfb2bd836589bc1cc44c75d3f9e1f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 18:28:44 +0000
Subject: [PATCH 0636/2012] chore(sync): update LLM Gateway model catalog
 (#6294)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway-providers/models/meta/muse-spark-1.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
index 55785b2a519..5ffdddfbb9a 100644
--- a/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
+++ b/providers/llmgateway-providers/models/meta/muse-spark-1.3.toml
@@ -7,7 +7,7 @@ name = "Muse Spark 1.3 (Meta)"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25

From dffd69bc490a02483e1d0873c439fe4953c63969 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 18:28:52 +0000
Subject: [PATCH 0637/2012] chore(sync): update OpenRouter model catalog
 (#6293)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ef9af884050..45d3a54eefe 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08806
-output = 0.17612
-cache_read = 0.017612
+input = 0.08778
+output = 0.17556
+cache_read = 0.017556
 
 [limit]
 context = 1_048_576

From fe8d31ef37357d66717ae4fc55fa4c5e7e36af00 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 18:29:02 +0000
Subject: [PATCH 0638/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6292)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/muse-spark-1.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/llmgateway/models/muse-spark-1.3.toml b/providers/llmgateway/models/muse-spark-1.3.toml
index 72dcd9f04a4..ebc722ec500 100644
--- a/providers/llmgateway/models/muse-spark-1.3.toml
+++ b/providers/llmgateway/models/muse-spark-1.3.toml
@@ -6,7 +6,7 @@ base_model = "meta/muse-spark-1.3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 1.25

From 00623bb5280d9b4b78eb1c961afbc75c350553ea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 13:32:16 -0500
Subject: [PATCH 0639/2012] fix(vercel): add verified Muse Spark 1.3 max effort
 (#6295)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/vercel/models/meta/muse-spark-1.3.toml | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/providers/vercel/models/meta/muse-spark-1.3.toml b/providers/vercel/models/meta/muse-spark-1.3.toml
index 2fb6a5058fe..22c06967dca 100644
--- a/providers/vercel/models/meta/muse-spark-1.3.toml
+++ b/providers/vercel/models/meta/muse-spark-1.3.toml
@@ -1,9 +1,12 @@
-# Vercel's documented unified effort levels stop at xhigh; native Meta max passthrough is unverified.
+# Effort: providerOptions.openai.reasoningEffort = minimal|low|medium|high|xhigh|max
+# Verified via Chat Completions with providerOptions.gateway.only = ["meta"] on 2026-09-04.
+# Native max succeeds on standard 1.3 and is rejected on Contributor/1.2; generic reasoning.effort also accepts those unsupported combinations.
 # https://research.meta.ai/blog/introducing-muse-spark-1-3
-# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/advanced
+# https://dev.meta.ai/docs/reasoning.md
 base_model = "meta/muse-spark-1.3"
 description = "Open Llama multimodal model for image understanding and text reasoning"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 1.25

From 378af9302364c12c4436c09ef03d66d7d4d61930 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 19:19:58 +0000
Subject: [PATCH 0640/2012] chore(sync): update OpenRouter model catalog
 (#6296)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../ling-3.0-flash-sante:free.toml            | 28 +++++++++++++++
 .../models/inclusionai/ling-3.0-flash.toml    |  2 +-
 .../meta/muse-spark-1.3-contributor.toml      |  2 +-
 .../models/meta/muse-spark-1.3.toml           |  2 +-
 .../models/openai/gpt-6-astra-pro.toml        | 36 +++++++++++++++++++
 .../openrouter/models/openai/gpt-6-astra.toml | 36 +++++++++++++++++++
 6 files changed, 103 insertions(+), 3 deletions(-)
 create mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-sante:free.toml
 create mode 100644 providers/openrouter/models/openai/gpt-6-astra-pro.toml
 create mode 100644 providers/openrouter/models/openai/gpt-6-astra.toml

diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-sante:free.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-sante:free.toml
new file mode 100644
index 00000000000..c07e7f799ed
--- /dev/null
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-sante:free.toml
@@ -0,0 +1,28 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ling 3.0 Flash Sante (free)"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "ling"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
index 59e76bd4c20..cf79271ade5 100644
--- a/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash.toml
@@ -1,6 +1,6 @@
 # Toggle: reasoning.enabled = true|false
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Ling-3.0-flash"
+name = "Ling 3.0 Flash"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 family = "ling"
 release_date = "2026-07-23"
diff --git a/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml b/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
index 747c8a17772..81136b11398 100644
--- a/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
+++ b/providers/openrouter/models/meta/muse-spark-1.3-contributor.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.1
diff --git a/providers/openrouter/models/meta/muse-spark-1.3.toml b/providers/openrouter/models/meta/muse-spark-1.3.toml
index f5242622051..1a5da91e2e5 100644
--- a/providers/openrouter/models/meta/muse-spark-1.3.toml
+++ b/providers/openrouter/models/meta/muse-spark-1.3.toml
@@ -3,7 +3,7 @@ description = "Open Llama multimodal model for image understanding and text reas
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.25
diff --git a/providers/openrouter/models/openai/gpt-6-astra-pro.toml b/providers/openrouter/models/openai/gpt-6-astra-pro.toml
new file mode 100644
index 00000000000..31554ae7caa
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-astra-pro.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Astra Pro"
+description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-6-astra.toml b/providers/openrouter/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..22bb9ba361f
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-astra.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Astra"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]

From 2acdca31bd1c0481260109c375444d61544455a5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 19:21:58 +0000
Subject: [PATCH 0641/2012] chore(sync): update Kilo model catalog (#6297)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inclusionai/ling-3.0-flash-fin.toml       |  2 +-
 .../inclusionai/ling-3.0-flash-fin:free.toml  |  2 +-
 .../ling-3.0-flash-sante:free.toml            | 27 +++++++++++++++++
 .../models/inclusionai/ling-3.0-flash.toml    |  2 +-
 .../meta/muse-spark-1.3-contributor.toml      |  2 +-
 .../kilo/models/meta/muse-spark-1.3.toml      |  2 +-
 providers/kilo/models/openai/gpt-6-astra.toml | 29 +++++++++++++++++++
 7 files changed, 61 insertions(+), 5 deletions(-)
 create mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-sante:free.toml
 create mode 100644 providers/kilo/models/openai/gpt-6-astra.toml

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml
index 0482b60a0bb..8eefe09bab0 100644
--- a/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-fin.toml
@@ -1,4 +1,4 @@
-name = "Ling 3.0 Flash Fin"
+name = "inclusionAI: Ling 3.0 Flash Fin"
 description = "Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for real-world investment..."
 family = "ling"
 release_date = "2026-08-27"
diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
index ecd3d2975fc..8a567141419 100644
--- a/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-fin:free.toml
@@ -1,4 +1,4 @@
-name = "Ling 3.0 Flash Fin (free)"
+name = "inclusionAI: Ling 3.0 Flash Fin (free)"
 description = "Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for real-world investment..."
 family = "ling"
 release_date = "2026-08-27"
diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-sante:free.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-sante:free.toml
new file mode 100644
index 00000000000..06431f38e8d
--- /dev/null
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-sante:free.toml
@@ -0,0 +1,27 @@
+name = "inclusionAI: Ling 3.0 Flash Sante (free)"
+description = "Ling 3.0 Flash Sante is a health and medicine-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for..."
+family = "ling"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash.toml b/providers/kilo/models/inclusionai/ling-3.0-flash.toml
index 3f459c7644b..aed0e0e6da0 100644
--- a/providers/kilo/models/inclusionai/ling-3.0-flash.toml
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash.toml
@@ -1,4 +1,4 @@
-name = "Ling-3.0-flash"
+name = "inclusionAI: Ling 3.0 Flash"
 description = "*Ling-3.0-flash* is a *124B-parameter Mixture-of-Experts (MoE) model*, with approximately *5.1B parameters activated per token*. The model is designed with *token efficiency and production-scale agentic inference* as key priorities, enabling developers..."
 family = "ling"
 release_date = "2026-07-23"
diff --git a/providers/kilo/models/meta/muse-spark-1.3-contributor.toml b/providers/kilo/models/meta/muse-spark-1.3-contributor.toml
index 04d56d4a05e..f0748ba0c05 100644
--- a/providers/kilo/models/meta/muse-spark-1.3-contributor.toml
+++ b/providers/kilo/models/meta/muse-spark-1.3-contributor.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.1
diff --git a/providers/kilo/models/meta/muse-spark-1.3.toml b/providers/kilo/models/meta/muse-spark-1.3.toml
index 29b23e8ab3a..42aa0a7f8db 100644
--- a/providers/kilo/models/meta/muse-spark-1.3.toml
+++ b/providers/kilo/models/meta/muse-spark-1.3.toml
@@ -3,7 +3,7 @@ description = "Muse Spark 1.3 is a multimodal reasoning model from Meta for long
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.25
diff --git a/providers/kilo/models/openai/gpt-6-astra.toml b/providers/kilo/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..0c3269f52f2
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-astra.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Astra ($$$$)"
+description = "GPT-6 Astra is a frontier reasoning model from OpenAI, suited for complex reasoning, coding, and agentic workflows."
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]

From ff0a8dda2ab8643c591544ce733b64d4b1c98d6f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 20:25:33 +0000
Subject: [PATCH 0642/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6300)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gpt-6-astra.toml | 29 ++++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/llmgateway/models/gpt-6-astra.toml

diff --git a/providers/llmgateway/models/gpt-6-astra.toml b/providers/llmgateway/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..2d589765056
--- /dev/null
+++ b/providers/llmgateway/models/gpt-6-astra.toml
@@ -0,0 +1,29 @@
+name = "GPT-6 Astra"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-03"
+last_updated = "2026-09-03"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+output = 1_050_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From acfeb574d428c8a01983fdf9d2da7cfc6ec6a81d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 20:25:42 +0000
Subject: [PATCH 0643/2012] chore(sync): update Merge Gateway model catalog
 (#6302)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index f30ae74cdb3..35e52350151 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 66500ebe410207d374ad47b599b77cc2edfbc863 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 20:25:47 +0000
Subject: [PATCH 0644/2012] chore(sync): update Kilo model catalog (#6303)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/{gpt-6-astra.toml => gpt-6-astra-pro.toml}  | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
 rename providers/kilo/models/openai/{gpt-6-astra.toml => gpt-6-astra-pro.toml} (56%)

diff --git a/providers/kilo/models/openai/gpt-6-astra.toml b/providers/kilo/models/openai/gpt-6-astra-pro.toml
similarity index 56%
rename from providers/kilo/models/openai/gpt-6-astra.toml
rename to providers/kilo/models/openai/gpt-6-astra-pro.toml
index 0c3269f52f2..e930a0ca690 100644
--- a/providers/kilo/models/openai/gpt-6-astra.toml
+++ b/providers/kilo/models/openai/gpt-6-astra-pro.toml
@@ -1,5 +1,5 @@
-name = "OpenAI: GPT-6 Astra ($$$$)"
-description = "GPT-6 Astra is a frontier reasoning model from OpenAI, suited for complex reasoning, coding, and agentic workflows."
+name = "OpenAI: GPT-6 Astra Pro ($$$$)"
+description = "GPT-6 Astra Pro is the same underlying model as [GPT-6 Astra](https://openrouter.ai/openai/gpt-6-astra), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode"
 family = "gpt"
 release_date = "2026-09-04"
 last_updated = "2026-09-04"

From 5c301b704cafb68ae39cffef6e534396fcce0fe5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 20:25:51 +0000
Subject: [PATCH 0645/2012] chore(sync): update LLM Gateway model catalog
 (#6299)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-6-astra.toml            | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/openai/gpt-6-astra.toml

diff --git a/providers/llmgateway-providers/models/openai/gpt-6-astra.toml b/providers/llmgateway-providers/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..bbb7e5d4ea9
--- /dev/null
+++ b/providers/llmgateway-providers/models/openai/gpt-6-astra.toml
@@ -0,0 +1,29 @@
+name = "GPT-6 Astra (OpenAI)"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-03"
+last_updated = "2026-09-03"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 53c5bdfd7551a4a498eda095e63f2e673e4dd8d2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 20:26:08 +0000
Subject: [PATCH 0646/2012] chore(sync): update OpenRouter model catalog
 (#6304)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 45d3a54eefe..81ffd0604e0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08778
-output = 0.17556
-cache_read = 0.017556
+input = 0.0875
+output = 0.175
+cache_read = 0.0175
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a7b4e69f3db..2b62860fa6a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.001892
-output = 2.003784
-cache_read = 0.083491
+input = 0.9918
+output = 1.9836
+cache_read = 0.08265
 
 [limit]
 context = 1_048_576

From 8425fdf76e8177cfde1eb186e63751a2ebe9560a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:23:07 +0000
Subject: [PATCH 0647/2012] chore(sync): update Venice model catalog (#6306)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-6-astra.toml     | 35 +++++++++++++++++++
 1 file changed, 35 insertions(+)
 create mode 100644 providers/venice/models/openai-gpt-6-astra.toml

diff --git a/providers/venice/models/openai-gpt-6-astra.toml b/providers/venice/models/openai-gpt-6-astra.toml
new file mode 100644
index 00000000000..f9ff45201ef
--- /dev/null
+++ b/providers/venice/models/openai-gpt-6-astra.toml
@@ -0,0 +1,35 @@
+name = "GPT-6 Astra"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 12.5
+output = 62.5
+cache_read = 1.25
+cache_write = 15.625
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 25
+output = 93.75
+cache_read = 2.5
+cache_write = 31.25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 8bda68338e494b571e7a4ec9ce421dfed39686bd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:23:10 +0000
Subject: [PATCH 0648/2012] chore(sync): update NanoGPT model catalog (#6307)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/openai/gpt-6-astra.toml   | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-astra.toml

diff --git a/providers/nano-gpt/models/openai/gpt-6-astra.toml b/providers/nano-gpt/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..08be47088bb
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-astra.toml
@@ -0,0 +1,29 @@
+name = "GPT 6 Astra"
+description = "OpenAI's most capable model for the hardest end-to-end work, with state-of-the-art performance in computer use, browsing, software engineering, science, and professional tasks. Astra is designed to carry long, multistep workflows across code, browsers, and professional software through to a finished result."
+family = "gpt"
+release_date = "2026-09-03"
+last_updated = "2026-09-03"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From c357021dc212c9eb0166a9da6f8e16637b8b2354 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:23:22 +0000
Subject: [PATCH 0649/2012] chore(sync): update Kilo model catalog (#6310)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-6-astra.toml | 29 +++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/kilo/models/openai/gpt-6-astra.toml

diff --git a/providers/kilo/models/openai/gpt-6-astra.toml b/providers/kilo/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..50757636ec0
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-astra.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Astra ($$$$)"
+description = "GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon..."
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]

From a7a697893b2860fcf0fd96324151ef80a0b59475 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:23:28 +0000
Subject: [PATCH 0650/2012] chore(sync): update Merge Gateway model catalog
 (#6309)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index b4bdd294444..0a7a2254ee8 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -13,8 +14,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 35e52350151..f30ae74cdb3 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From e21fdb543acc38dff779faf42d3f31b2540ea953 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:23:32 +0000
Subject: [PATCH 0651/2012] chore(sync): update OpenRouter model catalog
 (#6308)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 81ffd0604e0..45d3a54eefe 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0875
-output = 0.175
-cache_read = 0.0175
+input = 0.08778
+output = 0.17556
+cache_read = 0.017556
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 2b62860fa6a..edc88da0141 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.9918
-output = 1.9836
-cache_read = 0.08265
+input = 0.981708
+output = 1.963416
+cache_read = 0.081809
 
 [limit]
 context = 1_048_576

From 7c965ec5d9b3b25d88f583ffdb818f3e6cf4790d Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Fri, 4 Sep 2026 16:46:26 -0500
Subject: [PATCH 0652/2012] feat(vercel): sync AI Gateway models, add GPT-6
 Astra

---
 models/openai/gpt-6-astra-fast.toml           | 20 +++++++++++++++++
 models/openai/gpt-6-astra.toml                | 20 +++++++++++++++++
 packages/core/src/family.ts                   |  1 +
 .../ling-3.0-flash-sante-free.toml            | 22 +++++++++++++++++++
 .../inclusionai/ling-3.0-flash-sante.toml     | 22 +++++++++++++++++++
 .../models/openai/gpt-6-astra-fast.toml       | 11 ++++++++++
 .../vercel/models/openai/gpt-6-astra.toml     | 11 ++++++++++
 7 files changed, 107 insertions(+)
 create mode 100644 models/openai/gpt-6-astra-fast.toml
 create mode 100644 models/openai/gpt-6-astra.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-sante-free.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-sante.toml
 create mode 100644 providers/vercel/models/openai/gpt-6-astra-fast.toml
 create mode 100644 providers/vercel/models/openai/gpt-6-astra.toml

diff --git a/models/openai/gpt-6-astra-fast.toml b/models/openai/gpt-6-astra-fast.toml
new file mode 100644
index 00000000000..01b748aaf2f
--- /dev/null
+++ b/models/openai/gpt-6-astra-fast.toml
@@ -0,0 +1,20 @@
+name = "GPT-6 Astra (Fast)"
+description = "Fast variant of GPT-6 Astra for low-latency assistance and high-volume workloads."
+family = "gpt-astra"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/models/openai/gpt-6-astra.toml b/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..e40544c0dbf
--- /dev/null
+++ b/models/openai/gpt-6-astra.toml
@@ -0,0 +1,20 @@
+name = "GPT-6 Astra"
+description = "GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation."
+family = "gpt-astra"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/packages/core/src/family.ts b/packages/core/src/family.ts
index 789433fafd8..9f719d80376 100644
--- a/packages/core/src/family.ts
+++ b/packages/core/src/family.ts
@@ -16,6 +16,7 @@ export const ModelFamilyValues = [
   "gpt-sol",
   "gpt-terra",
   "gpt-luna",
+  "gpt-astra",
   "gpt-oss",
   "gpt-image",
 
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-sante-free.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-sante-free.toml
new file mode 100644
index 00000000000..c7910d31b2a
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-sante-free.toml
@@ -0,0 +1,22 @@
+name = "Ling 3.0 Flash Sante (Free)"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "ling"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-sante.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-sante.toml
new file mode 100644
index 00000000000..a4a58b331ba
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-sante.toml
@@ -0,0 +1,22 @@
+name = "Ling 3.0 Flash Sante"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "ling"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/openai/gpt-6-astra-fast.toml b/providers/vercel/models/openai/gpt-6-astra-fast.toml
new file mode 100644
index 00000000000..f5a6474f1e0
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-astra-fast.toml
@@ -0,0 +1,11 @@
+base_model = "openai/gpt-6-astra-fast"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 20
+output = 100
+cache_read = 2
+cache_write = 12.5
diff --git a/providers/vercel/models/openai/gpt-6-astra.toml b/providers/vercel/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..73666a92e17
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-astra.toml
@@ -0,0 +1,11 @@
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5

From f66cbabe0c4db60a22980a1166d1f8b3dcf25180 Mon Sep 17 00:00:00 2001
From: R-Taneja <47066511+R-Taneja@users.noreply.github.com>
Date: Fri, 4 Sep 2026 16:55:43 -0500
Subject: [PATCH 0653/2012] fix(vercel): add 272k context pricing tiers for
 gpt-6-astra from gateway API

---
 providers/vercel/models/openai/gpt-6-astra-fast.toml | 8 ++++++++
 providers/vercel/models/openai/gpt-6-astra.toml      | 8 ++++++++
 2 files changed, 16 insertions(+)

diff --git a/providers/vercel/models/openai/gpt-6-astra-fast.toml b/providers/vercel/models/openai/gpt-6-astra-fast.toml
index f5a6474f1e0..0ab38d43b74 100644
--- a/providers/vercel/models/openai/gpt-6-astra-fast.toml
+++ b/providers/vercel/models/openai/gpt-6-astra-fast.toml
@@ -1,3 +1,4 @@
+# Pricing: https://ai-gateway.vercel.sh/v1/models (2026-09-04), upper band starts at 272,001 prompt tokens
 base_model = "openai/gpt-6-astra-fast"
 
 [[reasoning_options]]
@@ -9,3 +10,10 @@ input = 20
 output = 100
 cache_read = 2
 cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 40
+output = 150
+cache_read = 4
+cache_write = 25
diff --git a/providers/vercel/models/openai/gpt-6-astra.toml b/providers/vercel/models/openai/gpt-6-astra.toml
index 73666a92e17..badf0a32466 100644
--- a/providers/vercel/models/openai/gpt-6-astra.toml
+++ b/providers/vercel/models/openai/gpt-6-astra.toml
@@ -1,3 +1,4 @@
+# Pricing: https://ai-gateway.vercel.sh/v1/models (2026-09-04), upper band starts at 272,001 prompt tokens
 base_model = "openai/gpt-6-astra"
 
 [[reasoning_options]]
@@ -9,3 +10,10 @@ input = 10
 output = 50
 cache_read = 1
 cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25

From 35d600a2a0d76189d40e7505569289af3f7ae7eb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:23:42 +0000
Subject: [PATCH 0654/2012] chore(sync): update Merge Gateway model catalog
 (#6316)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml     |  1 -
 .../models/deepseek/deepseek-v4-flash.toml          |  1 -
 .../models/deepseek/deepseek-v4-pro-0423.toml       |  1 -
 .../models/deepseek/deepseek-v4-pro-0813.toml       |  1 -
 .../models/deepseek/deepseek-v4-pro.toml            |  1 -
 .../gemini-2.5-computer-use-preview-10-2025.toml    |  2 +-
 .../merge-gateway/models/minimax/minimax-m2.1.toml  |  3 +--
 .../models/minimax/minimax-m2.5-highspeed.toml      |  3 +--
 .../merge-gateway/models/minimax/minimax-m2.5.toml  |  3 +--
 .../models/minimax/minimax-m2.7-highspeed.toml      |  3 +--
 .../merge-gateway/models/minimax/minimax-m2.7.toml  |  3 +--
 .../merge-gateway/models/minimax/minimax-m2.toml    |  2 +-
 .../merge-gateway/models/minimax/minimax-m3.toml    |  2 +-
 .../merge-gateway/models/openai/gpt-4-turbo.toml    |  2 +-
 .../models/openai/gpt-4o-2024-05-13.toml            |  2 --
 .../merge-gateway/models/openai/gpt-6-astra.toml    | 13 +++++++++++++
 providers/merge-gateway/models/qwen/qwen-flash.toml |  2 +-
 providers/merge-gateway/models/qwen/qwen-plus.toml  |  2 +-
 .../merge-gateway/models/qwen/qwen3-235b-a22b.toml  |  2 +-
 .../merge-gateway/models/qwen/qwen3-30b-a3b.toml    |  2 +-
 .../models/qwen/qwen3-coder-flash.toml              |  2 +-
 .../merge-gateway/models/qwen/qwen3-coder-plus.toml |  2 +-
 providers/merge-gateway/models/qwen/qwen3-max.toml  |  2 +-
 .../models/qwen/qwen3-next-80b-a3b-instruct.toml    |  2 +-
 .../models/qwen/qwen3-vl-235b-a22b-instruct.toml    |  1 -
 .../models/qwen/qwen3-vl-235b-a22b-thinking.toml    |  1 -
 .../merge-gateway/models/qwen/qwen3-vl-plus.toml    |  2 +-
 .../models/qwen/qwen3.5-122b-a10b.toml              |  8 +++-----
 .../merge-gateway/models/qwen/qwen3.5-27b.toml      |  8 ++++----
 .../merge-gateway/models/qwen/qwen3.5-35b-a3b.toml  |  8 +++-----
 .../models/qwen/qwen3.5-397b-a17b.toml              |  8 ++++----
 .../merge-gateway/models/qwen/qwen3.5-flash.toml    |  1 -
 .../merge-gateway/models/qwen/qwen3.5-plus.toml     |  2 +-
 .../merge-gateway/models/qwen/qwen3.6-35b-a3b.toml  |  1 -
 .../merge-gateway/models/qwen/qwen3.6-flash.toml    |  1 -
 .../models/qwen/qwen3.6-max-preview.toml            |  2 +-
 .../merge-gateway/models/qwen/qwen3.6-plus.toml     |  2 +-
 .../merge-gateway/models/qwen/qwen3.7-max.toml      |  2 +-
 .../merge-gateway/models/qwen/qwen3.7-plus.toml     |  2 +-
 .../merge-gateway/models/qwen/qwen3.8-max.toml      |  2 +-
 providers/merge-gateway/models/zai/glm-4.5v.toml    |  3 +--
 .../merge-gateway/models/zai/glm-5.3-flash.toml     |  7 +++----
 providers/merge-gateway/models/zai/glm-5.3.toml     |  1 -
 43 files changed, 55 insertions(+), 66 deletions(-)
 create mode 100644 providers/merge-gateway/models/openai/gpt-6-astra.toml

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
index 5105d3ff2de..d4498af2eee 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
@@ -1,5 +1,4 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
index da386447376..59726ca8398 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=deepseek%2Fdeepseek-v4-flash (accessed 2026-07-21)
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "deepseek/deepseek-v4-flash"
-structured_output = false
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0423.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0423.toml
index c3e727f5850..a65df5a3d3a 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0423.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0423.toml
@@ -1,5 +1,4 @@
 base_model = "deepseek/deepseek-v4-pro-0423"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
index 62ea3890104..bf23b6fc065 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
@@ -1,5 +1,4 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
index 9de89a96505..79cfd112ba1 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=deepseek%2Fdeepseek-v4-pro (accessed 2026-07-21)
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "deepseek/deepseek-v4-pro"
-structured_output = false
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/merge-gateway/models/google/gemini-2.5-computer-use-preview-10-2025.toml b/providers/merge-gateway/models/google/gemini-2.5-computer-use-preview-10-2025.toml
index d00ff2e821e..eaaa6ecc7df 100644
--- a/providers/merge-gateway/models/google/gemini-2.5-computer-use-preview-10-2025.toml
+++ b/providers/merge-gateway/models/google/gemini-2.5-computer-use-preview-10-2025.toml
@@ -1,6 +1,6 @@
 base_model = "google/gemini-2.5-computer-use-preview-10-2025"
 name = "Gemini 2.5 Computer Use Preview (10-2025)"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.1.toml b/providers/merge-gateway/models/minimax/minimax-m2.1.toml
index 87d1dde34eb..7334a8d1552 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.1.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.1.toml
@@ -1,8 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2.1 (accessed 2026-07-21)
-
 base_model = "minimax/MiniMax-M2.1"
 name = "MiniMax M2.1"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.5-highspeed.toml b/providers/merge-gateway/models/minimax/minimax-m2.5-highspeed.toml
index 7e4dc4ece19..8d3202d8c33 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.5-highspeed.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.5-highspeed.toml
@@ -1,8 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2.5-highspeed (accessed 2026-07-21)
-
 base_model = "minimax/MiniMax-M2.5-highspeed"
 name = "MiniMax M2.5 Highspeed"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.5.toml b/providers/merge-gateway/models/minimax/minimax-m2.5.toml
index c2c6eccd75c..8f4320f5d56 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.5.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.5.toml
@@ -1,8 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2.5 (accessed 2026-07-21)
-
 base_model = "minimax/MiniMax-M2.5"
 name = "MiniMax M2.5"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.7-highspeed.toml b/providers/merge-gateway/models/minimax/minimax-m2.7-highspeed.toml
index a8783c47c5d..1a16cd603e0 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.7-highspeed.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.7-highspeed.toml
@@ -1,8 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2.7-highspeed (accessed 2026-07-21)
-
 base_model = "minimax/MiniMax-M2.7-highspeed"
 name = "MiniMax M2.7 Highspeed"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.7.toml b/providers/merge-gateway/models/minimax/minimax-m2.7.toml
index 296c2ddf4b5..d634a90e2d3 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.7.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.7.toml
@@ -1,8 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2.7 (accessed 2026-07-21)
-
 base_model = "minimax/MiniMax-M2.7"
 name = "MiniMax M2.7"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m2.toml b/providers/merge-gateway/models/minimax/minimax-m2.toml
index 3b35af67523..75f762c9df6 100644
--- a/providers/merge-gateway/models/minimax/minimax-m2.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m2.toml
@@ -1,7 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m2 (accessed 2026-07-21)
 base_model = "minimax/MiniMax-M2"
 name = "MiniMax M2"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/minimax/minimax-m3.toml b/providers/merge-gateway/models/minimax/minimax-m3.toml
index c4c6f10cdf3..c78806e0ae1 100644
--- a/providers/merge-gateway/models/minimax/minimax-m3.toml
+++ b/providers/merge-gateway/models/minimax/minimax-m3.toml
@@ -1,7 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=minimax%2Fminimax-m3 (accessed 2026-07-21)
 base_model = "minimax/MiniMax-M3"
 name = "MiniMax M3"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/openai/gpt-4-turbo.toml b/providers/merge-gateway/models/openai/gpt-4-turbo.toml
index 73a5a029045..a67aace8bc9 100644
--- a/providers/merge-gateway/models/openai/gpt-4-turbo.toml
+++ b/providers/merge-gateway/models/openai/gpt-4-turbo.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=openai%2Fgpt-4-turbo (accessed 2026-07-21)
-
 base_model = "openai/gpt-4-turbo"
+structured_output = true
 
 [cost]
 input = 10
diff --git a/providers/merge-gateway/models/openai/gpt-4o-2024-05-13.toml b/providers/merge-gateway/models/openai/gpt-4o-2024-05-13.toml
index 06df984cfdf..d768c316752 100644
--- a/providers/merge-gateway/models/openai/gpt-4o-2024-05-13.toml
+++ b/providers/merge-gateway/models/openai/gpt-4o-2024-05-13.toml
@@ -1,7 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=openai%2Fgpt-4o-2024-05-13 (accessed 2026-07-21)
-
 base_model = "openai/gpt-4o-2024-05-13"
-structured_output = false
 
 [cost]
 input = 5
diff --git a/providers/merge-gateway/models/openai/gpt-6-astra.toml b/providers/merge-gateway/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..223b48abe9e
--- /dev/null
+++ b/providers/merge-gateway/models/openai/gpt-6-astra.toml
@@ -0,0 +1,13 @@
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen-flash.toml b/providers/merge-gateway/models/qwen/qwen-flash.toml
index 800d7f1a840..ede6824a83a 100644
--- a/providers/merge-gateway/models/qwen/qwen-flash.toml
+++ b/providers/merge-gateway/models/qwen/qwen-flash.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen-flash (accessed 2026-07-21)
 base_model = "alibaba/qwen-flash"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen-plus.toml b/providers/merge-gateway/models/qwen/qwen-plus.toml
index 60a8231c1cc..e6c1f46ca86 100644
--- a/providers/merge-gateway/models/qwen/qwen-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen-plus.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen-plus (accessed 2026-07-21)
 base_model = "alibaba/qwen-plus"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3-235b-a22b.toml b/providers/merge-gateway/models/qwen/qwen3-235b-a22b.toml
index d25433e2540..da7f4cacbff 100644
--- a/providers/merge-gateway/models/qwen/qwen3-235b-a22b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-235b-a22b.toml
@@ -1,7 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-235b-a22b (accessed 2026-07-21)
 base_model = "alibaba/qwen3-235b-a22b"
 name = "Qwen3 235B A22B"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3-30b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3-30b-a3b.toml
index 76084676532..5a9891d7a0e 100644
--- a/providers/merge-gateway/models/qwen/qwen3-30b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-30b-a3b.toml
@@ -1,5 +1,5 @@
 base_model = "alibaba/qwen3-30b-a3b"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3-coder-flash.toml b/providers/merge-gateway/models/qwen/qwen3-coder-flash.toml
index 0fc7a0eddcb..ff224869997 100644
--- a/providers/merge-gateway/models/qwen/qwen3-coder-flash.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-coder-flash.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-coder-flash (accessed 2026-07-21)
 base_model = "alibaba/qwen3-coder-flash"
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.144
diff --git a/providers/merge-gateway/models/qwen/qwen3-coder-plus.toml b/providers/merge-gateway/models/qwen/qwen3-coder-plus.toml
index 11d24d63469..8f25aa3170a 100644
--- a/providers/merge-gateway/models/qwen/qwen3-coder-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-coder-plus.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-coder-plus (accessed 2026-07-21)
 base_model = "alibaba/qwen3-coder-plus"
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.574
diff --git a/providers/merge-gateway/models/qwen/qwen3-max.toml b/providers/merge-gateway/models/qwen/qwen3-max.toml
index ad1210b88ad..858d3349fdb 100644
--- a/providers/merge-gateway/models/qwen/qwen3-max.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-max.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-max (accessed 2026-07-21)
 base_model = "alibaba/qwen3-max"
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.359
diff --git a/providers/merge-gateway/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/merge-gateway/models/qwen/qwen3-next-80b-a3b-instruct.toml
index b94096ac22e..04fc76a7597 100644
--- a/providers/merge-gateway/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-next-80b-a3b-instruct (accessed 2026-07-21)
 base_model = "alibaba/qwen3-next-80b-a3b-instruct"
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.144
diff --git a/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-instruct.toml b/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-instruct.toml
index 488fd9577a4..1d35975631a 100644
--- a/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-instruct.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "alibaba/qwen3-vl-235b-a22b-instruct"
 name = "Qwen3-VL 235B A22B Instruct"
-structured_output = false
 
 [cost]
 input = 0.287
diff --git a/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-thinking.toml b/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-thinking.toml
index 9651a10f77e..534d3c93920 100644
--- a/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-thinking.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-vl-235b-a22b-thinking.toml
@@ -1,6 +1,5 @@
 base_model = "alibaba/qwen3-vl-235b-a22b-thinking"
 name = "Qwen3-VL 235B A22B Thinking"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3-vl-plus.toml b/providers/merge-gateway/models/qwen/qwen3-vl-plus.toml
index 5b007d27718..3658b3c731b 100644
--- a/providers/merge-gateway/models/qwen/qwen3-vl-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen3-vl-plus.toml
@@ -1,7 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3-vl-plus (accessed 2026-07-21)
 base_model = "alibaba/qwen3-vl-plus"
 attachment = true
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index c680bd0f847..b86a0e589cd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,8 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +11,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index 3ef83d21d9d..d7d8d42decc 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-structured_output = false
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 0a7a2254ee8..42aa3ae2951 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,8 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -14,8 +12,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index f30ae74cdb3..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-structured_output = false
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-flash.toml b/providers/merge-gateway/models/qwen/qwen3.5-flash.toml
index 7169f2ffc97..aa448d645ab 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-flash.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-flash.toml
@@ -1,6 +1,5 @@
 base_model = "alibaba/qwen3.5-flash"
 attachment = false
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-plus.toml b/providers/merge-gateway/models/qwen/qwen3.5-plus.toml
index 3e86233bbd7..797948065e1 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-plus.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-plus (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-plus"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.6-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.6-35b-a3b.toml
index 281db930778..f1fa86a3996 100644
--- a/providers/merge-gateway/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.6-35b-a3b.toml
@@ -2,7 +2,6 @@
 base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B A3B"
 attachment = false
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.6-flash.toml b/providers/merge-gateway/models/qwen/qwen3.6-flash.toml
index b5c0bdfda0b..269b1a8b224 100644
--- a/providers/merge-gateway/models/qwen/qwen3.6-flash.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.6-flash.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.6-flash (accessed 2026-07-21)
 base_model = "alibaba/qwen3.6-flash"
 attachment = false
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.6-max-preview.toml b/providers/merge-gateway/models/qwen/qwen3.6-max-preview.toml
index abc8de8a662..26707800ab6 100644
--- a/providers/merge-gateway/models/qwen/qwen3.6-max-preview.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.6-max-preview.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.6-max-preview (accessed 2026-07-21)
 base_model = "alibaba/qwen3.6-max-preview"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/qwen/qwen3.6-plus.toml b/providers/merge-gateway/models/qwen/qwen3.6-plus.toml
index 05999f0ce0f..75d339be349 100644
--- a/providers/merge-gateway/models/qwen/qwen3.6-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.6-plus.toml
@@ -2,7 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.6-plus"
 attachment = false
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.7-max.toml b/providers/merge-gateway/models/qwen/qwen3.7-max.toml
index a719edc7dd2..f3a0658fe5a 100644
--- a/providers/merge-gateway/models/qwen/qwen3.7-max.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.7-max.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.7-max (accessed 2026-07-21)
 base_model = "alibaba/qwen3.7-max"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/qwen/qwen3.7-plus.toml b/providers/merge-gateway/models/qwen/qwen3.7-plus.toml
index 516b5d39a49..e9dd354d93a 100644
--- a/providers/merge-gateway/models/qwen/qwen3.7-plus.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.7-plus.toml
@@ -1,6 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.7-plus (accessed 2026-07-21)
 base_model = "alibaba/qwen3.7-plus"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
diff --git a/providers/merge-gateway/models/qwen/qwen3.8-max.toml b/providers/merge-gateway/models/qwen/qwen3.8-max.toml
index b87024c4974..f79621f2711 100644
--- a/providers/merge-gateway/models/qwen/qwen3.8-max.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.8-max.toml
@@ -1,5 +1,5 @@
 base_model = "alibaba/qwen3.8-max"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/merge-gateway/models/zai/glm-4.5v.toml b/providers/merge-gateway/models/zai/glm-4.5v.toml
index 822f172a0ae..ec63ddc5966 100644
--- a/providers/merge-gateway/models/zai/glm-4.5v.toml
+++ b/providers/merge-gateway/models/zai/glm-4.5v.toml
@@ -1,9 +1,8 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=zai%2Fglm-4.5v (accessed 2026-07-21)
-
 base_model = "zhipuai/glm-4.5v"
 name = "Glm 4.5V"
+structured_output = true
 reasoning_options = []
-structured_output = false
 
 [cost]
 input = 0.6
diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index 3e406e2e86a..9423fa79264 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -1,12 +1,11 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM-5.3 Flash"
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.015
-output = 0.05
-cache_read = 0.003
+input = 0.075
+output = 0.25
+cache_read = 0.015
diff --git a/providers/merge-gateway/models/zai/glm-5.3.toml b/providers/merge-gateway/models/zai/glm-5.3.toml
index 91a5bf985e4..9091b31da95 100644
--- a/providers/merge-gateway/models/zai/glm-5.3.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3.toml
@@ -1,5 +1,4 @@
 base_model = "zhipuai/glm-5.3"
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"

From 87ce4dc0ba75fcc26de7c3c5e3ca3d814211cd97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:23:46 +0000
Subject: [PATCH 0655/2012] chore(sync): update LLM Gateway model catalog
 (#6315)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google-ai-studio/gemini-3.8-flash.toml | 12 ++++++++++++
 .../models/google-vertex/gemini-3.8-flash.toml    |  2 +-
 .../models/novita/glm-5.3-flash.toml              | 15 +++++++++++++++
 3 files changed, 28 insertions(+), 1 deletion(-)
 create mode 100644 providers/llmgateway-providers/models/google-ai-studio/gemini-3.8-flash.toml
 create mode 100644 providers/llmgateway-providers/models/novita/glm-5.3-flash.toml

diff --git a/providers/llmgateway-providers/models/google-ai-studio/gemini-3.8-flash.toml b/providers/llmgateway-providers/models/google-ai-studio/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..74f0925212a
--- /dev/null
+++ b/providers/llmgateway-providers/models/google-ai-studio/gemini-3.8-flash.toml
@@ -0,0 +1,12 @@
+base_model = "google/gemini-3.8-flash"
+name = "Gemini 3.8 Flash (Google AI Studio)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.08333
diff --git a/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml b/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml
index 85d6071eb17..77a81139d72 100644
--- a/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml
+++ b/providers/llmgateway-providers/models/google-vertex/gemini-3.8-flash.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["low", "medium", "high"]
 
 [cost]
 input = 0.75
diff --git a/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml b/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml
new file mode 100644
index 00000000000..15ec49d085e
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (NovitaAI)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+
+[limit]
+context = 1_048_576

From 13241cf5debebc7c27367bb1a0ced4a9cf494374 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:23:48 +0000
Subject: [PATCH 0656/2012] chore(sync): update Kilo model catalog (#6318)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-6-astra.toml | 19 +------------------
 1 file changed, 1 insertion(+), 18 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-6-astra.toml b/providers/kilo/models/openai/gpt-6-astra.toml
index 50757636ec0..3cd84fe9c9f 100644
--- a/providers/kilo/models/openai/gpt-6-astra.toml
+++ b/providers/kilo/models/openai/gpt-6-astra.toml
@@ -1,14 +1,5 @@
-name = "OpenAI: GPT-6 Astra ($$$$)"
+base_model = "openai/gpt-6-astra"
 description = "GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon..."
-family = "gpt"
-release_date = "2026-09-04"
-last_updated = "2026-09-04"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,11 +10,3 @@ input = 10
 output = 50
 cache_read = 1
 cache_write = 12.5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]

From 1fd7ddfd2a6cf0f3e4ddfdb3daf025dea6d9e219 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:23:53 +0000
Subject: [PATCH 0657/2012] chore(sync): update NanoGPT model catalog (#6314)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ibm-granite/granite-4.1-8b.toml    | 25 ----------------
 .../models/openai/gpt-6-astra-pro.toml        | 29 +++++++++++++++++++
 .../nano-gpt/models/openai/gpt-6-astra.toml   | 18 ++----------
 3 files changed, 31 insertions(+), 41 deletions(-)
 delete mode 100644 providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-astra-pro.toml

diff --git a/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml b/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
deleted file mode 100644
index c17c8b75739..00000000000
--- a/providers/nano-gpt/models/ibm-granite/granite-4.1-8b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Included in subscription
-name = "Granite 4.1 8B"
-description = "Tool-capable chat model for instruction following and agentic application workflows"
-family = "granite"
-release_date = "2026-04-29"
-last_updated = "2026-04-29"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.05
-output = 0.1
-cache_read = 0.05
-
-[limit]
-context = 131_072
-input = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-6-astra-pro.toml b/providers/nano-gpt/models/openai/gpt-6-astra-pro.toml
new file mode 100644
index 00000000000..7d7ab129d1a
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-astra-pro.toml
@@ -0,0 +1,29 @@
+name = "GPT 6 Astra Pro"
+description = "GPT 6 Astra in Pro reasoning mode. Uses additional model work for difficult tasks, with higher latency and token usage at the same per-token rates. Reasoning effort remains independently configurable."
+family = "gpt"
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-6-astra.toml b/providers/nano-gpt/models/openai/gpt-6-astra.toml
index 08be47088bb..6d87b91b289 100644
--- a/providers/nano-gpt/models/openai/gpt-6-astra.toml
+++ b/providers/nano-gpt/models/openai/gpt-6-astra.toml
@@ -1,17 +1,9 @@
+base_model = "openai/gpt-6-astra"
 name = "GPT 6 Astra"
-description = "OpenAI's most capable model for the hardest end-to-end work, with state-of-the-art performance in computer use, browsing, software engineering, science, and professional tasks. Astra is designed to carry long, multistep workflows across code, browsers, and professional software through to a finished result."
-family = "gpt"
-release_date = "2026-09-03"
-last_updated = "2026-09-03"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 10
@@ -20,10 +12,4 @@ cache_read = 1
 cache_write = 12.5
 
 [limit]
-context = 1_050_000
 input = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]

From c6a981d7b063a4ec20d613a095e6020784b41121 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:24:02 +0000
Subject: [PATCH 0658/2012] chore(sync): update Venice model catalog (#6317)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/openai-gpt-6-astra.toml | 14 +-------------
 1 file changed, 1 insertion(+), 13 deletions(-)

diff --git a/providers/venice/models/openai-gpt-6-astra.toml b/providers/venice/models/openai-gpt-6-astra.toml
index f9ff45201ef..a063cf28964 100644
--- a/providers/venice/models/openai-gpt-6-astra.toml
+++ b/providers/venice/models/openai-gpt-6-astra.toml
@@ -1,13 +1,6 @@
-name = "GPT-6 Astra"
+base_model = "openai/gpt-6-astra"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt"
-release_date = "2026-09-04"
-last_updated = "2026-09-04"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -26,10 +19,5 @@ output = 93.75
 cache_read = 2.5
 cache_write = 31.25
 
-[limit]
-context = 1_050_000
-output = 128_000
-
 [modalities]
 input = ["text", "image"]
-output = ["text"]

From 12f7b0baff8fb098119a183543bd86b13357576c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:24:05 +0000
Subject: [PATCH 0659/2012] chore(sync): update xAI model catalog (#6313)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/xai/models/grok-imagine-image-2.0.toml     | 3 +++
 providers/xai/models/grok-imagine-image-quality.toml | 2 +-
 providers/xai/models/grok-imagine-image.toml         | 2 +-
 3 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/providers/xai/models/grok-imagine-image-2.0.toml b/providers/xai/models/grok-imagine-image-2.0.toml
index ca1c58ac153..0f00ef4f383 100644
--- a/providers/xai/models/grok-imagine-image-2.0.toml
+++ b/providers/xai/models/grok-imagine-image-2.0.toml
@@ -5,6 +5,9 @@
 # Pricing: $0.04 per image (not token-based; no [cost] authored, same as sibling imagine image models)
 base_model = "xai/grok-imagine-image-2.0"
 
+[limit]
+context = 64_000
+
 [modalities]
 input = ["text", "image", "pdf"]
 output = ["image", "pdf"]
diff --git a/providers/xai/models/grok-imagine-image-quality.toml b/providers/xai/models/grok-imagine-image-quality.toml
index 5781dc2839b..12faca1152a 100644
--- a/providers/xai/models/grok-imagine-image-quality.toml
+++ b/providers/xai/models/grok-imagine-image-quality.toml
@@ -10,7 +10,7 @@ tool_call = false
 open_weights = false
 
 [limit]
-context = 8_000
+context = 16_000
 output = 0
 
 [modalities]
diff --git a/providers/xai/models/grok-imagine-image.toml b/providers/xai/models/grok-imagine-image.toml
index 1a5fe140ad0..cd78dffb344 100644
--- a/providers/xai/models/grok-imagine-image.toml
+++ b/providers/xai/models/grok-imagine-image.toml
@@ -10,7 +10,7 @@ tool_call = false
 open_weights = false
 
 [limit]
-context = 8_000
+context = 16_000
 output = 0
 
 [modalities]

From b19f3c454ca117b906ec56c41f6bae9b00a44696 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:24:17 +0000
Subject: [PATCH 0660/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6319)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gemini-3.8-flash.toml | 2 +-
 providers/llmgateway/models/glm-5.3-flash.toml    | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/gemini-3.8-flash.toml b/providers/llmgateway/models/gemini-3.8-flash.toml
index 8eee5addfa4..cac3cbe0aaf 100644
--- a/providers/llmgateway/models/gemini-3.8-flash.toml
+++ b/providers/llmgateway/models/gemini-3.8-flash.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["minimal", "low", "medium", "high"]
+values = ["low", "medium", "high"]
 
 [cost]
 input = 0.75
diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index 035cb3a9470..b27a2911357 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.1
+input = 0.075
 output = 0.25
-cache_read = 0.02
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 4a9f3c217266f77ccbd40948235be79fffcd31cb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 22:24:29 +0000
Subject: [PATCH 0661/2012] chore(sync): update OpenRouter model catalog
 (#6320)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +++---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 +++---
 .../openrouter/models/openai/gpt-6-astra.toml | 19 +------------------
 3 files changed, 7 insertions(+), 24 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 45d3a54eefe..81ffd0604e0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08778
-output = 0.17556
-cache_read = 0.017556
+input = 0.0875
+output = 0.175
+cache_read = 0.0175
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index edc88da0141..f96b57ddc6f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.981708
-output = 1.963416
-cache_read = 0.081809
+input = 0.971442
+output = 1.942884
+cache_read = 0.080954
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-6-astra.toml b/providers/openrouter/models/openai/gpt-6-astra.toml
index 22bb9ba361f..9342c064d10 100644
--- a/providers/openrouter/models/openai/gpt-6-astra.toml
+++ b/providers/openrouter/models/openai/gpt-6-astra.toml
@@ -1,14 +1,5 @@
-name = "GPT-6 Astra"
+base_model = "openai/gpt-6-astra"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-09-04"
-last_updated = "2026-09-04"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -26,11 +17,3 @@ input = 20
 output = 75
 cache_read = 2
 cache_write = 25
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]

From fe5b09c6ea3ffd4f82e72b3d4bb6a8b670688b56 Mon Sep 17 00:00:00 2001
From: etonlels <etonlels@gmail.com>
Date: Fri, 4 Sep 2026 17:05:11 -0600
Subject: [PATCH 0662/2012] feat(openai): add `gpt-6-astra` provider entry
 (#6321)

Co-authored-by: OpenCode openai/gpt-6-astra <noreply@opencode.ai>
---
 providers/openai/models/gpt-6-astra.toml | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/openai/models/gpt-6-astra.toml

diff --git a/providers/openai/models/gpt-6-astra.toml b/providers/openai/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..4d59049f8a5
--- /dev/null
+++ b/providers/openai/models/gpt-6-astra.toml
@@ -0,0 +1,24 @@
+# Pricing: https://developers.openai.com/api/docs/models/gpt-6-astra
+# Reasoning and modes: https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-astra
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.00
+output = 50.00
+cache_read = 1.00
+cache_write = 12.50
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 20.00
+output = 75.00
+cache_read = 2.00
+cache_write = 25.00
+
+[experimental.modes.fast]
+cost = { input = 20.00, output = 100.00, cache_read = 2.00, cache_write = 25.00 }
+provider = { body = { service_tier = "priority" } }
+
+[experimental.modes.pro]
+provider = { body = { reasoning = { mode = "pro" } } }

From 21612f19b29fa920980fc24d59cf57acb4d225f8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 18:14:57 -0500
Subject: [PATCH 0663/2012] chore(sync): update Weights & Biases model catalog
 (#6301)

* chore(sync): update Weights & Biases model catalog

* fix(wandb): add GLM-5.3-Flash reasoning effort options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../wandb/models/zai-org/GLM-5.3-Flash.toml   | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/wandb/models/zai-org/GLM-5.3-Flash.toml

diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..778444a49a9
--- /dev/null
+++ b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,19 @@
+# W&B lists always-on reasoning with low/high/max effort; no on/off toggle.
+# https://trace.wandb.ai/inference/catalog/models
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
+open_weights = true
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]

From 86be6c9c5b4331a2a3119ed7f9396c57aaca766c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 23:22:56 +0000
Subject: [PATCH 0664/2012] chore(sync): update OpenRouter model catalog
 (#6323)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/qwen/qwen3-14b.toml             | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 81ffd0604e0..b0cbab2ec04 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0875
-output = 0.175
-cache_read = 0.0175
+input = 0.08722
+output = 0.17444
+cache_read = 0.017444
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index b10030292e3..b6ce24c73e0 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -17,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.12
-output = 0.24
+input = 0.2275
+output = 0.91
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
 
 [modalities]
 input = ["text"]

From 4a1ddefbe3b7275723e73a6e6bcb98eb43535782 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 23:22:58 +0000
Subject: [PATCH 0665/2012] chore(sync): update Kilo model catalog (#6324)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-14b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 2ce8bc46042..67828967f41 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 40_960
-output = 16_384
+context = 131_072
+output = 8_192
 
 [modalities]
 input = ["text"]

From fe421d95d66db4926e6e76f8dd7a22ee4f007d63 Mon Sep 17 00:00:00 2001
From: Karl Sabo <karl.sabo@gmail.com>
Date: Fri, 4 Sep 2026 17:24:38 -0600
Subject: [PATCH 0666/2012] feat(fireworks): add Qwen3.8 2.4T A95B (#6290)

* feat(fireworks): add Qwen3.8 Max

* Address reviewer feedback in Fireworks model
---
 .../fireworks/models/qwen3p8-2p4t-a95b.toml   | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-2p4t-a95b.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-2p4t-a95b.toml b/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-2p4t-a95b.toml
new file mode 100644
index 00000000000..c026897dfd3
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-2p4t-a95b.toml
@@ -0,0 +1,21 @@
+# Fireworks' page retains a qwen3p8-max URL slug, but identifies the served model as
+# Qwen3.8-2.4T-A95B with API path accounts/fireworks/models/qwen3p8-2p4t-a95b;
+# it also documents serverless pricing, 262K context, and text-only input:
+# https://fireworks.ai/models/fireworks/qwen3p8-max (accessed 2026-09-04)
+# The model requires thinking and accepts reasoning_effort low|medium|xhigh:
+# https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B/raw/main/README.md (accessed 2026-09-04)
+# Fireworks returns reasoning in reasoning_content:
+# https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-09-04)
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.25

From 0890e0faf028e92593efa71e402f055d9c7b194e Mon Sep 17 00:00:00 2001
From: Essam Almansouri <me@ealmansouri.com>
Date: Fri, 4 Sep 2026 16:25:47 -0700
Subject: [PATCH 0667/2012] feat(github-copilot): add `gpt-6-astra` provider
 entry (#6325)

Co-authored-by: Essam Almansouri <me@eralmansouri.com>
---
 .../github-copilot/models/gpt-6-astra.toml    | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/github-copilot/models/gpt-6-astra.toml

diff --git a/providers/github-copilot/models/gpt-6-astra.toml b/providers/github-copilot/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..73551409384
--- /dev/null
+++ b/providers/github-copilot/models/gpt-6-astra.toml
@@ -0,0 +1,20 @@
+# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# GPT-6 Astra rates: ≤272K $10/$1/$12.50/$50; >272K $20/$2/$25/$75 per 1M (input/cached/cache write/output)
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25

From b8d041c93422248311086d4f55289c92af383450 Mon Sep 17 00:00:00 2001
From: Tony Yu <49832190+ysntony@users.noreply.github.com>
Date: Sat, 5 Sep 2026 07:27:03 +0800
Subject: [PATCH 0668/2012] fix(kimi-for-coding): inherit attachment support
 for K3 models (#6102)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Both k3 and k3-256k inherit from the moonshotai/kimi-k3 lab base, which already declares attachment = true, so per the repo's override-only rule the correct fix is to drop the stale attachment = false override rather than flip it.
On the Kimi For Coding endpoint both models accept image input (k3 also accepts video), per Kimi Code's own subscription model catalog: k3 lists image_in and video_in, k3-256k lists image_in.
The declared [modalities] already match this (k3 inherits text/image/video from the base; k3-256k declares text/image), so the inherited flag is now consistent with modalities — same class of fix as #5793.
---
 providers/kimi-for-coding/models/k3-256k.toml | 1 -
 providers/kimi-for-coding/models/k3.toml      | 1 -
 2 files changed, 2 deletions(-)

diff --git a/providers/kimi-for-coding/models/k3-256k.toml b/providers/kimi-for-coding/models/k3-256k.toml
index d28baff592d..2e7cd5d2d1f 100644
--- a/providers/kimi-for-coding/models/k3-256k.toml
+++ b/providers/kimi-for-coding/models/k3-256k.toml
@@ -6,7 +6,6 @@
 base_model = "moonshotai/kimi-k3"
 name = "Kimi K3-256K"
 description = "256K-context version of Kimi K3, reducing token consumption for shorter coding sessions"
-attachment = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kimi-for-coding/models/k3.toml b/providers/kimi-for-coding/models/k3.toml
index 3b31cf148db..dad0dadecd8 100644
--- a/providers/kimi-for-coding/models/k3.toml
+++ b/providers/kimi-for-coding/models/k3.toml
@@ -3,7 +3,6 @@
 #   output_config.effort = "low" | "high" | "max"
 # Cost is zeroed per this provider's subscription convention.
 base_model = "moonshotai/kimi-k3"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"

From 43f7f23b9060e5ce12e101392cd3ad9d2ec9d189 Mon Sep 17 00:00:00 2001
From: Emmanuel Acheampong <achampion.emma@gmail.com>
Date: Fri, 4 Sep 2026 16:27:15 -0700
Subject: [PATCH 0669/2012] feat(crusoe): add Qwen3-235B-A22B-Instruct-2507 and
 Nemotron 3 Nano/Super (#6267)

Follow-up to #3769: adds three models that are in Crusoe's live catalog
and on the public pricing page but were dropped from the original PR
pending verification. Model ids match the live /v1/models listing
(NVIDIA- prefix on the Nemotron pair, Qwen org casing). Pricing from
https://www.crusoe.ai/cloud/pricing (accessed 2026-09-04).
---
 .../models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml |  7 +++++++
 .../nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml     | 14 ++++++++++++++
 .../nvidia/NVIDIA-Nemotron-3-Super-120B-A12B.toml  | 14 ++++++++++++++
 3 files changed, 35 insertions(+)
 create mode 100644 providers/crusoe/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml
 create mode 100644 providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
 create mode 100644 providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B.toml

diff --git a/providers/crusoe/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml b/providers/crusoe/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml
new file mode 100644
index 00000000000..c92a72b8f0e
--- /dev/null
+++ b/providers/crusoe/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml
@@ -0,0 +1,7 @@
+base_model = "alibaba/qwen3-235b-a22b-instruct-2507"
+
+# Pricing: https://www.crusoe.ai/cloud/pricing (accessed 2026-09-04)
+[cost]
+input = 0.22
+output = 0.80
+cache_read = 0.11
diff --git a/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml b/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
new file mode 100644
index 00000000000..5174564f492
--- /dev/null
+++ b/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B.toml
@@ -0,0 +1,14 @@
+# Not yet live-probed on this endpoint (unlike the entries from #3769 —
+# no API access at submission time). NVIDIA Nemotron 3 models are
+# hybrid-reasoning with an on/off toggle, and the probed sibling on this
+# endpoint (Nemotron-3-Nano-Omni-Reasoning-30B-A3B) behaves as a toggle
+# with inert low/medium/high values, so this is encoded the same way.
+# No interleaved claim is made.
+base_model = "nvidia/nemotron-3-nano-30b-a3b"
+reasoning_options = [{ type = "toggle" }]
+
+# Pricing: https://www.crusoe.ai/cloud/pricing (accessed 2026-09-04)
+[cost]
+input = 0.05
+output = 0.20
+cache_read = 0.03
diff --git a/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B.toml b/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B.toml
new file mode 100644
index 00000000000..c5b8212623f
--- /dev/null
+++ b/providers/crusoe/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B.toml
@@ -0,0 +1,14 @@
+# Not yet live-probed on this endpoint (unlike the entries from #3769 —
+# no API access at submission time). NVIDIA Nemotron 3 models are
+# hybrid-reasoning with an on/off toggle, and the probed sibling on this
+# endpoint (Nemotron-3-Nano-Omni-Reasoning-30B-A3B) behaves as a toggle
+# with inert low/medium/high values, so this is encoded the same way.
+# No interleaved claim is made.
+base_model = "nvidia/nemotron-3-super-120b-a12b"
+reasoning_options = [{ type = "toggle" }]
+
+# Pricing: https://www.crusoe.ai/cloud/pricing (accessed 2026-09-04)
+[cost]
+input = 0.30
+output = 2.40
+cache_read = 0.15

From ef71a4865a5821d8cd24f171a0100156a02e808e Mon Sep 17 00:00:00 2001
From: Abdul Munim <iam@munim.net>
Date: Sat, 5 Sep 2026 01:27:31 +0200
Subject: [PATCH 0670/2012] feat(cline-pass): add glm-5.3-flash (#6053)

* feat(cline-pass): add glm-5.3-flash

glm-5.3-flash is available in ClinePass but missing from the provider
data. Adds providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
with base_model pointing at the existing zhipuai lab entry and
reference pricing matching glm-5.3.

* fix(cline-pass): correct glm-5.3-flash pricing

$0.075 input / $0.25 output per 1M tokens; cache read $0.015.

* fix(cline-pass): glm-5.3-flash actual pricing

$0.15 input / $0.50 output / $0.03 cache read per 1M tokens.

* fix(cline-pass): inline glm-5.3-flash entry

No upstream lab mapping on ClinePass for this slug. Inline the full
model metadata (name/description from the ClinePass catalog, specs
matching Z.ai's GLM-5.3-Flash) instead of base_model.
---
 .../models/cline-pass/glm-5.3-flash.toml      | 36 +++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 providers/cline-pass/models/cline-pass/glm-5.3-flash.toml

diff --git a/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml b/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
new file mode 100644
index 00000000000..7f9e9f9ec7d
--- /dev/null
+++ b/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
@@ -0,0 +1,36 @@
+# https://docs.cline.bot/getting-started/clinepass (accessed 2026-09-02)
+# Model ID cline-pass/glm-5.3-flash; reference pricing $0.15 / $0.50 / $0.03 per 1M tokens (input/output/cache read).
+# Name/description from the ClinePass model catalog. ClinePass does not declare an
+# upstream lab mapping for this entry; specs (limits, modalities, reasoning) match
+# Z.ai's GLM-5.3-Flash (https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash).
+name = "cline-pass/glm-5.3-flash"
+description = "Latest natively multimodal model in the GLM-5 series"
+family = "glm"
+release_date = "2026-08-26"
+last_updated = "2026-08-26"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03

From 9e59515cbd50acbdd7304c60d6f9eb0deb5a9717 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 00:57:23 +0000
Subject: [PATCH 0671/2012] chore(sync): update Kilo model catalog (#6327)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-14b.toml            | 4 ++--
 providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml | 2 +-
 providers/kilo/models/tencent/hy3.toml               | 6 +++---
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 67828967f41..2ce8bc46042 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 131_072
-output = 8_192
+context = 40_960
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
index b356342358d..3f37ded3629 100644
--- a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
@@ -16,7 +16,7 @@ output = 0.598
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144

From beb06bd3ed68b7635f44186926b02236330d0caf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 00:57:41 +0000
Subject: [PATCH 0672/2012] chore(sync): update OpenRouter model catalog
 (#6326)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 providers/openrouter/models/qwen/qwen3-14b.toml            | 6 +++---
 providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml | 7 +++----
 providers/openrouter/models/tencent/hy3.toml               | 6 +++---
 5 files changed, 15 insertions(+), 16 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..09d98bc8130 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 1.12068
+output = 3.36204
+cache_read = 0.037356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f96b57ddc6f..edc88da0141 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.971442
-output = 1.942884
-cache_read = 0.080954
+input = 0.981708
+output = 1.963416
+cache_read = 0.081809
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index b6ce24c73e0..b10030292e3 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -17,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.2275
-output = 0.91
+input = 0.12
+output = 0.24
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index 0d691163001..aa706cc488b 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,13 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.0875
-output = 0.35
-cache_read = 0.0175
+input = 0.09
+output = 0.55
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 9afd6fab3ce3f492327caa4fc80ce09b71fecdd3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 01:31:40 +0000
Subject: [PATCH 0673/2012] chore(sync): update Venice model catalog (#6331)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/openai-gpt-6-astra.toml | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/providers/venice/models/openai-gpt-6-astra.toml b/providers/venice/models/openai-gpt-6-astra.toml
index a063cf28964..3ed7e7cfb48 100644
--- a/providers/venice/models/openai-gpt-6-astra.toml
+++ b/providers/venice/models/openai-gpt-6-astra.toml
@@ -1,23 +1,24 @@
 base_model = "openai/gpt-6-astra"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt"
+release_date = "2026-09-05"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 12.5
-output = 62.5
-cache_read = 1.25
-cache_write = 15.625
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 25
-output = 93.75
-cache_read = 2.5
-cache_write = 31.25
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
 
 [modalities]
 input = ["text", "image"]

From 5b72c95b6de93b19700eee5a836aaa80f7a8dd96 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 01:31:47 +0000
Subject: [PATCH 0674/2012] chore(sync): update OpenRouter model catalog
 (#6330)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index edc88da0141..f96b57ddc6f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.981708
-output = 1.963416
-cache_read = 0.081809
+input = 0.971442
+output = 1.942884
+cache_read = 0.080954
 
 [limit]
 context = 1_048_576

From 9e3bc69f2cadf8fcb8fd9d4fdc75144068ef8d42 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Fri, 4 Sep 2026 21:57:48 -0400
Subject: [PATCH 0675/2012] feat(opencode): add Zen models

---
 .../models/deepseek-v4-flash-vision-exp.toml   | 12 ++++++++++++
 providers/opencode/models/glm-5.3-flash.toml   | 10 ++++++++++
 providers/opencode/models/glm-5.3.toml         | 10 ++++++++++
 providers/opencode/models/gpt-6-astra.toml     | 18 ++++++++++++++++++
 providers/opencode/models/muse-spark-1.3.toml  | 10 ++++++++++
 5 files changed, 60 insertions(+)
 create mode 100644 providers/opencode/models/deepseek-v4-flash-vision-exp.toml
 create mode 100644 providers/opencode/models/glm-5.3-flash.toml
 create mode 100644 providers/opencode/models/glm-5.3.toml
 create mode 100644 providers/opencode/models/gpt-6-astra.toml
 create mode 100644 providers/opencode/models/muse-spark-1.3.toml

diff --git a/providers/opencode/models/deepseek-v4-flash-vision-exp.toml b/providers/opencode/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..f6858d10c75
--- /dev/null
+++ b/providers/opencode/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.028
diff --git a/providers/opencode/models/glm-5.3-flash.toml b/providers/opencode/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..828e40e41b9
--- /dev/null
+++ b/providers/opencode/models/glm-5.3-flash.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
diff --git a/providers/opencode/models/glm-5.3.toml b/providers/opencode/models/glm-5.3.toml
new file mode 100644
index 00000000000..a61a9939174
--- /dev/null
+++ b/providers/opencode/models/glm-5.3.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/opencode/models/gpt-6-astra.toml b/providers/opencode/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..af9ab833561
--- /dev/null
+++ b/providers/opencode/models/gpt-6-astra.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[provider]
+npm = "@ai-sdk/openai"
diff --git a/providers/opencode/models/muse-spark-1.3.toml b/providers/opencode/models/muse-spark-1.3.toml
new file mode 100644
index 00000000000..cfaa3f534d3
--- /dev/null
+++ b/providers/opencode/models/muse-spark-1.3.toml
@@ -0,0 +1,10 @@
+base_model = "meta/muse-spark-1.3"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[provider]
+npm = "@ai-sdk/openai"

From 6e5efcd056370b0853db07ce9b4e02391c8a2d55 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Fri, 4 Sep 2026 22:04:24 -0400
Subject: [PATCH 0676/2012] update zen models

---
 providers/opencode/models/gpt-6-astra.toml | 18 ------------------
 1 file changed, 18 deletions(-)
 delete mode 100644 providers/opencode/models/gpt-6-astra.toml

diff --git a/providers/opencode/models/gpt-6-astra.toml b/providers/opencode/models/gpt-6-astra.toml
deleted file mode 100644
index af9ab833561..00000000000
--- a/providers/opencode/models/gpt-6-astra.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "openai/gpt-6-astra"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
-input = 20
-output = 75
-cache_read = 2
-cache_write = 25
-
-[provider]
-npm = "@ai-sdk/openai"

From 12c9088ada76f509452850640815a36033c4e9f8 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Fri, 4 Sep 2026 22:48:42 -0400
Subject: [PATCH 0677/2012] Revert "update zen models"

This reverts commit 6e5efcd056370b0853db07ce9b4e02391c8a2d55.
---
 providers/opencode/models/gpt-6-astra.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/opencode/models/gpt-6-astra.toml

diff --git a/providers/opencode/models/gpt-6-astra.toml b/providers/opencode/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..af9ab833561
--- /dev/null
+++ b/providers/opencode/models/gpt-6-astra.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[provider]
+npm = "@ai-sdk/openai"

From 8771965aef29fd598d48b86d08411d24e5ce90a6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 4 Sep 2026 21:54:48 -0500
Subject: [PATCH 0678/2012] chore(sync): update Weights & Biases model catalog
 (#6322)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/zai-org/GLM-5.3-Flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
index 778444a49a9..c3b6661bb43 100644
--- a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
+++ b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
@@ -4,7 +4,7 @@ base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
 open_weights = true
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+reasoning_options = []
 
 [cost]
 input = 0.15

From e736443eaae86bc4d0d5ed0a160e18adbf61ebf0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 03:28:58 +0000
Subject: [PATCH 0679/2012] chore(sync): update OpenRouter model catalog
 (#6333)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b0cbab2ec04..8414db82b51 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08722
-output = 0.17444
-cache_read = 0.017444
+input = 0.08694
+output = 0.17388
+cache_read = 0.017388
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f96b57ddc6f..544ad412ba2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.971442
-output = 1.942884
-cache_read = 0.080954
+input = 0.96135
+output = 1.9227
+cache_read = 0.080113
 
 [limit]
 context = 1_048_576

From f1ffd4bb59b31af3c1f7182541ba1e3d1793174a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 04:28:10 +0000
Subject: [PATCH 0680/2012] chore(sync): update OpenRouter model catalog
 (#6334)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 8414db82b51..2ac691bbe96 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08694
-output = 0.17388
-cache_read = 0.017388
+input = 0.08666
+output = 0.17332
+cache_read = 0.017332
 
 [limit]
 context = 1_048_576

From 1abf8914cbc0d715a91e393f25842d959c0eece6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 04:28:16 +0000
Subject: [PATCH 0681/2012] chore(sync): update Venice model catalog (#6335)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-6-astra-pro.toml | 35 +++++++++++++++++++
 1 file changed, 35 insertions(+)
 create mode 100644 providers/venice/models/openai-gpt-6-astra-pro.toml

diff --git a/providers/venice/models/openai-gpt-6-astra-pro.toml b/providers/venice/models/openai-gpt-6-astra-pro.toml
new file mode 100644
index 00000000000..8d380a44b1d
--- /dev/null
+++ b/providers/venice/models/openai-gpt-6-astra-pro.toml
@@ -0,0 +1,35 @@
+name = "GPT-6 Astra Pro"
+description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
+family = "gpt"
+release_date = "2026-09-05"
+last_updated = "2026-09-05"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 12.5
+output = 62.5
+cache_read = 1.25
+cache_write = 15.625
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 25
+output = 93.75
+cache_read = 2.5
+cache_write = 31.25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 968b0b8e321e9b8dce8a6dd226fccd722e0e302c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 04:28:21 +0000
Subject: [PATCH 0682/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6287)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/openai/gpt-5.6-luna-fast.toml  | 2 +-
 providers/vercel/models/openai/gpt-5.6-sol-fast.toml   | 2 +-
 providers/vercel/models/openai/gpt-5.6-terra-fast.toml | 2 +-
 providers/vercel/models/openai/gpt-6-astra-fast.toml   | 2 +-
 4 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/vercel/models/openai/gpt-5.6-luna-fast.toml b/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
index acca99f77f7..0677db18d76 100644
--- a/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
@@ -9,4 +9,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 0.4
 output = 2.4
 cache_read = 0.04
-cache_write = 0.25
+cache_write = 0.5
diff --git a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
index 93628796761..adf0da2fcb0 100644
--- a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
@@ -9,4 +9,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 4
 output = 20
 cache_read = 0.4
-cache_write = 2.5
+cache_write = 5
diff --git a/providers/vercel/models/openai/gpt-5.6-terra-fast.toml b/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
index e8749237b36..c5e03cad536 100644
--- a/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
@@ -9,4 +9,4 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 4
 output = 24
 cache_read = 0.4
-cache_write = 2.5
+cache_write = 5
diff --git a/providers/vercel/models/openai/gpt-6-astra-fast.toml b/providers/vercel/models/openai/gpt-6-astra-fast.toml
index 0ab38d43b74..ad67cd2908a 100644
--- a/providers/vercel/models/openai/gpt-6-astra-fast.toml
+++ b/providers/vercel/models/openai/gpt-6-astra-fast.toml
@@ -9,7 +9,7 @@ values = ["none", "low", "medium", "high", "xhigh", "max"]
 input = 20
 output = 100
 cache_read = 2
-cache_write = 12.5
+cache_write = 25
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_001 }

From 333d17ec00dcc4e660693d304e2e8a34e6d839da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 05:25:29 +0000
Subject: [PATCH 0683/2012] chore(sync): update Kilo model catalog (#6336)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-max-0902.toml | 16 ++++++++++++++++
 providers/kilo/models/qwen/qwen3.8-max.toml      | 16 ----------------
 2 files changed, 16 insertions(+), 16 deletions(-)
 create mode 100644 providers/kilo/models/qwen/qwen3.8-max-0902.toml
 delete mode 100644 providers/kilo/models/qwen/qwen3.8-max.toml

diff --git a/providers/kilo/models/qwen/qwen3.8-max-0902.toml b/providers/kilo/models/qwen/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..cb4bc8d0cb3
--- /dev/null
+++ b/providers/kilo/models/qwen/qwen3.8-max-0902.toml
@@ -0,0 +1,16 @@
+base_model = "alibaba/qwen3.8-max-0902"
+description = "Qwen3.8 Max 0902 is an updated snapshot of Qwen3.8 Max from Alibaba's Qwen team. It is a 2.4-trillion-parameter mixture-of-experts model that accepts text, image, and video input and returns text,..."
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+cache_write = 2.5
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.8-max.toml b/providers/kilo/models/qwen/qwen3.8-max.toml
deleted file mode 100644
index 169746d865f..00000000000
--- a/providers/kilo/models/qwen/qwen3.8-max.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "alibaba/qwen3.8-max"
-description = "Qwen3.8 Max is the flagship model in Alibaba's Qwen3.8 series, the general-availability successor to the Qwen3.8 Max Preview. It is a multimodal reasoning model intended for complex reasoning, visual understanding,..."
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
-
-[cost]
-input = 2
-output = 6
-cache_read = 0.25
-cache_write = 2.5
-
-[modalities]
-input = ["text", "image", "video"]

From 765a0f3fda3854e100507cef8cd6ee59a3207616 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 05:26:51 +0000
Subject: [PATCH 0684/2012] chore(sync): update Eden AI model catalog (#6339)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 1f11f70d4ba..d5d8773c4b6 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 7e31b605e19..05bf93bee1e 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From a7c0017da35bdc8ebb5cf155e250b10553abce2a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 05:26:57 +0000
Subject: [PATCH 0685/2012] chore(sync): update OpenRouter model catalog
 (#6338)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../models/qwen/{qwen3.8-max.toml => qwen3.8-max-0902.toml} | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)
 rename providers/openrouter/models/qwen/{qwen3.8-max.toml => qwen3.8-max-0902.toml} (89%)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 2ac691bbe96..6c62b40176e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08666
-output = 0.17332
-cache_read = 0.017332
+input = 0.08638
+output = 0.17276
+cache_read = 0.017276
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 544ad412ba2..5aba42455d2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.96135
-output = 1.9227
-cache_read = 0.080113
+input = 0.951258
+output = 1.902516
+cache_read = 0.079272
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-max.toml b/providers/openrouter/models/qwen/qwen3.8-max-0902.toml
similarity index 89%
rename from providers/openrouter/models/qwen/qwen3.8-max.toml
rename to providers/openrouter/models/qwen/qwen3.8-max-0902.toml
index d86d862997e..78275205850 100644
--- a/providers/openrouter/models/qwen/qwen3.8-max.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-max-0902.toml
@@ -1,4 +1,4 @@
-base_model = "alibaba/qwen3.8-max"
+base_model = "alibaba/qwen3.8-max-0902"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 structured_output = true
 

From 5943298ebcd043c1ecf1a9e6874a494fb3c176d9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 06:36:05 +0000
Subject: [PATCH 0686/2012] chore(sync): update LLM Gateway model catalog
 (#6342)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/azure/gpt-6-astra.toml                  | 12 ++++++++++++
 .../models/runware/kimi-k3.toml                    | 14 ++++++++++++++
 2 files changed, 26 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/azure/gpt-6-astra.toml
 create mode 100644 providers/llmgateway-providers/models/runware/kimi-k3.toml

diff --git a/providers/llmgateway-providers/models/azure/gpt-6-astra.toml b/providers/llmgateway-providers/models/azure/gpt-6-astra.toml
new file mode 100644
index 00000000000..d2ebe475720
--- /dev/null
+++ b/providers/llmgateway-providers/models/azure/gpt-6-astra.toml
@@ -0,0 +1,12 @@
+base_model = "openai/gpt-6-astra"
+name = "GPT-6 Astra (Azure)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
diff --git a/providers/llmgateway-providers/models/runware/kimi-k3.toml b/providers/llmgateway-providers/models/runware/kimi-k3.toml
new file mode 100644
index 00000000000..26d21545c02
--- /dev/null
+++ b/providers/llmgateway-providers/models/runware/kimi-k3.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (Runware)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+output = 1_048_576

From 72e5909a078bd54b56c2eafc461f40caa3540cdf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 06:36:14 +0000
Subject: [PATCH 0687/2012] chore(sync): update Eden AI model catalog (#6341)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/qwen/qwen3.8-max-0902.toml      | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/edenai/models/qwen/qwen3.8-max-0902.toml

diff --git a/providers/edenai/models/qwen/qwen3.8-max-0902.toml b/providers/edenai/models/qwen/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..4ecec330d98
--- /dev/null
+++ b/providers/edenai/models/qwen/qwen3.8-max-0902.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-max-0902"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+cache_write = 2.5
+
+[modalities]
+input = ["text", "image", "video"]

From 08b502fba2911de9d3e9e324d629264f781f7262 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 06:36:20 +0000
Subject: [PATCH 0688/2012] chore(sync): update OpenRouter model catalog
 (#6343)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6c62b40176e..83f020b67b5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08638
-output = 0.17276
-cache_read = 0.017276
+input = 0.0861
+output = 0.1722
+cache_read = 0.01722
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 5aba42455d2..8b309925acc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.951258
-output = 1.902516
-cache_read = 0.079272
+input = 0.941166
+output = 1.882332
+cache_read = 0.078431
 
 [limit]
 context = 1_048_576

From 3eed7cf03695901db6c50dc5c4be011086307f0f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 07:24:46 +0000
Subject: [PATCH 0689/2012] chore(sync): update OpenRouter model catalog
 (#6344)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 83f020b67b5..c278261398d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0861
-output = 0.1722
-cache_read = 0.01722
+input = 0.08582
+output = 0.17164
+cache_read = 0.017164
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 8b309925acc..9d07c19a143 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.941166
-output = 1.882332
-cache_read = 0.078431
+input = 0.9309
+output = 1.8618
+cache_read = 0.077575
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 052fb05abd3..70251750002 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 14
-cache_read = 0.29
+input = 2.55
+output = 12.75
+cache_read = 0.256
 
 [limit]
 context = 1_048_576

From f6baae501fb7e6e1715046160b170930fa7c3f0b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 07:24:55 +0000
Subject: [PATCH 0690/2012] chore(sync): update Kilo model catalog (#6345)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 6481e4e1993..de31079e2dd 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 14
-cache_read = 0.29
+input = 2.55
+output = 12.75
+cache_read = 0.256
 
 [limit]
 context = 1_048_576

From 26a9ab3b8fbdccd9d288ad577e18f028f4cae96c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 08:28:58 +0000
Subject: [PATCH 0691/2012] chore(sync): update NanoGPT model catalog (#6346)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/claude-opus-4-5-20251101.toml       | 1 -
 .../nano-gpt/models/claude-opus-4-5-20251101:thinking.toml    | 1 -
 .../nano-gpt/models/gemini-2.5-flash-lite-preview-06-17.toml  | 4 ++--
 .../gemini-2.5-flash-lite-preview-09-2025-thinking.toml       | 4 ++--
 .../models/gemini-2.5-flash-lite-preview-09-2025.toml         | 4 ++--
 providers/nano-gpt/models/gemini-2.5-flash-lite.toml          | 3 +--
 providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml    | 4 ++--
 providers/nano-gpt/models/gemini-2.5-flash-preview-04-17.toml | 4 ++--
 .../models/gemini-2.5-flash-preview-04-17:thinking.toml       | 4 ++--
 .../models/gemini-2.5-flash-preview-09-2025-thinking.toml     | 4 ++--
 .../nano-gpt/models/gemini-2.5-flash-preview-09-2025.toml     | 4 ++--
 providers/nano-gpt/models/gemini-2.5-flash.toml               | 3 +--
 providers/nano-gpt/models/gemini-2.5-pro-exp-03-25.toml       | 4 ++--
 providers/nano-gpt/models/gemini-2.5-pro-preview-03-25.toml   | 4 ++--
 providers/nano-gpt/models/gemini-2.5-pro-preview-05-06.toml   | 4 ++--
 providers/nano-gpt/models/gemini-2.5-pro-preview-06-05.toml   | 4 ++--
 providers/nano-gpt/models/gemini-2.5-pro.toml                 | 3 +--
 providers/nano-gpt/models/gemini-3-pro-image-preview.toml     | 4 +---
 .../models/google/gemini-3-flash-preview-thinking.toml        | 3 +--
 providers/nano-gpt/models/google/gemini-3-flash-preview.toml  | 3 +--
 .../models/google/gemini-3.1-pro-preview-customtools.toml     | 3 +--
 .../nano-gpt/models/google/gemini-3.1-pro-preview-high.toml   | 4 ++--
 .../nano-gpt/models/google/gemini-3.1-pro-preview-low.toml    | 4 ++--
 providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml  | 3 +--
 providers/nano-gpt/models/google/gemini-pro-latest.toml       | 4 ++--
 25 files changed, 38 insertions(+), 49 deletions(-)

diff --git a/providers/nano-gpt/models/claude-opus-4-5-20251101.toml b/providers/nano-gpt/models/claude-opus-4-5-20251101.toml
index c55048c722d..6e38409f610 100644
--- a/providers/nano-gpt/models/claude-opus-4-5-20251101.toml
+++ b/providers/nano-gpt/models/claude-opus-4-5-20251101.toml
@@ -13,4 +13,3 @@ cache_read = 0.5
 
 [limit]
 input = 200_000
-output = 32_000
diff --git a/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml b/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml
index 817feef1609..0cfb293c2f4 100644
--- a/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml
+++ b/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml
@@ -13,4 +13,3 @@ cache_read = 0.5
 
 [limit]
 input = 200_000
-output = 32_000
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-06-17.toml b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-06-17.toml
index b903f64e19a..6a1c528190a 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-06-17.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-06-17.toml
@@ -19,8 +19,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025-thinking.toml b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025-thinking.toml
index bd866af82db..d58e411eea9 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025-thinking.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025-thinking.toml
@@ -19,8 +19,8 @@ output = 0.4
 cache_read = 0.01
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025.toml b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025.toml
index c8324b3784c..8f3240de66f 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-lite-preview-09-2025.toml
@@ -19,8 +19,8 @@ output = 0.4
 cache_read = 0.01
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-lite.toml b/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
index 5d2f7caaff7..8537dcb2c30 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
@@ -12,8 +12,7 @@ output = 0.4
 cache_read = 0.01
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml b/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
index a2c247811b3..772e6034dae 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
@@ -15,8 +15,8 @@ output = 2.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17.toml b/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17.toml
index d13a6edc9e6..632f808bdc0 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17.toml
@@ -19,8 +19,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17:thinking.toml b/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17:thinking.toml
index 4d6bfe143d9..2efb99f0fc6 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17:thinking.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-preview-04-17:thinking.toml
@@ -19,8 +19,8 @@ output = 3.5
 cache_read = 0.015
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025-thinking.toml b/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025-thinking.toml
index 1f1c0a7ee42..e243098e6a5 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025-thinking.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025-thinking.toml
@@ -19,8 +19,8 @@ output = 2.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025.toml b/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025.toml
index b81e312b337..54927974f4e 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-preview-09-2025.toml
@@ -19,8 +19,8 @@ output = 2.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash.toml b/providers/nano-gpt/models/gemini-2.5-flash.toml
index 12a931a7eb4..8cf079ce92f 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash.toml
@@ -11,8 +11,7 @@ output = 2.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro-exp-03-25.toml b/providers/nano-gpt/models/gemini-2.5-pro-exp-03-25.toml
index f2f864700d1..20085a3efa2 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro-exp-03-25.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro-exp-03-25.toml
@@ -19,8 +19,8 @@ output = 10
 cache_read = 0.25
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro-preview-03-25.toml b/providers/nano-gpt/models/gemini-2.5-pro-preview-03-25.toml
index 0ea091e9591..b55fdddb797 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro-preview-03-25.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro-preview-03-25.toml
@@ -19,8 +19,8 @@ output = 10
 cache_read = 0.25
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro-preview-05-06.toml b/providers/nano-gpt/models/gemini-2.5-pro-preview-05-06.toml
index 1111e07a90c..49636161cbf 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro-preview-05-06.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro-preview-05-06.toml
@@ -19,8 +19,8 @@ output = 10
 cache_read = 0.25
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro-preview-06-05.toml b/providers/nano-gpt/models/gemini-2.5-pro-preview-06-05.toml
index 37cd71988b3..138dd1a9f94 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro-preview-06-05.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro-preview-06-05.toml
@@ -19,8 +19,8 @@ output = 10
 cache_read = 0.25
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro.toml b/providers/nano-gpt/models/gemini-2.5-pro.toml
index 63c7f8c7722..aac0d659349 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro.toml
@@ -12,8 +12,7 @@ cache_read = 0.125
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/gemini-3-pro-image-preview.toml b/providers/nano-gpt/models/gemini-3-pro-image-preview.toml
index 7258a07ee50..1ee700a2e54 100644
--- a/providers/nano-gpt/models/gemini-3-pro-image-preview.toml
+++ b/providers/nano-gpt/models/gemini-3-pro-image-preview.toml
@@ -9,9 +9,7 @@ output = 12
 cache_read = 0.2
 
 [limit]
-context = 1_048_756
-input = 1_048_756
-output = 65_536
+input = 65_536
 
 [modalities]
 output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml b/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
index c1b7f94a508..8e849ee5ac4 100644
--- a/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
+++ b/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
@@ -13,8 +13,7 @@ output = 3
 cache_read = 0.05
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3-flash-preview.toml b/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
index 5bb90311b02..dd42a4b85e1 100644
--- a/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
+++ b/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
@@ -11,8 +11,7 @@ output = 3
 cache_read = 0.05
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
index f50df40cada..ef98ef336f8 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
@@ -13,8 +13,7 @@ cache_read = 0.2
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
index 2e9c672892f..253bc7b4555 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
@@ -18,8 +18,8 @@ cache_read = 0.2
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-low.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-low.toml
index df7e4a728c3..b78d52075b6 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-low.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-low.toml
@@ -18,8 +18,8 @@ cache_read = 0.2
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
index dcc5c328aa0..264445e5b25 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
@@ -13,8 +13,7 @@ cache_read = 0.2
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+input = 1_048_576
 
 [modalities]
 input = ["text", "image", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-pro-latest.toml b/providers/nano-gpt/models/google/gemini-pro-latest.toml
index c7299c217fc..7f53052baf1 100644
--- a/providers/nano-gpt/models/google/gemini-pro-latest.toml
+++ b/providers/nano-gpt/models/google/gemini-pro-latest.toml
@@ -21,8 +21,8 @@ cache_read = 0.2
 cache_write = 0.375
 
 [limit]
-context = 1_048_756
-input = 1_048_756
+context = 1_048_576
+input = 1_048_576
 output = 65_536
 
 [modalities]

From a151b4bea50dab22b6fd5f991d19faeee63f446f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 08:29:01 +0000
Subject: [PATCH 0692/2012] chore(sync): update OpenRouter model catalog
 (#6347)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c278261398d..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08582
-output = 0.17164
-cache_read = 0.017164
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 9d07c19a143..50522fea9a3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.9309
-output = 1.8618
-cache_read = 0.077575
+input = 0.920808
+output = 1.841616
+cache_read = 0.076734
 
 [limit]
 context = 1_048_576

From 7a42caabb2c8397ca145f72d5a1d5c7a4031d593 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 09:24:48 +0000
Subject: [PATCH 0693/2012] chore(sync): update OpenRouter model catalog
 (#6348)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..e3961bdffdf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.08526
+output = 0.17052
+cache_read = 0.017052
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 50522fea9a3..838a06c3431 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.920808
-output = 1.841616
-cache_read = 0.076734
+input = 0.910716
+output = 1.821432
+cache_read = 0.075893
 
 [limit]
 context = 1_048_576

From 5d0ade66731ba63550d1fe659e689b1f3b538975 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 10:23:18 +0000
Subject: [PATCH 0694/2012] chore(sync): update OpenRouter model catalog
 (#6350)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e3961bdffdf..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08526
-output = 0.17052
-cache_read = 0.017052
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 838a06c3431..4ace763636d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.910716
-output = 1.821432
-cache_read = 0.075893
+input = 0.90045
+output = 1.8009
+cache_read = 0.075038
 
 [limit]
 context = 1_048_576

From 55c30ce9ca65938faf7be7d6da8f4a6bfa298c0c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 10:23:22 +0000
Subject: [PATCH 0695/2012] chore(sync): update LLM Gateway model catalog
 (#6349)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/deepseek-v4-flash.toml           | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
index e0bff4fb512..e28f16c5e0a 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.08
+input = 0.07
 output = 0.16
-cache_read = 0.02
+cache_read = 0.1
 
 [limit]
 context = 524_288

From 0f17620db1d9ec22adf4fc1517a877deea7b9a62 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 10:23:28 +0000
Subject: [PATCH 0696/2012] chore(sync): update Eden AI model catalog (#6351)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/openai/gpt-6-astra.toml      | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/edenai/models/openai/gpt-6-astra.toml

diff --git a/providers/edenai/models/openai/gpt-6-astra.toml b/providers/edenai/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..d3cae944f01
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-6-astra.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25

From 4816b02b6d4d4ff82c36a8b43d4e3b1e35ef84e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 11:21:28 +0000
Subject: [PATCH 0697/2012] chore(sync): update Eden AI model catalog (#6357)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/openai/gpt-latest.toml      | 20 +++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/edenai/models/openai/gpt-latest.toml b/providers/edenai/models/openai/gpt-latest.toml
index c32a0af9b15..d3cae944f01 100644
--- a/providers/edenai/models/openai/gpt-latest.toml
+++ b/providers/edenai/models/openai/gpt-latest.toml
@@ -1,18 +1,18 @@
-base_model = "openai/gpt-5.6-sol"
+base_model = "openai/gpt-6-astra"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 4
-output = 20
-cache_read = 0.4
-cache_write = 5
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 8
-output = 30
-cache_read = 0.8
-cache_write = 10
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25

From dca7479a94196a010527aaa5fa89ae093a57eb4d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 11:21:30 +0000
Subject: [PATCH 0698/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6355)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3-flash.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index b27a2911357..035cb3a9470 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.075
+input = 0.1
 output = 0.25
-cache_read = 0.015
+cache_read = 0.02
 
 [limit]
 context = 1_048_576

From 4ebbc384523e05fc51a9f498900bcb634b80222e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 11:21:51 +0000
Subject: [PATCH 0699/2012] chore(sync): update OpenRouter model catalog
 (#6354)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..e3961bdffdf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.08526
+output = 0.17052
+cache_read = 0.017052
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4ace763636d..cacaaa73367 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.90045
-output = 1.8009
-cache_read = 0.075038
+input = 0.890358
+output = 1.780716
+cache_read = 0.074197
 
 [limit]
 context = 1_048_576

From d0363af59128282080b85469b3915b7f895f338f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 11:22:03 +0000
Subject: [PATCH 0700/2012] chore(sync): update LLM Gateway model catalog
 (#6356)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/novita/glm-5.3-flash.toml   | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml b/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml
index 15ec49d085e..91fc18af8c4 100644
--- a/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/novita/glm-5.3-flash.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_048_576

From ba4cc5447e8c86104992e43a64b041a241426981 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 11:22:06 +0000
Subject: [PATCH 0701/2012] chore(sync): update CrossModel model catalog
 (#6358)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/openai/gpt-6-astra.toml | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/crossmodel/models/openai/gpt-6-astra.toml

diff --git a/providers/crossmodel/models/openai/gpt-6-astra.toml b/providers/crossmodel/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..3f1fcc6d753
--- /dev/null
+++ b/providers/crossmodel/models/openai/gpt-6-astra.toml
@@ -0,0 +1,21 @@
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[modalities]
+input = ["text", "image"]

From df0ae2e6b537b14ecf0001beb13823ff6c0b7da9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 12:31:35 +0000
Subject: [PATCH 0702/2012] chore(sync): update CrossModel model catalog
 (#6363)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/openai/gpt-5.6-sol.toml    | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/crossmodel/models/openai/gpt-5.6-sol.toml b/providers/crossmodel/models/openai/gpt-5.6-sol.toml
index f7399bf4b2f..ceb6dc11d8c 100644
--- a/providers/crossmodel/models/openai/gpt-5.6-sol.toml
+++ b/providers/crossmodel/models/openai/gpt-5.6-sol.toml
@@ -5,17 +5,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 10
-output = 45
-cache_read = 1
-cache_write = 12.5
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10
 
 [modalities]
 input = ["text", "image"]

From 849bde5063decf187d7d3da49def24585f76010f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 12:31:42 +0000
Subject: [PATCH 0703/2012] chore(sync): update OpenRouter model catalog
 (#6362)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e3961bdffdf..9afb3f39949 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08526
-output = 0.17052
-cache_read = 0.017052
+input = 0.08498
+output = 0.16996
+cache_read = 0.016996
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cacaaa73367..dddcf3958a6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.890358
-output = 1.780716
-cache_read = 0.074197
+input = 0.880266
+output = 1.760532
+cache_read = 0.073356
 
 [limit]
 context = 1_048_576

From a73f16d18107cae0ddc653a8108d5e19dc7b67a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 12:31:47 +0000
Subject: [PATCH 0704/2012] chore(sync): update Eden AI model catalog (#6365)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/google/gemini-flash-latest.toml | 2 +-
 providers/edenai/models/vertex/gemini-flash-latest.toml | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/edenai/models/google/gemini-flash-latest.toml b/providers/edenai/models/google/gemini-flash-latest.toml
index 1a64f9681ee..e1d0abd66e5 100644
--- a/providers/edenai/models/google/gemini-flash-latest.toml
+++ b/providers/edenai/models/google/gemini-flash-latest.toml
@@ -1,4 +1,4 @@
-base_model = "google/gemini-3.7-flash"
+base_model = "google/gemini-3.8-flash"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-flash-latest.toml b/providers/edenai/models/vertex/gemini-flash-latest.toml
index 1a64f9681ee..e1d0abd66e5 100644
--- a/providers/edenai/models/vertex/gemini-flash-latest.toml
+++ b/providers/edenai/models/vertex/gemini-flash-latest.toml
@@ -1,4 +1,4 @@
-base_model = "google/gemini-3.7-flash"
+base_model = "google/gemini-3.8-flash"
 
 [[reasoning_options]]
 type = "effort"

From 0f8b63b03094560b8d30d1665091576e18183aca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 12:31:55 +0000
Subject: [PATCH 0705/2012] chore(sync): update NanoGPT model catalog (#6364)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TheDrummer/Anubis-70B-v1.1.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/nano-gpt/models/TheDrummer/Anubis-70B-v1.1.toml b/providers/nano-gpt/models/TheDrummer/Anubis-70B-v1.1.toml
index 36fffe8c945..117252e6472 100644
--- a/providers/nano-gpt/models/TheDrummer/Anubis-70B-v1.1.toml
+++ b/providers/nano-gpt/models/TheDrummer/Anubis-70B-v1.1.toml
@@ -15,8 +15,8 @@ output = 0.31
 cache_read = 0.155
 
 [limit]
-context = 131_072
-input = 131_072
+context = 32_000
+input = 32_000
 output = 16_384
 
 [modalities]

From 5c600a037417cf778ee6eb3ea2ce0f17abc12130 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 13:22:31 +0000
Subject: [PATCH 0706/2012] chore(sync): update OpenRouter model catalog
 (#6366)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index dddcf3958a6..b16dacbc4e8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.880266
-output = 1.760532
-cache_read = 0.073356
+input = 0.870174
+output = 1.740348
+cache_read = 0.072515
 
 [limit]
 context = 1_048_576

From c51e62240ee9984f90b98f9914d7814d81fc2d7d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 14:23:05 +0000
Subject: [PATCH 0707/2012] chore(sync): update Merge Gateway model catalog
 (#6367)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/zai/glm-5.3-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index 9423fa79264..4067b378b24 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -6,6 +6,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.015
+output = 0.05
+cache_read = 0.003

From 2ff01a65831b497708efbc6090fab6ce91227a28 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 14:23:08 +0000
Subject: [PATCH 0708/2012] chore(sync): update OpenRouter model catalog
 (#6368)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 9afb3f39949..3d8746e59d0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08498
-output = 0.16996
-cache_read = 0.016996
+input = 0.0847
+output = 0.1694
+cache_read = 0.01694
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b16dacbc4e8..206243d372f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.870174
-output = 1.740348
-cache_read = 0.072515
+input = 0.859908
+output = 1.719816
+cache_read = 0.071659
 
 [limit]
 context = 1_048_576

From 6e4ede8419cc8e38dd97eff2b29dff5d4ead3ed5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:09 +0000
Subject: [PATCH 0709/2012] chore(sync): update Merge Gateway model catalog
 (#6370)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/zai/glm-5.3-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index 4067b378b24..c26b1f6f0ee 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -8,4 +8,4 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.015
 output = 0.05
-cache_read = 0.003
+cache_read = 0.03

From a2b42d1f128b5127f2356b89f024a125d603e03f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:18 +0000
Subject: [PATCH 0710/2012] chore(sync): update Eden AI model catalog (#6373)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index d5d8773c4b6..1f11f70d4ba 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 05bf93bee1e..7e31b605e19 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424

From 48e596efb19dd0a1483aa6f020903adc8d000807 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:20 +0000
Subject: [PATCH 0711/2012] chore(sync): update OpenRouter model catalog
 (#6371)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3d8746e59d0..d53d185dadb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0847
-output = 0.1694
-cache_read = 0.01694
+input = 0.08442
+output = 0.16884
+cache_read = 0.016884
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 09d98bc8130..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12068
-output = 3.36204
-cache_read = 0.037356
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 206243d372f..1b4ffd8a683 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.859908
-output = 1.719816
-cache_read = 0.071659
+input = 0.849816
+output = 1.699632
+cache_read = 0.070818
 
 [limit]
 context = 1_048_576

From 435b72a4dc8f1a451195562fd039ec7524b3d12d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:21 +0000
Subject: [PATCH 0712/2012] chore(sync): update NanoGPT model catalog (#6374)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../NousResearch/Hermes-4-70B:thinking.toml   | 25 -------------------
 1 file changed, 25 deletions(-)
 delete mode 100644 providers/nano-gpt/models/NousResearch/Hermes-4-70B:thinking.toml

diff --git a/providers/nano-gpt/models/NousResearch/Hermes-4-70B:thinking.toml b/providers/nano-gpt/models/NousResearch/Hermes-4-70B:thinking.toml
deleted file mode 100644
index a207e324b43..00000000000
--- a/providers/nano-gpt/models/NousResearch/Hermes-4-70B:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Included in subscription
-name = "Hermes 4 (Thinking)"
-description = "General-purpose chat model for instruction following, writing, and analysis"
-family = "nousresearch"
-release_date = "2025-09-17"
-last_updated = "2025-09-17"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.2006
-output = 0.3995
-cache_read = 0.1003
-
-[limit]
-context = 128_000
-input = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 2557a4de0a8013d07bd2219bab6744507608f510 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:26 +0000
Subject: [PATCH 0713/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6375)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/deepseek-v4-flash.toml b/providers/llmgateway/models/deepseek-v4-flash.toml
index 9e1dd314ff5..18580e3a2f4 100644
--- a/providers/llmgateway/models/deepseek-v4-flash.toml
+++ b/providers/llmgateway/models/deepseek-v4-flash.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.051
-output = 0.104
-cache_read = 0.0097
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_050_000

From 754bc4b624c7c5646f3ddd80b1e4a997db7325df Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 15:22:41 +0000
Subject: [PATCH 0714/2012] chore(sync): update LLM Gateway model catalog
 (#6372)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/deepseek-v4-flash.toml         | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
index e28f16c5e0a..38a7c0ad2ce 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.16
-cache_read = 0.1
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 524_288

From ddfe8fde92191e388302eef628b642f745c28aa1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 16:24:30 +0000
Subject: [PATCH 0715/2012] chore(sync): update Kilo model catalog (#6378)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 5b7d01b0a4c51c3634d7005bcb8e85fe90e6e9ee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 16:24:32 +0000
Subject: [PATCH 0716/2012] chore(sync): update OpenRouter model catalog
 (#6377)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From f57a900926160a597e96b62dab5de44732e95e35 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 16:25:03 +0000
Subject: [PATCH 0717/2012] chore(sync): update Charm Hyper model catalog
 (#6379)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/inkling.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/hyper/models/inkling.toml

diff --git a/providers/hyper/models/inkling.toml b/providers/hyper/models/inkling.toml
new file mode 100644
index 00000000000..5379b31fe1a
--- /dev/null
+++ b/providers/hyper/models/inkling.toml
@@ -0,0 +1,18 @@
+base_model = "thinkingmachines/inkling"
+release_date = "2026-08-15"
+last_updated = "2026-09-05"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.0888
+output = 4.40964
+cache_read = 0.185096
+
+[limit]
+output = 32_768
+
+[modalities]
+input = ["text", "image"]

From cb3ca28fc99a5d33eb02470b60cf81ba17fe389f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 17:20:34 +0000
Subject: [PATCH 0718/2012] chore(sync): update OpenRouter model catalog
 (#6381)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d53d185dadb..eaeffec98f1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08442
-output = 0.16884
-cache_read = 0.016884
+input = 0.08414
+output = 0.16828
+cache_read = 0.016828
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1b4ffd8a683..f35c0a05aa4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.849816
-output = 1.699632
-cache_read = 0.070818
+input = 0.839724
+output = 1.679448
+cache_read = 0.069977
 
 [limit]
 context = 1_048_576

From 0061499964cb4ada6b64849a68ffd311e3620f38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 18:26:29 +0000
Subject: [PATCH 0719/2012] chore(sync): update Kilo model catalog (#6384)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.5-35b-a3b.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index f9494ea15f2..bdebd95519c 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -10,6 +10,7 @@ input = 0.1625
 output = 1.3
 
 [limit]
+context = 256_000
 output = 16_384
 
 [modalities]

From c85f3c09d71849783617aaf8f379439675a6db0f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 18:26:39 +0000
Subject: [PATCH 0720/2012] chore(sync): update Merge Gateway model catalog
 (#6382)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/zai/glm-5.3-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/merge-gateway/models/zai/glm-5.3-flash.toml b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
index c26b1f6f0ee..4067b378b24 100644
--- a/providers/merge-gateway/models/zai/glm-5.3-flash.toml
+++ b/providers/merge-gateway/models/zai/glm-5.3-flash.toml
@@ -8,4 +8,4 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.015
 output = 0.05
-cache_read = 0.03
+cache_read = 0.003

From 88bceeb97d554fa148699188279abe6a372375f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 18:27:06 +0000
Subject: [PATCH 0721/2012] chore(sync): update OpenRouter model catalog
 (#6383)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml     | 5 +++--
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f35c0a05aa4..501bc3d086f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.839724
-output = 1.679448
-cache_read = 0.069977
+input = 0.829458
+output = 1.658916
+cache_read = 0.069122
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index ed439da675c..8257248c271 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -6,8 +6,9 @@ base_model = "alibaba/qwen3.5-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.08
-output = 0.75
+input = 0.3125
+output = 1.25
+cache_read = 0.15625
 
 [limit]
 output = 16_384

From 7a8d93aa1efdf5c6aedb1ce79efaff89d9a63196 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 19:20:05 +0000
Subject: [PATCH 0722/2012] chore(sync): update OpenRouter model catalog
 (#6388)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index eaeffec98f1..f7cde4582ac 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08414
-output = 0.16828
-cache_read = 0.016828
+input = 0.08386
+output = 0.16772
+cache_read = 0.016772
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 501bc3d086f..cc4bcd1c726 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.829458
-output = 1.658916
-cache_read = 0.069122
+input = 0.819366
+output = 1.638732
+cache_read = 0.068281
 
 [limit]
 context = 1_048_576

From 5cafadf8deb64d83296bd5034b3946b5a3606cd1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 19:20:08 +0000
Subject: [PATCH 0723/2012] chore(sync): update NanoGPT model catalog (#6389)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/openai/gpt-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/openai/gpt-latest.toml b/providers/nano-gpt/models/openai/gpt-latest.toml
index a080340442c..4ee95a57363 100644
--- a/providers/nano-gpt/models/openai/gpt-latest.toml
+++ b/providers/nano-gpt/models/openai/gpt-latest.toml
@@ -12,13 +12,13 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
 
 [limit]
 context = 1_050_000

From fb44fc26073f0a66703e108930d221bca2e6b6db Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 20:23:46 +0000
Subject: [PATCH 0724/2012] chore(sync): update OpenRouter model catalog
 (#6391)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f7cde4582ac..569bccd0120 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08386
-output = 0.16772
-cache_read = 0.016772
+input = 0.08358
+output = 0.16716
+cache_read = 0.016716
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cc4bcd1c726..501bc3d086f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.819366
-output = 1.638732
-cache_read = 0.068281
+input = 0.829458
+output = 1.658916
+cache_read = 0.069122
 
 [limit]
 context = 1_048_576

From 4d8c3fe04132cc4047b3996546ddcc56d348f010 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 21:21:20 +0000
Subject: [PATCH 0725/2012] chore(sync): update OpenRouter model catalog
 (#6392)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 501bc3d086f..cc4bcd1c726 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.829458
-output = 1.658916
-cache_read = 0.069122
+input = 0.819366
+output = 1.638732
+cache_read = 0.068281
 
 [limit]
 context = 1_048_576

From 4b5b511b4ca166de019924e0209f8910325492f2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 22:21:59 +0000
Subject: [PATCH 0726/2012] chore(sync): update OpenRouter model catalog
 (#6394)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 569bccd0120..de4c4b47d39 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08358
-output = 0.16716
-cache_read = 0.016716
+input = 0.0833
+output = 0.1666
+cache_read = 0.01666
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cc4bcd1c726..a21494f5a5a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.819366
-output = 1.638732
-cache_read = 0.068281
+input = 0.809274
+output = 1.618548
+cache_read = 0.06744
 
 [limit]
 context = 1_048_576

From f5313cbf409592687d5a8e84fcc95b1ba58de914 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 23:21:12 +0000
Subject: [PATCH 0727/2012] chore(sync): update Kilo model catalog (#6395)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/thedrummer/unslopnemo-12b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/thedrummer/unslopnemo-12b.toml b/providers/kilo/models/thedrummer/unslopnemo-12b.toml
index 5cce6362295..f1b04078f03 100644
--- a/providers/kilo/models/thedrummer/unslopnemo-12b.toml
+++ b/providers/kilo/models/thedrummer/unslopnemo-12b.toml
@@ -14,8 +14,8 @@ input = 0.4
 output = 0.4
 
 [limit]
-context = 32_768
-output = 26_214
+context = 1_024_000
+output = 819_200
 
 [modalities]
 input = ["text"]

From dc649a39af1e77acbe98ac5cd8186c5c763a6fcd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 5 Sep 2026 23:21:19 +0000
Subject: [PATCH 0728/2012] chore(sync): update OpenRouter model catalog
 (#6396)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/thedrummer/unslopnemo-12b.toml  | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index de4c4b47d39..38f8158b5c8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0833
-output = 0.1666
-cache_read = 0.01666
+input = 0.08302
+output = 0.16604
+cache_read = 0.016604
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a21494f5a5a..263d2ca4a64 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.809274
-output = 1.618548
-cache_read = 0.06744
+input = 0.799182
+output = 1.598364
+cache_read = 0.066599
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/thedrummer/unslopnemo-12b.toml b/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
index 4d6dfbacd79..2fee039a839 100644
--- a/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
+++ b/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
@@ -16,7 +16,7 @@ output = 0.4
 
 [limit]
 context = 1_024_000
-output = 26_214
+output = 819_200
 
 [modalities]
 input = ["text"]

From 332c0b596df36ff05ae157a2a8ce69a069f4bd85 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 01:05:24 +0000
Subject: [PATCH 0729/2012] chore(sync): update OpenRouter model catalog
 (#6397)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml             | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 6 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 38f8158b5c8..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08302
-output = 0.16604
-cache_read = 0.016604
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..09d98bc8130 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 1.12068
+output = 3.36204
+cache_read = 0.037356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 263d2ca4a64..e418673776b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.799182
-output = 1.598364
-cache_read = 0.066599
+input = 0.778824
+output = 1.557648
+cache_read = 0.064902
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 6e636f26b9c..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.5
+output = 2
+cache_read = 0.1
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 51793fd9815..d6026834e83 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.15
-output = 3.5
-cache_read = 0.1
+input = 1.17
+output = 3.96
+cache_read = 0.234
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From f291f2f21e5b1a410e18313bdbb8afaff1650d79 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 01:05:43 +0000
Subject: [PATCH 0730/2012] chore(sync): update Kilo model catalog (#6398)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml      |  6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml     |  9 ++++++---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 3 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fee21b1f2d8..fd68fe0c68b 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.5
+output = 2
+cache_read = 0.1
+
+[limit]
+context = 202_752
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 82fdc44362b..a2f41897231 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.15
-output = 3.5
-cache_read = 0.1
+input = 1.17
+output = 3.96
+cache_read = 0.234
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 5c4dc2b296cf08d0f4cfa3f81f764f6040a99b8f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 02:33:17 +0000
Subject: [PATCH 0731/2012] chore(sync): update Kilo model catalog (#6400)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index a2f41897231..a88240b8a86 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -20,8 +20,8 @@ output = 3.96
 cache_read = 0.234
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From ed3bb55cd3c737908624d1c73e380b13fb1c5893 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 02:33:21 +0000
Subject: [PATCH 0732/2012] chore(sync): update OpenRouter model catalog
 (#6401)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index d6026834e83..2eab1890c5a 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.234
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From c32b77c41e94ee91e2264eeffce90f666f15d2ca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 02:33:43 +0000
Subject: [PATCH 0733/2012] chore(sync): update Venice model catalog (#6399)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-56-luna-pro.toml    | 16 ++++++++--------
 providers/venice/models/openai-gpt-56-luna.toml  | 15 +++++++++++----
 2 files changed, 19 insertions(+), 12 deletions(-)

diff --git a/providers/venice/models/openai-gpt-56-luna-pro.toml b/providers/venice/models/openai-gpt-56-luna-pro.toml
index c0bc3e7d243..0ecf54ab9e3 100644
--- a/providers/venice/models/openai-gpt-56-luna-pro.toml
+++ b/providers/venice/models/openai-gpt-56-luna-pro.toml
@@ -8,17 +8,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 1.25
-output = 7.5
-cache_read = 0.125
-cache_write = 1.5625
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.3125
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 2.5
-output = 11.25
-cache_read = 0.25
-cache_write = 3.125
+input = 0.5
+output = 2.25
+cache_read = 0.05
+cache_write = 0.625
 
 [limit]
 context = 1_000_000
diff --git a/providers/venice/models/openai-gpt-56-luna.toml b/providers/venice/models/openai-gpt-56-luna.toml
index edfe64bb25b..1f5dcbce8d6 100644
--- a/providers/venice/models/openai-gpt-56-luna.toml
+++ b/providers/venice/models/openai-gpt-56-luna.toml
@@ -7,10 +7,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.26666667
-output = 1.6
-cache_read = 0.02666667
-cache_write = 0.33333334
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.3125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.5
+output = 2.25
+cache_read = 0.05
+cache_write = 0.625
 
 [limit]
 context = 1_000_000

From 7e94b8c65767e7c76548b8baa000551090973b30 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 03:29:47 +0000
Subject: [PATCH 0734/2012] chore(sync): update Venice model catalog (#6402)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-56-sol-pro.toml     | 16 ++++++++--------
 providers/venice/models/openai-gpt-56-sol.toml   | 15 +++++++++++----
 .../venice/models/openai-gpt-56-terra-pro.toml   | 16 ++++++++--------
 providers/venice/models/openai-gpt-56-terra.toml | 15 +++++++++++----
 4 files changed, 38 insertions(+), 24 deletions(-)

diff --git a/providers/venice/models/openai-gpt-56-sol-pro.toml b/providers/venice/models/openai-gpt-56-sol-pro.toml
index 20e4ecf6d2f..36093d01914 100644
--- a/providers/venice/models/openai-gpt-56-sol-pro.toml
+++ b/providers/venice/models/openai-gpt-56-sol-pro.toml
@@ -8,17 +8,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 6.25
-output = 37.5
-cache_read = 0.625
-cache_write = 7.8125
+input = 2.5
+output = 12.5
+cache_read = 0.25
+cache_write = 3.125
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 12.5
-output = 56.25
-cache_read = 1.25
-cache_write = 15.625
+input = 5
+output = 18.75
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 1_000_000
diff --git a/providers/venice/models/openai-gpt-56-sol.toml b/providers/venice/models/openai-gpt-56-sol.toml
index c38aaba119d..dedfc528097 100644
--- a/providers/venice/models/openai-gpt-56-sol.toml
+++ b/providers/venice/models/openai-gpt-56-sol.toml
@@ -7,10 +7,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 6.25
-output = 37.5
-cache_read = 0.625
-cache_write = 7.8125
+input = 2.5
+output = 12.5
+cache_read = 0.25
+cache_write = 3.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 5
+output = 18.75
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 1_000_000
diff --git a/providers/venice/models/openai-gpt-56-terra-pro.toml b/providers/venice/models/openai-gpt-56-terra-pro.toml
index f96eba9db6e..945c54a580e 100644
--- a/providers/venice/models/openai-gpt-56-terra-pro.toml
+++ b/providers/venice/models/openai-gpt-56-terra-pro.toml
@@ -8,17 +8,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 3.125
-output = 18.75
-cache_read = 0.3125
-cache_write = 3.90625
+input = 2.5
+output = 15
+cache_read = 0.25
+cache_write = 3.125
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 6.25
-output = 28.125
-cache_read = 0.625
-cache_write = 7.8125
+input = 5
+output = 22.5
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 1_000_000
diff --git a/providers/venice/models/openai-gpt-56-terra.toml b/providers/venice/models/openai-gpt-56-terra.toml
index 5c14cdce5a4..da4706dfde0 100644
--- a/providers/venice/models/openai-gpt-56-terra.toml
+++ b/providers/venice/models/openai-gpt-56-terra.toml
@@ -7,10 +7,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 3.125
-output = 18.75
-cache_read = 0.3125
-cache_write = 3.90625
+input = 2.5
+output = 15
+cache_read = 0.25
+cache_write = 3.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 5
+output = 22.5
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 1_000_000

From 34d79c5923b8e76afd17542dba9917f266a72762 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 03:29:54 +0000
Subject: [PATCH 0735/2012] chore(sync): update OpenRouter model catalog
 (#6403)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..2852933e3df 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08218
+output = 0.16436
+cache_read = 0.016436
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e418673776b..60abfac3e72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.778824
-output = 1.557648
-cache_read = 0.064902
+input = 0.768732
+output = 1.537464
+cache_read = 0.064061
 
 [limit]
 context = 1_048_576

From 6c1584e0574c86263a25258dc49472cdd14568a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 04:28:36 +0000
Subject: [PATCH 0736/2012] chore(sync): update OpenRouter model catalog
 (#6404)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 2852933e3df..afef050d16a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08218
-output = 0.16436
-cache_read = 0.016436
+input = 0.0819
+output = 0.1638
+cache_read = 0.01638
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 60abfac3e72..ec612bbcbdb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.768732
-output = 1.537464
-cache_read = 0.064061
+input = 0.75864
+output = 1.51728
+cache_read = 0.06322
 
 [limit]
 context = 1_048_576

From 3160639458a8c57647ae44e259caa9148b38819f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 05:24:22 +0000
Subject: [PATCH 0737/2012] chore(sync): update Eden AI model catalog (#6407)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/databricks/databricks-inkling.toml  | 18 ++++++++++++++++++
 .../models/qwen/deepseek-v4-flash-0731.toml    |  4 ++--
 .../models/qwen/deepseek-v4-pro-0813.toml      |  4 ++--
 3 files changed, 22 insertions(+), 4 deletions(-)
 create mode 100644 providers/edenai/models/databricks/databricks-inkling.toml

diff --git a/providers/edenai/models/databricks/databricks-inkling.toml b/providers/edenai/models/databricks/databricks-inkling.toml
new file mode 100644
index 00000000000..e32923d89a4
--- /dev/null
+++ b/providers/edenai/models/databricks/databricks-inkling.toml
@@ -0,0 +1,18 @@
+base_model = "thinkingmachines/inkling"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "max"]
+
+[cost]
+input = 1.00002
+output = 4.04999
+cache_read = 0.17003
+cache_write = 1.00002
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 1f11f70d4ba..d5d8773c4b6 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 7e31b605e19..05bf93bee1e 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From 384895e14392fc87696659eb5acec27cb8e620b6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 05:24:28 +0000
Subject: [PATCH 0738/2012] chore(sync): update OpenRouter model catalog
 (#6406)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ec612bbcbdb..d6b763edc78 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.75864
-output = 1.51728
-cache_read = 0.06322
+input = 0.748374
+output = 1.496748
+cache_read = 0.062365
 
 [limit]
 context = 1_048_576

From 515574e3f00764443aa7c7d666a3b8ae3d9997bb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 06:36:03 +0000
Subject: [PATCH 0739/2012] chore(sync): update Deep Infra model catalog
 (#6409)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml  | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index 8ec374899bd..1e6d2eb6400 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -4,9 +4,9 @@ base_model = "deepseek/deepseek-v4-flash-0731"
 type = "toggle"
 
 [cost]
-input = 0.08
+input = 0.06
 output = 0.18
-cache_read = 0.016
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 893c99615fc543f22e885ff469649304c1ca10cb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 06:36:06 +0000
Subject: [PATCH 0740/2012] chore(sync): update OpenRouter model catalog
 (#6410)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml               | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index afef050d16a..b0cf0e4c88e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0819
-output = 0.1638
-cache_read = 0.01638
+input = 0.08162
+output = 0.16324
+cache_read = 0.016324
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index d6b763edc78..1530b62b57b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.748374
-output = 1.496748
-cache_read = 0.062365
+input = 0.738282
+output = 1.476564
+cache_read = 0.061524
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..6e636f26b9c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.55
+output = 2.2
+cache_read = 0.11

From 4cf35d0eaa8573152e01bf7c5801981eadb7ee2d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 06:36:15 +0000
Subject: [PATCH 0741/2012] chore(sync): update Kilo model catalog (#6408)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..fee21b1f2d8 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
-
-[limit]
-context = 202_752
+input = 0.55
+output = 2.2
+cache_read = 0.11

From a37a81c82477fa02c8bcedff27aa76ec640cd508 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 07:25:35 +0000
Subject: [PATCH 0742/2012] chore(sync): update OpenRouter model catalog
 (#6411)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b0cf0e4c88e..140db44f179 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08162
-output = 0.16324
-cache_read = 0.016324
+input = 0.08134
+output = 0.16268
+cache_read = 0.016268
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1530b62b57b..e010cc1159a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.738282
-output = 1.476564
-cache_read = 0.061524
+input = 0.72819
+output = 1.45638
+cache_read = 0.060683
 
 [limit]
 context = 1_048_576

From f2ac278924c9e6a27de4ed0df0711d9266695639 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 08:29:44 +0000
Subject: [PATCH 0743/2012] chore(sync): update Kilo model catalog (#6412)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
index 9f316a31c44..7cb4fe7a46b 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
@@ -12,4 +12,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072

From 474681271c795ced097421d986d0297bbade9f7a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 08:30:04 +0000
Subject: [PATCH 0744/2012] chore(sync): update OpenRouter model catalog
 (#6413)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 8 ++++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index d06dd2cd8d1..26fedac913a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,10 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.016
+input = 0.04998
+output = 0.09996
+cache_read = 0.009996
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e010cc1159a..fe90d3804bd 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.72819
-output = 1.45638
-cache_read = 0.060683
+input = 0.717924
+output = 1.435848
+cache_read = 0.059827
 
 [limit]
 context = 1_048_576

From f7d03035a6076024f16c2223cb0a0a6180f504d4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 09:24:39 +0000
Subject: [PATCH 0745/2012] chore(sync): update Kilo model catalog (#6416)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 5b7d7be3b0d..67b240e58e2 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04998
-output = 0.09996
-cache_read = 0.009996
+input = 0.045
+output = 0.09
+cache_read = 0.009
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 3937dc83abe06c5a9a3b350b035593645d98f6b1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 09:24:42 +0000
Subject: [PATCH 0746/2012] chore(sync): update OpenRouter model catalog
 (#6415)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 140db44f179..eed3333110b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08134
-output = 0.16268
-cache_read = 0.016268
+input = 0.08106
+output = 0.16212
+cache_read = 0.016212
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index fe90d3804bd..18cb700cd9e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.717924
-output = 1.435848
-cache_read = 0.059827
+input = 0.707832
+output = 1.415664
+cache_read = 0.058986
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 4e3d7984114..a6efa176c68 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04998
-output = 0.09996
-cache_read = 0.009996
+input = 0.045
+output = 0.09
+cache_read = 0.009
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 98383f3755693d8b173ec1a2ff8bd0ae851ef207 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 09:24:50 +0000
Subject: [PATCH 0747/2012] chore(sync): update NanoGPT model catalog (#6417)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml  | 2 +-
 providers/nano-gpt/models/openai/gpt-5.1-2025-11-13.toml | 6 +++---
 providers/nano-gpt/models/openai/gpt-5.4.toml            | 2 +-
 providers/nano-gpt/models/openai/gpt-5.5.toml            | 3 +--
 4 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
index a8fda51f1ef..ed81087ddc3 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
+++ b/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
@@ -3,7 +3,7 @@ structured_output = true
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "high"]
+values = ["none", "low", "medium", "high", "max"]
 
 [cost]
 input = 2
diff --git a/providers/nano-gpt/models/openai/gpt-5.1-2025-11-13.toml b/providers/nano-gpt/models/openai/gpt-5.1-2025-11-13.toml
index 562a40ae89a..7ad09d53794 100644
--- a/providers/nano-gpt/models/openai/gpt-5.1-2025-11-13.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.1-2025-11-13.toml
@@ -19,9 +19,9 @@ output = 10
 cache_read = 0.125
 
 [limit]
-context = 1_000_000
-input = 1_000_000
-output = 32_768
+context = 400_000
+input = 400_000
+output = 128_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-5.4.toml b/providers/nano-gpt/models/openai/gpt-5.4.toml
index 90fff170418..4b213fbe090 100644
--- a/providers/nano-gpt/models/openai/gpt-5.4.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.4.toml
@@ -12,4 +12,4 @@ output = 15
 cache_read = 0.25
 
 [limit]
-context = 922_000
+input = 1_050_000
diff --git a/providers/nano-gpt/models/openai/gpt-5.5.toml b/providers/nano-gpt/models/openai/gpt-5.5.toml
index cfb2b4240b9..1cf345d52f8 100644
--- a/providers/nano-gpt/models/openai/gpt-5.5.toml
+++ b/providers/nano-gpt/models/openai/gpt-5.5.toml
@@ -12,5 +12,4 @@ output = 30
 cache_read = 0.5
 
 [limit]
-context = 1_000_000
-input = 1_000_000
+input = 1_050_000

From 83e16488a02329e077f0378be03bccea9c100dc6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 10:23:32 +0000
Subject: [PATCH 0748/2012] chore(sync): update OpenRouter model catalog
 (#6419)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index eed3333110b..403d518788f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08106
-output = 0.16212
-cache_read = 0.016212
+input = 0.08078
+output = 0.16156
+cache_read = 0.016156
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 18cb700cd9e..2f5edfbb4ed 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.707832
-output = 1.415664
-cache_read = 0.058986
+input = 0.69774
+output = 1.39548
+cache_read = 0.058145
 
 [limit]
 context = 1_048_576

From 5e7a140bc50960ba5271a5e9eb68f6f853e7b481 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 10:23:42 +0000
Subject: [PATCH 0749/2012] chore(sync): update NanoGPT model catalog (#6418)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/TheDrummer/Artemis-v1.1.toml       | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/nano-gpt/models/TheDrummer/Artemis-v1.1.toml

diff --git a/providers/nano-gpt/models/TheDrummer/Artemis-v1.1.toml b/providers/nano-gpt/models/TheDrummer/Artemis-v1.1.toml
new file mode 100644
index 00000000000..92b0778c727
--- /dev/null
+++ b/providers/nano-gpt/models/TheDrummer/Artemis-v1.1.toml
@@ -0,0 +1,24 @@
+name = "TheDrummer/Artemis v1.1"
+description = "TheDrummer's Artemis v1.1 is a Gemma 4 31B fine-tune for creative writing, expressive dialogue, and roleplay, with optional thinking and a 262K context window."
+release_date = "2026-09-06"
+last_updated = "2026-09-06"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.1
+output = 0.45
+cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 8b4eefbb5dd56edac59dde2b93dff688225733c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 11:21:31 +0000
Subject: [PATCH 0750/2012] chore(sync): update OpenRouter model catalog
 (#6420)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 2f5edfbb4ed..f373d7eeeb6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.69774
-output = 1.39548
-cache_read = 0.058145
+input = 0.687648
+output = 1.375296
+cache_read = 0.057304
 
 [limit]
 context = 1_048_576

From 89dfccad54909c8e91432eb5a44b5cde8246232b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 12:32:02 +0000
Subject: [PATCH 0751/2012] chore(sync): update OpenRouter model catalog
 (#6421)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 403d518788f..03a5191e437 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08078
-output = 0.16156
-cache_read = 0.016156
+input = 0.0805
+output = 0.161
+cache_read = 0.0161
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f373d7eeeb6..1ca663922d9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.687648
-output = 1.375296
-cache_read = 0.057304
+input = 0.677382
+output = 1.354764
+cache_read = 0.056449
 
 [limit]
 context = 1_048_576

From 39149991cf800c3bb0d83d2129e4673460e86aea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 13:22:26 +0000
Subject: [PATCH 0752/2012] chore(sync): update OpenRouter model catalog
 (#6422)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml     |  6 +++---
 .../openrouter/models/z-ai/glm-5.2:free.toml   | 18 ------------------
 2 files changed, 3 insertions(+), 21 deletions(-)
 delete mode 100644 providers/openrouter/models/z-ai/glm-5.2:free.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 03a5191e437..404f8d66b38 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0805
-output = 0.161
-cache_read = 0.0161
+input = 0.08022
+output = 0.16044
+cache_read = 0.016044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.2:free.toml b/providers/openrouter/models/z-ai/glm-5.2:free.toml
deleted file mode 100644
index 4139b667397..00000000000
--- a/providers/openrouter/models/z-ai/glm-5.2:free.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2 (free)"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "xhigh"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 256_000
-output = 230_400

From 9e669cd2f32416d81fb64cc8360a5586361c3265 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 14:23:04 +0000
Subject: [PATCH 0753/2012] chore(sync): update OpenRouter model catalog
 (#6423)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 404f8d66b38..c19d2ff4995 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08022
-output = 0.16044
-cache_read = 0.016044
+input = 0.07994
+output = 0.15988
+cache_read = 0.015988
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1ca663922d9..61d1cf5f807 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.677382
-output = 1.354764
-cache_read = 0.056449
+input = 0.66729
+output = 1.33458
+cache_read = 0.055608
 
 [limit]
 context = 1_048_576

From fa9def07c1dec51be5b9f6cfe1a3c392a3ac59da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 15:22:46 +0000
Subject: [PATCH 0754/2012] chore(sync): update OpenRouter model catalog
 (#6425)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c19d2ff4995..404f8d66b38 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07994
-output = 0.15988
-cache_read = 0.015988
+input = 0.08022
+output = 0.16044
+cache_read = 0.016044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 09d98bc8130..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12068
-output = 3.36204
-cache_read = 0.037356
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576

From c4635967a23282bfdcd5afae252ccd87ff6d24e7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 15:22:53 +0000
Subject: [PATCH 0755/2012] chore(sync): update Eden AI model catalog (#6424)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml  | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml        | 4 ++--
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index e065c5297a8..ac478b62e04 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.08
+input = 0.06
 output = 0.18
-cache_read = 0.016
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index d5d8773c4b6..1f11f70d4ba 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 05bf93bee1e..7e31b605e19 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424

From 599254b0f0de5ef3ed7d0b7c7ccf9d543803b0c1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 16:25:16 +0000
Subject: [PATCH 0756/2012] chore(sync): update OpenRouter model catalog
 (#6427)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 404f8d66b38..c19d2ff4995 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08022
-output = 0.16044
-cache_read = 0.016044
+input = 0.07994
+output = 0.15988
+cache_read = 0.015988
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 61d1cf5f807..96c9123e736 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.66729
-output = 1.33458
-cache_read = 0.055608
+input = 0.657198
+output = 1.314396
+cache_read = 0.054767
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From a68e46a7ebffa2a2ab6737abc7385ae2497e284c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 16:25:19 +0000
Subject: [PATCH 0757/2012] chore(sync): update Kilo model catalog (#6426)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 052514de6a801d71401487f5f4aa36bfa1ca347c Mon Sep 17 00:00:00 2001
From: Ori Perry <48057913+oriori1703@users.noreply.github.com>
Date: Sun, 6 Sep 2026 19:39:17 +0300
Subject: [PATCH 0758/2012] feat(models): add GPT-6 Astra benchmarks (#6414)

---
 models/openai/gpt-6-astra.toml | 145 +++++++++++++++++++++++++++++++++
 1 file changed, 145 insertions(+)

diff --git a/models/openai/gpt-6-astra.toml b/models/openai/gpt-6-astra.toml
index e40544c0dbf..9676745a8b8 100644
--- a/models/openai/gpt-6-astra.toml
+++ b/models/openai/gpt-6-astra.toml
@@ -18,3 +18,148 @@ output = 128_000
 [modalities]
 input = ["text", "image", "pdf"]
 output = ["text"]
+
+[[benchmarks]]
+name = "Agents' Last Exam"
+score = 59.3
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "OSWorld"
+score = 72.6
+metric = "partial score"
+dataset = "V2-Offline (v2026.08.08)"
+version = "2.0"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "ScreenSpot-Pro"
+score = 92.7
+metric = "accuracy"
+variant = "no tools"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "AutomationBench"
+score = 41.4
+metric = "success rate"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "BenchCAD"
+score = 95.9
+metric = "geometric overlap"
+variant = "with tools"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "BrowseComp"
+score = 91.5
+metric = "accuracy"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "Terminal-Bench"
+score = 57.9
+metric = "success rate"
+version = "4.0"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "DeepSWE"
+score = 74.1
+metric = "resolve rate"
+version = "1.1"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "FrontierCode"
+score = 64.5
+metric = "score"
+dataset = "Extended"
+version = "1.1"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "Terminal-Bench Science"
+score = 64.6
+metric = "success rate"
+version = "0.1"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "FrontierMath"
+score = 97.6
+metric = "accuracy"
+dataset = "Tier 4"
+version = "v2"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "GPQA Diamond"
+score = 96.0
+metric = "accuracy"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "Humanity's Last Exam"
+score = 57.2
+metric = "accuracy"
+variant = "with tools"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "ExploitBench"
+score = 100.0
+metric = "success rate"
+variant = "without production safeguards"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "SRE-Bench"
+score = 88.0
+metric = "success rate"
+variant = "single attempt"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "ARC-AGI-3"
+score = 99.9
+metric = "RHAE"
+harness = "Responses API"
+source = "https://openai.com/index/gpt-6-astra/"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "Artificial Analysis Intelligence Index"
+score = 61
+metric = "index score"
+variant = "max"
+version = "4.1.1"
+source = "https://artificialanalysis.ai/articles/benchmarking-gpt-6-astra"
+date = "2026-09-03"
+
+[[benchmarks]]
+name = "Artificial Analysis Coding Agent Index"
+score = 67
+metric = "index score"
+harness = "Codex"
+variant = "max"
+version = "1.4"
+source = "https://artificialanalysis.ai/articles/benchmarking-gpt-6-astra"
+date = "2026-09-03"

From ed60706c77b5543f84b44cdb14038a07cfad6bae Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E9=9B=AA=E9=9C=81=E6=B8=85=E5=AF=92?= <cn_hui@163.com>
Date: Mon, 7 Sep 2026 00:40:00 +0800
Subject: [PATCH 0759/2012] feat(sensenova): add deepseek-v4-pro and kimi-k3
 (#6405)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Add DeepSeek V4 Pro 0813 and Kimi K3 to the SenseNova provider based on
the live GET /v1/models catalog, host docs, and live API probes.

deepseek-v4-pro: base_model deepseek/deepseek-v4-pro-0813 (host serves
the 0813 snapshot); reasoning toggle via thinking.type plus effort
high/max (docs: reasoning_effort low/high/max, medium/xhigh mapped to
high; live probes show low == high reasoning volume); catalog limits
context 1048576 / output 65536; zero pricing.

kimi-k3: native multimodal agent model — image input via base64
image_url verified live (video undocumented, so input is text+image);
reasoning_effort low/high/max per docs and confirmed graded by live
probes; catalog output 65536; zero pricing.

Both expose the reasoning side channel as reasoning_content.

Co-authored-by: lzhuii <lzhuii@users.noreply.github.com>
---
 .../sensenova/models/deepseek-v4-pro.toml     | 30 +++++++++++++++++++
 providers/sensenova/models/kimi-k3.toml       | 30 +++++++++++++++++++
 2 files changed, 60 insertions(+)
 create mode 100644 providers/sensenova/models/deepseek-v4-pro.toml
 create mode 100644 providers/sensenova/models/kimi-k3.toml

diff --git a/providers/sensenova/models/deepseek-v4-pro.toml b/providers/sensenova/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..c794055d51a
--- /dev/null
+++ b/providers/sensenova/models/deepseek-v4-pro.toml
@@ -0,0 +1,30 @@
+# Sources (accessed 2026-09-06):
+# https://platform.sensenova.cn/docs
+# GET /v1/models: context 1048576, max output 65536, fp8, text -> text,
+# supported features tools+json_mode+reasoning; pricing all zero (public beta
+# free on TokenPlan).
+# Version note: the host serves DeepSeek-V4-Pro 0813, so base_model points at
+# the 0813 snapshot. DeepSeek V4 Pro is text-only (no image input).
+# Toggle: thinking.type = enabled|disabled (default enabled)
+# Effort: reasoning_effort = low|high|max; medium and xhigh are accepted but
+# mapped to high. Live calls confirm none disables reasoning and that no
+# better-than-high achievable output differs, so the lab-effective levels
+# (high/max) are authored per AGENTS.md policy for DeepSeek V4.
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
\ No newline at end of file
diff --git a/providers/sensenova/models/kimi-k3.toml b/providers/sensenova/models/kimi-k3.toml
new file mode 100644
index 00000000000..46fe6450863
--- /dev/null
+++ b/providers/sensenova/models/kimi-k3.toml
@@ -0,0 +1,30 @@
+# Sources (accessed 2026-09-06):
+# https://platform.sensenova.cn/docs
+# GET /v1/models: context 1048576, max output 65536, fp8, text -> text,
+# supported features tools+json_mode+reasoning; pricing all zero (public beta
+# free on TokenPlan).
+# Reasoning: host docs give reasoning_effort low|high|max (default max); live
+# calls show clear grading (low = 1-14 reasoning tokens, high = 7-36, max =
+# 593-626 on the same prompt). No separate on/off toggle is documented on this
+# host, so the documented effort levels are authored as-is.
+# Image input via base64 image_url is verified live and documented (native
+# multimodal agent model); host docs require base64 (public URLs rejected) and
+# do not document video, so input is limited to text+image.
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
\ No newline at end of file

From def9517ce331f4edeaa03ed71f28d5a2fd064f08 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 11:40:09 -0500
Subject: [PATCH 0760/2012] fix: dev CI failure (#6380)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cloudflare-ai-gateway/curation.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/cloudflare-ai-gateway/curation.toml b/providers/cloudflare-ai-gateway/curation.toml
index 697217a5ffb..916be4ca9a8 100644
--- a/providers/cloudflare-ai-gateway/curation.toml
+++ b/providers/cloudflare-ai-gateway/curation.toml
@@ -33,6 +33,7 @@ skip = [
   "google/gemini-3.5-flash",
   "google/gemini-3.6-flash",
   "google/gemini-3.7-flash",
+  "google/gemini-3.8-flash",
 ]
 
 [models."anthropic/claude-fable-5"]

From 6fc598a466e735b1a202d3d7a120c389086c41a9 Mon Sep 17 00:00:00 2001
From: Minato Minami <me@mina.to>
Date: Mon, 7 Sep 2026 01:40:30 +0900
Subject: [PATCH 0761/2012] Add CrofAI Deepseek v4 flash vision exp (#6359)

---
 .../models/deepseek-v4-flash-vision-exp.toml     | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/crof/models/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/crof/models/deepseek-v4-flash-vision-exp.toml b/providers/crof/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..c9af088f346
--- /dev/null
+++ b/providers/crof/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.08
+output = 0.2
+cache_read = 0.007
+
+[limit]
+output = 131_072
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
\ No newline at end of file

From 18eab0042661aaff6edff3abedab90334921d86e Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 11:40:44 -0500
Subject: [PATCH 0762/2012] fix: [missing-model] ofox: openai/gpt-6-astra
 (#6340)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/openai/gpt-6-astra.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/openai/gpt-6-astra.toml

diff --git a/providers/ofox/models/openai/gpt-6-astra.toml b/providers/ofox/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..4bf30a6b531
--- /dev/null
+++ b/providers/ofox/models/openai/gpt-6-astra.toml
@@ -0,0 +1,17 @@
+# Sources:
+# - Ofox model page: https://ofox.ai/models/openai/gpt-6-astra
+# - Ofox catalog API: https://api.ofox.ai/v2/models/catalog?include=provider_price&search=gpt-6-astra
+# - OpenAI model docs (reasoning effort): https://developers.openai.com/api/docs/models/gpt-6-astra
+# Ofox publishes flat $10/$50 (+ cache) with no >=272k context tier in the catalog, so no [[cost.tiers]].
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/openai"
+api = "https://api.ofox.ai/v1"

From ec65575f5ea791c96fe45834f23e12a93550a34f Mon Sep 17 00:00:00 2001
From: Varun Jhaveri <85669579+CYCLOP5@users.noreply.github.com>
Date: Sun, 6 Sep 2026 22:11:19 +0530
Subject: [PATCH 0763/2012] feat(providers): add GPT-6 Astra hosted model
 (#6390)

---
 providers/azure/models/gpt-6-astra.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/azure/models/gpt-6-astra.toml

diff --git a/providers/azure/models/gpt-6-astra.toml b/providers/azure/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..ad77bce32d7
--- /dev/null
+++ b/providers/azure/models/gpt-6-astra.toml
@@ -0,0 +1,17 @@
+# Azure pricing page publishes GPT-6 pricing in the blog rather than the service pricing page; https://azure.microsoft.com/en-us/blog/gpt-6-astra-frontier-intelligence-for-work-now-generally-available-in-microsoft-foundry/
+base_model = "openai/gpt-6-astra"
+status = "beta"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.00
+output = 50.00
+cache_read = 1.00
+cache_write = 12.50
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20.00
+output = 75.00
+cache_read = 2.00
+cache_write = 25.00

From 44fa5bfa0a8e80a722508c7b0a2adab27d079fba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 11:41:36 -0500
Subject: [PATCH 0764/2012] chore(sync): update Tinfoil model catalog (#6328)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/gpt-oss-safeguard-120b.toml | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/providers/tinfoil/models/gpt-oss-safeguard-120b.toml b/providers/tinfoil/models/gpt-oss-safeguard-120b.toml
index 4ff8d43f992..f1c52ce51c8 100644
--- a/providers/tinfoil/models/gpt-oss-safeguard-120b.toml
+++ b/providers/tinfoil/models/gpt-oss-safeguard-120b.toml
@@ -1,11 +1,8 @@
 base_model = "openai/gpt-oss-safeguard-120b"
 name = "gpt-oss-safeguard-120b"
+reasoning = false
 knowledge = "2024-06"
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
 [cost]
 input = 0.15
 output = 0.6

From 7f495e623abe0478e2b11526de8295b059b6b7cd Mon Sep 17 00:00:00 2001
From: Andre Landgraf <andre.timo.landgraf@gmail.com>
Date: Sun, 6 Sep 2026 09:41:51 -0700
Subject: [PATCH 0765/2012] Neon: add gpt-6-astra (#6386)

* Neon: add gpt-6-astra

* Put GPT-6 Astra's knowledge cutoff on the lab model.
---
 models/openai/gpt-6-astra.toml         |  1 +
 providers/neon/models/gpt-6-astra.toml | 23 +++++++++++++++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 providers/neon/models/gpt-6-astra.toml

diff --git a/models/openai/gpt-6-astra.toml b/models/openai/gpt-6-astra.toml
index 9676745a8b8..ea7015c41e4 100644
--- a/models/openai/gpt-6-astra.toml
+++ b/models/openai/gpt-6-astra.toml
@@ -8,6 +8,7 @@ reasoning = true
 temperature = false
 tool_call = true
 structured_output = true
+knowledge = "2026-04-30"
 open_weights = false
 
 [limit]
diff --git a/providers/neon/models/gpt-6-astra.toml b/providers/neon/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..2c0f285f346
--- /dev/null
+++ b/providers/neon/models/gpt-6-astra.toml
@@ -0,0 +1,23 @@
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[modalities]
+output = ["text", "image"]
+
+[provider]
+npm = "@ai-sdk/openai"
+api = "${NEON_AI_GATEWAY_BASE_URL}/openai/v1"
+shape = "responses"

From 744d18fc86f6ec01bd6a57f977dd84ab1eb9da78 Mon Sep 17 00:00:00 2001
From: Andre Landgraf <andre.timo.landgraf@gmail.com>
Date: Sun, 6 Sep 2026 09:42:20 -0700
Subject: [PATCH 0766/2012] Neon: add claude-fable-5-1, glm-5-3-flash and
 grok-4-6 (#6284)

* Neon: add claude-fable-5-1, glm-5-3-flash, grok-4-6

Mirror the neon.com/models entries. Grok has no Neon [cost] yet. Fable 5.1 structured_output is false; GLM list prices are not the first-party promo.

* fix: encode Neon grok/glm reasoning_options per models.dev rules

Grok off is effort=none, so the thinking toggle cannot sit beside it. GLM keeps toggle plus the measured effort list, with the wire path at the top of the file.

* docs: describe Fable structured_output rejection without naming the translator

The public PR body already uses "the gateway"; keep the file comment matched.

* fix: drop GLM/Grok toggles that do not turn reasoning off

glm-5-3-flash thinking.type=disabled still returns reasoning_content.
Grok off is effort=none. Inherit lab knowledge and modalities.

* Narrow Neon GLM and Grok effort lists to lab/peer values with measured effect.

Keep grok on Responses; keep Fable cache_read at the neon.com/models rate.
---
 providers/neon/models/claude-fable-5-1.toml | 19 +++++++++++++++++
 providers/neon/models/glm-5-3-flash.toml    | 20 ++++++++++++++++++
 providers/neon/models/grok-4-6.toml         | 23 +++++++++++++++++++++
 3 files changed, 62 insertions(+)
 create mode 100644 providers/neon/models/claude-fable-5-1.toml
 create mode 100644 providers/neon/models/glm-5-3-flash.toml
 create mode 100644 providers/neon/models/grok-4-6.toml

diff --git a/providers/neon/models/claude-fable-5-1.toml b/providers/neon/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..26a08223218
--- /dev/null
+++ b/providers/neon/models/claude-fable-5-1.toml
@@ -0,0 +1,19 @@
+# Chat Completions `reasoning_effort` is rejected (`Extra inputs are not permitted`).
+# Effort is the Anthropic Messages control: thinking.type adaptive + output_config.effort.
+# No toggle: thinking.type.disabled is rejected, same as claude-fable-5.
+# json_schema on /v1/chat/completions is translated via forced tool use; this
+# model rejects that (INVALID_PARAMETER_VALUE). Measured 2026-09-04.
+# [cost] is the neon.com/models list for this id (10 / 50 / cache_read 0.25 /
+# cache_write 12.5). Sibling claude-fable-5 publishes cache_read = 1; do not
+# copy that onto 5.1.
+base_model = "anthropic/claude-fable-5-1"
+structured_output = false
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/neon/models/glm-5-3-flash.toml b/providers/neon/models/glm-5-3-flash.toml
new file mode 100644
index 00000000000..d2dad8e846c
--- /dev/null
+++ b/providers/neon/models/glm-5-3-flash.toml
@@ -0,0 +1,20 @@
+# Effort: chat completions reasoning_effort.
+# Lab/peers: low|high|max. Neon accepts extra values (minimal, medium, xhigh)
+# with HTTP 200, but a 2026-09-04 effect probe did not show distinct tiers
+# (reasoning_content length is not monotonic). `none` is rejected (thinking-only).
+# thinking.type=disabled returns 200 and still emits reasoning_content.
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash"
+reasoning_options = [
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+# List rates. First-party zhipuai currently publishes a 50% promo that Neon does not.
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+# Lab context is 1_000_000. neon.com/models publishes 1_048_576.
+[limit]
+context = 1_048_576
diff --git a/providers/neon/models/grok-4-6.toml b/providers/neon/models/grok-4-6.toml
new file mode 100644
index 00000000000..b17b6990ea0
--- /dev/null
+++ b/providers/neon/models/grok-4-6.toml
@@ -0,0 +1,23 @@
+# Effort: Chat Completions `reasoning_effort` and Responses `reasoning.effort`.
+# Native dialect is openai-responses: POST /openai/v1/responses returns 200.
+# Chat Completions also returns 200 and accepts `reasoning_effort`.
+# Lab/peers: low|medium|high|xhigh. `max` is rejected.
+# `none` and `minimal` return 200 but do not expose a reasoning channel to
+# compare against, and first-party xAI does not treat this model as disableable.
+# Do not advertise an off switch we cannot verify.
+base_model = "xai/grok-4.6"
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh"] },
+]
+
+# Neon rate is not published yet. Do not copy first-party xAI prices.
+
+# Lab output is 500_000. The gateway names 524_288:
+# 'max_tokens' (100000000) exceeds model maximum (524288). Measured 2026-09-04.
+[limit]
+output = 524_288
+
+[provider]
+npm = "@ai-sdk/openai"
+api = "${NEON_AI_GATEWAY_BASE_URL}/openai/v1"
+shape = "responses"

From c7c07f4783d7f4366fddccb1ffde896ed1dcda82 Mon Sep 17 00:00:00 2001
From: Tony Yu <49832190+ysntony@users.noreply.github.com>
Date: Mon, 7 Sep 2026 00:42:50 +0800
Subject: [PATCH 0767/2012] fix(moonshotai): remove retired models that 404 on
 first-party APIs (#6101)

Moonshot retired these models on its own platform (api.moonshot.ai / api.moonshot.cn) and calls now return 404 model-not-found: the kimi-k2 series (0711/0905/turbo previews, thinking, thinking-turbo) was discontinued 2026-05-25, and kimi-k2.5 was discontinued 2026-08-31.
Source: https://platform.kimi.ai/docs/models
Third-party hosts that still serve these models (e.g. OpenRouter's entries under providers/openrouter/models/moonshotai/) are untouched, as are the lab metadata files under models/moonshotai/.
---
 .../models/kimi-k2-0711-preview.toml          |  1 -
 .../models/kimi-k2-0905-preview.toml          |  1 -
 .../models/kimi-k2-thinking-turbo.toml        |  1 -
 .../models/kimi-k2-thinking.toml              |  1 -
 .../models/kimi-k2-turbo-preview.toml         |  1 -
 providers/moonshotai-cn/models/kimi-k2.5.toml |  1 -
 .../models/kimi-k2-0711-preview.toml          | 24 --------------
 .../models/kimi-k2-0905-preview.toml          | 24 --------------
 .../models/kimi-k2-thinking-turbo.toml        | 28 -----------------
 .../moonshotai/models/kimi-k2-thinking.toml   | 28 -----------------
 .../models/kimi-k2-turbo-preview.toml         | 24 --------------
 providers/moonshotai/models/kimi-k2.5.toml    | 31 -------------------
 12 files changed, 165 deletions(-)
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2-0711-preview.toml
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2-0905-preview.toml
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2-thinking-turbo.toml
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2-thinking.toml
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2-turbo-preview.toml
 delete mode 120000 providers/moonshotai-cn/models/kimi-k2.5.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2-0711-preview.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2-0905-preview.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2-thinking-turbo.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2-thinking.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2-turbo-preview.toml
 delete mode 100644 providers/moonshotai/models/kimi-k2.5.toml

diff --git a/providers/moonshotai-cn/models/kimi-k2-0711-preview.toml b/providers/moonshotai-cn/models/kimi-k2-0711-preview.toml
deleted file mode 120000
index 8e4e60785e7..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2-0711-preview.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2-0711-preview.toml
\ No newline at end of file
diff --git a/providers/moonshotai-cn/models/kimi-k2-0905-preview.toml b/providers/moonshotai-cn/models/kimi-k2-0905-preview.toml
deleted file mode 120000
index d74c6ce2f56..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2-0905-preview.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2-0905-preview.toml
\ No newline at end of file
diff --git a/providers/moonshotai-cn/models/kimi-k2-thinking-turbo.toml b/providers/moonshotai-cn/models/kimi-k2-thinking-turbo.toml
deleted file mode 120000
index 1f1081c786a..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2-thinking-turbo.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2-thinking-turbo.toml
\ No newline at end of file
diff --git a/providers/moonshotai-cn/models/kimi-k2-thinking.toml b/providers/moonshotai-cn/models/kimi-k2-thinking.toml
deleted file mode 120000
index 31cc2adf45b..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2-thinking.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2-thinking.toml
\ No newline at end of file
diff --git a/providers/moonshotai-cn/models/kimi-k2-turbo-preview.toml b/providers/moonshotai-cn/models/kimi-k2-turbo-preview.toml
deleted file mode 120000
index f2e9b0093e5..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2-turbo-preview.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2-turbo-preview.toml
\ No newline at end of file
diff --git a/providers/moonshotai-cn/models/kimi-k2.5.toml b/providers/moonshotai-cn/models/kimi-k2.5.toml
deleted file mode 120000
index ec89508b7cd..00000000000
--- a/providers/moonshotai-cn/models/kimi-k2.5.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../moonshotai/models/kimi-k2.5.toml
\ No newline at end of file
diff --git a/providers/moonshotai/models/kimi-k2-0711-preview.toml b/providers/moonshotai/models/kimi-k2-0711-preview.toml
deleted file mode 100644
index 81ef972cb71..00000000000
--- a/providers/moonshotai/models/kimi-k2-0711-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kimi K2 0711"
-description = "Kimi model for long-context chat, coding, and agentic reasoning"
-family = "kimi-k2"
-release_date = "2025-07-14"
-last_updated = "2025-07-14"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = true
-
-[cost]
-input = 0.6
-output = 2.5
-cache_read = 0.15
-
-[limit]
-context = 131_072
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/moonshotai/models/kimi-k2-0905-preview.toml b/providers/moonshotai/models/kimi-k2-0905-preview.toml
deleted file mode 100644
index cab06b8acdd..00000000000
--- a/providers/moonshotai/models/kimi-k2-0905-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kimi K2 0905"
-description = "Kimi model for long-context chat, coding, and agentic reasoning"
-family = "kimi-k2"
-release_date = "2025-09-05"
-last_updated = "2025-09-05"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = true
-
-[cost]
-input = 0.6
-output = 2.5
-cache_read = 0.15
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/moonshotai/models/kimi-k2-thinking-turbo.toml b/providers/moonshotai/models/kimi-k2-thinking-turbo.toml
deleted file mode 100644
index 2f36a534239..00000000000
--- a/providers/moonshotai/models/kimi-k2-thinking-turbo.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Kimi K2 Thinking Turbo"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
-family = "kimi-thinking"
-release_date = "2025-11-06"
-last_updated = "2025-11-06"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2024-08"
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.15
-output = 8
-cache_read = 0.15
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/moonshotai/models/kimi-k2-thinking.toml b/providers/moonshotai/models/kimi-k2-thinking.toml
deleted file mode 100644
index ed8742f8f36..00000000000
--- a/providers/moonshotai/models/kimi-k2-thinking.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Kimi K2 Thinking"
-description = "Thinking Kimi model for slower research passes, planning, and hard technical questions"
-family = "kimi-thinking"
-release_date = "2025-11-06"
-last_updated = "2025-11-06"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2024-08"
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.6
-output = 2.5
-cache_read = 0.15
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/moonshotai/models/kimi-k2-turbo-preview.toml b/providers/moonshotai/models/kimi-k2-turbo-preview.toml
deleted file mode 100644
index 262b7ad957e..00000000000
--- a/providers/moonshotai/models/kimi-k2-turbo-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kimi K2 Turbo"
-description = "Fast Kimi model for responsive chat, coding help, and agent loops"
-family = "kimi-k2"
-release_date = "2025-09-05"
-last_updated = "2025-09-05"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = true
-
-[cost]
-input = 2.4
-output = 10
-cache_read = 0.6
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/moonshotai/models/kimi-k2.5.toml b/providers/moonshotai/models/kimi-k2.5.toml
deleted file mode 100644
index 9db35770f6b..00000000000
--- a/providers/moonshotai/models/kimi-k2.5.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "Kimi K2.5"
-description = "Earlier Kimi frontier model for long-context agents, coding, and multimodal work"
-family = "kimi-k2"
-release_date = "2026-01"
-last_updated = "2026-01"
-attachment = true
-reasoning = true
-structured_output = true
-temperature = false
-tool_call = true
-knowledge = "2025-01"
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.6
-output = 3.0
-cache_read = 0.1
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]

From ebcc9fb52648ea64f82de6b5308c0d523b6e1a6e Mon Sep 17 00:00:00 2001
From: Seb Duerr <sebastian.duerr@cerebras.net>
Date: Sun, 6 Sep 2026 09:43:13 -0700
Subject: [PATCH 0768/2012] chore(cerebras): replace Gemma 4 with Qwen 3.8
 (#6199)

* feat(cerebras): add Qwen 3.8 27B

* chore(cerebras): remove Gemma 4
---
 providers/cerebras/models/gemma-4-31b.toml  | 17 -----------------
 providers/cerebras/models/qwen-3.8-27b.toml | 17 +++++++++++++++++
 2 files changed, 17 insertions(+), 17 deletions(-)
 delete mode 100644 providers/cerebras/models/gemma-4-31b.toml
 create mode 100644 providers/cerebras/models/qwen-3.8-27b.toml

diff --git a/providers/cerebras/models/gemma-4-31b.toml b/providers/cerebras/models/gemma-4-31b.toml
deleted file mode 100644
index 177b9021fe6..00000000000
--- a/providers/cerebras/models/gemma-4-31b.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "google/gemma-4-31b-it"
-last_updated = "2026-07-01"
-status = "beta"
-# Model-specific reasoning HTTP values (accessed 2026-07-01):
-# reasoning_effort = "none" (default) | "low"|"medium"|"high" (enable reasoning).
-# Sources:
-# https://inference-docs.cerebras.ai/capabilities/reasoning#gemma-4-31b-reasoning_effort
-# https://inference-docs.cerebras.ai/models/gemma-4-31b
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
-
-[cost]
-input = 0.99
-output = 1.49
-
-[limit]
-context = 131_072
-output = 40_960
diff --git a/providers/cerebras/models/qwen-3.8-27b.toml b/providers/cerebras/models/qwen-3.8-27b.toml
new file mode 100644
index 00000000000..17a36db9d02
--- /dev/null
+++ b/providers/cerebras/models/qwen-3.8-27b.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-27b"
+last_updated = "2026-09-03"
+# Model-specific reasoning HTTP values from the Cerebras launch contract:
+# reasoning_effort = "none" | "low" | "medium" | "high"; default is "high".
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[cost]
+input = 0.99
+output = 1.49
+
+[limit]
+context = 65_536
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 643ece828087e127cdd81b5063c71909d02d73e4 Mon Sep 17 00:00:00 2001
From: far-ouq <125839498+far-ouq@users.noreply.github.com>
Date: Sun, 6 Sep 2026 12:43:44 -0400
Subject: [PATCH 0769/2012] fix(gitlab): add reasoning effort options (#6276)

---
 providers/gitlab/models/duo-chat-gpt-5-6-luna.toml |  2 +-
 providers/gitlab/models/duo-chat-gpt-5-6-sol.toml  |  2 +-
 .../gitlab/models/duo-chat-gpt-5-6-terra.toml      |  2 +-
 providers/gitlab/models/duo-chat-opus-4-8.toml     |  2 +-
 providers/gitlab/provider.toml                     | 14 ++++++--------
 5 files changed, 10 insertions(+), 12 deletions(-)

diff --git a/providers/gitlab/models/duo-chat-gpt-5-6-luna.toml b/providers/gitlab/models/duo-chat-gpt-5-6-luna.toml
index 8c5a15f1803..7a4793159c4 100644
--- a/providers/gitlab/models/duo-chat-gpt-5-6-luna.toml
+++ b/providers/gitlab/models/duo-chat-gpt-5-6-luna.toml
@@ -1,6 +1,6 @@
 base_model = "openai/gpt-5.6-luna"
 name = "Agentic Chat (GPT-5.6 Luna)"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 0
diff --git a/providers/gitlab/models/duo-chat-gpt-5-6-sol.toml b/providers/gitlab/models/duo-chat-gpt-5-6-sol.toml
index a7e940cad4b..98fbc75111f 100644
--- a/providers/gitlab/models/duo-chat-gpt-5-6-sol.toml
+++ b/providers/gitlab/models/duo-chat-gpt-5-6-sol.toml
@@ -1,6 +1,6 @@
 base_model = "openai/gpt-5.6-sol"
 name = "Agentic Chat (GPT-5.6 Sol)"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 0
diff --git a/providers/gitlab/models/duo-chat-gpt-5-6-terra.toml b/providers/gitlab/models/duo-chat-gpt-5-6-terra.toml
index 7dc3cceb9c1..14e0e617968 100644
--- a/providers/gitlab/models/duo-chat-gpt-5-6-terra.toml
+++ b/providers/gitlab/models/duo-chat-gpt-5-6-terra.toml
@@ -1,6 +1,6 @@
 base_model = "openai/gpt-5.6-terra"
 name = "Agentic Chat (GPT-5.6 Terra)"
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 0
diff --git a/providers/gitlab/models/duo-chat-opus-4-8.toml b/providers/gitlab/models/duo-chat-opus-4-8.toml
index 6855a9124e2..bc0ab16ca76 100644
--- a/providers/gitlab/models/duo-chat-opus-4-8.toml
+++ b/providers/gitlab/models/duo-chat-opus-4-8.toml
@@ -5,7 +5,7 @@ release_date = "2026-05-28"
 last_updated = "2026-05-28"
 attachment = true
 reasoning = true
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 temperature = false
 tool_call = true
 knowledge = "2026-01-31"
diff --git a/providers/gitlab/provider.toml b/providers/gitlab/provider.toml
index d9c42b94a22..fda87c1b1f0 100644
--- a/providers/gitlab/provider.toml
+++ b/providers/gitlab/provider.toml
@@ -1,7 +1,4 @@
-name = "GitLab Duo"
-env = ["GITLAB_TOKEN"]
-npm = "gitlab-ai-provider"
-# Raw HTTP reasoning controls (sources accessed 2026-06-25):
+# Reasoning controls (sources accessed 2026-09-04):
 # - OpenAI Chat: POST /ai/v1/proxy/openai/v1/chat/completions uses top-level
 #   `reasoning_effort`; Responses: POST /ai/v1/proxy/openai/v1/responses uses
 #   `reasoning = { effort = "..." }`.
@@ -11,11 +8,12 @@ npm = "gitlab-ai-provider"
 #   `thinking = { type = "enabled", budget_tokens = N }` or
 #   `thinking = { type = "disabled" }`; effort is `output_config.effort`.
 #   https://docs.anthropic.com/en/api/messages
-# The npm provider obtains a direct-access token, points the native SDKs at
-# those proxy bases, but builds fixed request bodies without reasoning fields;
-# it does not expose a reasoning passthrough. These raw shapes therefore do not
-# make `reasoning_options` available through this configured npm integration.
+# gitlab-ai-provider 6.14.0 exposes OpenAI `reasoningEffort` and Anthropic
+# `thinking` options through model defaults and per-call provider options.
 # https://gitlab.com/vglafirov/gitlab-ai-provider/-/blob/main/src/gitlab-direct-access.ts
 # https://gitlab.com/vglafirov/gitlab-ai-provider/-/blob/main/src/gitlab-openai-language-model.ts
 # https://gitlab.com/vglafirov/gitlab-ai-provider/-/blob/main/src/gitlab-anthropic-language-model.ts
+name = "GitLab Duo"
+env = ["GITLAB_TOKEN"]
+npm = "gitlab-ai-provider"
 doc = "https://docs.gitlab.com/user/duo_agent_platform/"

From 828f0f835679d4090f2f24109867a644b902856f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 11:44:35 -0500
Subject: [PATCH 0770/2012] fix(azure): remove stale beta flags for GA models
 (#6428)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/azure-cognitive-services/models/claude-sonnet-5.toml | 1 -
 providers/azure-cognitive-services/models/gpt-5.6-luna.toml    | 1 -
 providers/azure-cognitive-services/models/gpt-5.6-sol.toml     | 1 -
 providers/azure-cognitive-services/models/gpt-5.6-terra.toml   | 1 -
 providers/azure/models/claude-sonnet-5.toml                    | 1 -
 providers/azure/models/gpt-5.6-luna.toml                       | 1 -
 providers/azure/models/gpt-5.6-sol.toml                        | 1 -
 providers/azure/models/gpt-5.6-terra.toml                      | 1 -
 providers/azure/models/gpt-6-astra.toml                        | 1 -
 providers/azure/models/grok-4-1-fast-non-reasoning.toml        | 3 +--
 providers/azure/models/grok-4-1-fast-reasoning.toml            | 1 -
 11 files changed, 1 insertion(+), 12 deletions(-)

diff --git a/providers/azure-cognitive-services/models/claude-sonnet-5.toml b/providers/azure-cognitive-services/models/claude-sonnet-5.toml
index 4506be12f1c..bb31c7eac7d 100644
--- a/providers/azure-cognitive-services/models/claude-sonnet-5.toml
+++ b/providers/azure-cognitive-services/models/claude-sonnet-5.toml
@@ -5,7 +5,6 @@
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 structured_output = true
-status = "beta"
 
 [cost]
 input = 2.00
diff --git a/providers/azure-cognitive-services/models/gpt-5.6-luna.toml b/providers/azure-cognitive-services/models/gpt-5.6-luna.toml
index 963997ecc3b..c92a9cacd1b 100644
--- a/providers/azure-cognitive-services/models/gpt-5.6-luna.toml
+++ b/providers/azure-cognitive-services/models/gpt-5.6-luna.toml
@@ -1,6 +1,5 @@
 # Pricing aligned with Azure (no Terra/Luna cut from 2026-07-30): https://learn.microsoft.com/en-us/answers/questions/5962521/pricing-of-openai-gpt-5-6-models-luna-terra
 base_model = "openai/gpt-5.6-luna"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure-cognitive-services/models/gpt-5.6-sol.toml b/providers/azure-cognitive-services/models/gpt-5.6-sol.toml
index e681a386bfc..50c91c8e91e 100644
--- a/providers/azure-cognitive-services/models/gpt-5.6-sol.toml
+++ b/providers/azure-cognitive-services/models/gpt-5.6-sol.toml
@@ -1,5 +1,4 @@
 base_model = "openai/gpt-5.6-sol"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure-cognitive-services/models/gpt-5.6-terra.toml b/providers/azure-cognitive-services/models/gpt-5.6-terra.toml
index eaec53d6dc4..c0a3f062b04 100644
--- a/providers/azure-cognitive-services/models/gpt-5.6-terra.toml
+++ b/providers/azure-cognitive-services/models/gpt-5.6-terra.toml
@@ -1,6 +1,5 @@
 # Pricing aligned with Azure (no Terra/Luna cut from 2026-07-30): https://learn.microsoft.com/en-us/answers/questions/5962521/pricing-of-openai-gpt-5-6-models-luna-terra
 base_model = "openai/gpt-5.6-terra"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure/models/claude-sonnet-5.toml b/providers/azure/models/claude-sonnet-5.toml
index b946d285cad..0e027e6ef69 100644
--- a/providers/azure/models/claude-sonnet-5.toml
+++ b/providers/azure/models/claude-sonnet-5.toml
@@ -5,7 +5,6 @@
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 structured_output = true
-status = "beta"
 
 [cost]
 input = 2.00
diff --git a/providers/azure/models/gpt-5.6-luna.toml b/providers/azure/models/gpt-5.6-luna.toml
index aebac5426cb..fdbaf22e36e 100644
--- a/providers/azure/models/gpt-5.6-luna.toml
+++ b/providers/azure/models/gpt-5.6-luna.toml
@@ -1,6 +1,5 @@
 # Pricing: https://developers.openai.com/api/docs/pricing
 base_model = "openai/gpt-5.6-luna"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure/models/gpt-5.6-sol.toml b/providers/azure/models/gpt-5.6-sol.toml
index 2eb13824b76..7d37256b5ec 100644
--- a/providers/azure/models/gpt-5.6-sol.toml
+++ b/providers/azure/models/gpt-5.6-sol.toml
@@ -3,7 +3,6 @@
 # $4 input / $20 output promotion: 2026-09-01 through at least 2026-11-30.
 # Keep published cache and long-context rates; the announcement does not specify discounts for them.
 base_model = "openai/gpt-5.6-sol"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure/models/gpt-5.6-terra.toml b/providers/azure/models/gpt-5.6-terra.toml
index bfbe29ccd4a..4ee1ef141a7 100644
--- a/providers/azure/models/gpt-5.6-terra.toml
+++ b/providers/azure/models/gpt-5.6-terra.toml
@@ -1,6 +1,5 @@
 # Pricing: https://developers.openai.com/api/docs/pricing
 base_model = "openai/gpt-5.6-terra"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure/models/gpt-6-astra.toml b/providers/azure/models/gpt-6-astra.toml
index ad77bce32d7..fec9a699c7a 100644
--- a/providers/azure/models/gpt-6-astra.toml
+++ b/providers/azure/models/gpt-6-astra.toml
@@ -1,6 +1,5 @@
 # Azure pricing page publishes GPT-6 pricing in the blog rather than the service pricing page; https://azure.microsoft.com/en-us/blog/gpt-6-astra-frontier-intelligence-for-work-now-generally-available-in-microsoft-foundry/
 base_model = "openai/gpt-6-astra"
-status = "beta"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
diff --git a/providers/azure/models/grok-4-1-fast-non-reasoning.toml b/providers/azure/models/grok-4-1-fast-non-reasoning.toml
index 0e50d185531..8aac9716a00 100644
--- a/providers/azure/models/grok-4-1-fast-non-reasoning.toml
+++ b/providers/azure/models/grok-4-1-fast-non-reasoning.toml
@@ -8,7 +8,6 @@ reasoning = false
 temperature = true
 tool_call = true
 open_weights = false
-status = "beta"
 
 [cost]
 input = 0.20
@@ -22,4 +21,4 @@ output = 8_192
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
+output = ["text"]
diff --git a/providers/azure/models/grok-4-1-fast-reasoning.toml b/providers/azure/models/grok-4-1-fast-reasoning.toml
index 3914e2ca803..f5eb65b6207 100644
--- a/providers/azure/models/grok-4-1-fast-reasoning.toml
+++ b/providers/azure/models/grok-4-1-fast-reasoning.toml
@@ -9,7 +9,6 @@ reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false
-status = "beta"
 
 [cost]
 input = 0.20

From e7afb6dfec9a06fe69db0e907cfcef4417cdd987 Mon Sep 17 00:00:00 2001
From: Rubin Simons <me@rubin55.org>
Date: Sun, 6 Sep 2026 18:46:05 +0200
Subject: [PATCH 0771/2012] Fix name duplicates for Eden AI (#6387)

* fix(edenai): name latest aliases distinctly from their targets

Eden AI "latest" aliases inherit the versioned model's name via
base_model, so UIs that key on name (like /models in opencode) show
duplicate entries. This change to the Eden AI sync provider names them
as (for example) "Claude Fable Latest (Claude Fable 5.1)". Leave case-
only alias_of IDs unchanged.

* fix(edenai): distinguish latest aliases and host display names

Eden AI "latest" aliases inherit the versioned model's name via
base_model, and the same lab model on Vertex, Deep Infra, Groq, and
other hosts shares that name too. Name "latest" aliases as "Claude Fable
Latest (Claude Fable 5.1)", put extra labels in one parenthetical
("GPT OSS 120B (Deep Infra)", "Gemini 3.8 Flash (Vertex AI, US)"),
and leave the lab's own API unsuffixed.

* chore(sync): distinguish Eden AI latest alias and host display names

* fix(google): mark Nano Banana preview models in their names

gemini-3.1-flash-image-preview and gemini-3-pro-image-preview shared
the GA names Nano Banana 2 and Nano Banana Pro. Name them Nano Banana
2 Preview and Nano Banana Pro Preview so hosts that inherit via
base_model stay distinct.
---
 models/google/gemini-3-pro-image-preview.toml |   2 +-
 .../gemini-3.1-flash-image-preview.toml       |   2 +-
 packages/core/src/sync/providers/edenai.ts    |  72 ++++++++++-
 packages/core/test/sync.test.ts               | 113 +++++++++++++++++-
 .../amazon/moonshot.kimi-k2-thinking.toml     |   1 +
 .../models/amazon/moonshotai.kimi-k2.5.toml   |   1 +
 .../models/amazon/zai.glm-4.7-flash.toml      |   1 +
 .../models/amazon/zai.glm-4.7-flash@us.toml   |   2 +-
 .../models/anthropic/claude-fable-latest.toml |   1 +
 .../models/anthropic/claude-opus-latest.toml  |   1 +
 .../anthropic/claude-sonnet-latest.toml       |   1 +
 .../models/azure/gpt-5.1-codex-max.toml       |   1 +
 .../models/azure/gpt-5.1-codex-mini.toml      |   1 +
 .../edenai/models/azure/gpt-5.1-codex.toml    |   1 +
 .../edenai/models/azure/gpt-5.2-codex.toml    |   1 +
 .../edenai/models/cerebras/gpt-oss-120b.toml  |   1 +
 .../aisingapore/gemma-sea-lion-v4-27b-it.toml |   1 +
 .../deepseek-ai/deepseek-v4-flash-0731.toml   |   1 +
 .../@cf/deepseek-ai/deepseek-v4-pro-0813.toml |   1 +
 .../cloudflare/@cf/meta/llama-guard-3-8b.toml |   1 +
 .../cloudflare/@cf/openai/gpt-oss-120b.toml   |   1 +
 .../cloudflare/@cf/openai/gpt-oss-20b.toml    |   1 +
 .../@cf/qwen/qwen2.5-coder-32b-instruct.toml  |   1 +
 .../cloudflare/@cf/zai-org/glm-4.7-flash.toml |   1 +
 .../databricks-deepseek-v4-flash-0731.toml    |   1 +
 .../databricks-deepseek-v4-pro-0813.toml      |   1 +
 .../databricks/databricks-gpt-oss-120b.toml   |   1 +
 .../databricks-gpt-oss-120b@eu.toml           |   2 +-
 .../databricks/databricks-gpt-oss-20b.toml    |   1 +
 .../databricks/databricks-gpt-oss-20b@eu.toml |   2 +-
 .../deepinfra/ByteDance/Seed-2.0-code.toml    |   1 +
 .../deepinfra/ByteDance/Seed-2.0-mini.toml    |   1 +
 .../deepinfra/deepseek-ai/DeepSeek-R1.toml    |   1 +
 .../deepseek-ai/DeepSeek-V3-0324.toml         |   1 +
 .../deepinfra/deepseek-ai/DeepSeek-V3.toml    |   1 +
 .../deepseek-ai/DeepSeek-V4-Flash-0731.toml   |   1 +
 .../deepseek-ai/DeepSeek-V4-Pro-0813.toml     |   1 +
 .../Llama-3.2-11B-Vision-Instruct.toml        |   1 +
 .../meta-llama/Llama-3.3-70B-Instruct.toml    |   1 +
 .../meta-llama/Llama-Guard-3-8B.toml          |   1 +
 .../meta-models/Muse-Glimmer-30B.toml         |   1 +
 .../deepinfra/moonshotai/Kimi-K2.5.toml       |   1 +
 .../deepinfra/nemotron-3-ultra-550b-a55b.toml |   1 +
 .../Llama-3.1-Nemotron-70B-Instruct.toml      |   1 +
 .../nvidia/Nemotron-3-Nano-30B-A3B.toml       |   1 +
 .../models/deepinfra/openai/gpt-oss-120b.toml |   1 +
 .../models/deepinfra/openai/gpt-oss-20b.toml  |   1 +
 .../deepinfra/stepfun-ai/Step-3.5-Flash.toml  |   1 +
 .../deepinfra/stepfun-ai/Step-3.7-Flash.toml  |   1 +
 .../edenai/models/deepinfra/tencent/Hy3.toml  |   1 +
 .../thinkingmachines/Inkling-Small.toml       |   1 +
 .../deepinfra/thinkingmachines/Inkling.toml   |   1 +
 .../deepinfra/zai-org/GLM-4.7-Flash.toml      |   1 +
 .../models/deepseek-v4-flash-0731.toml        |   1 +
 .../models/deepseek-v4-pro-0813.toml          |   1 +
 .../accounts/fireworks/models/inkling.toml    |   1 +
 .../fireworks/models/muse-glimmer-30b.toml    |   1 +
 .../models/fireworks_ai/gpt-oss-120b.toml     |   1 +
 .../models/flexai/Muse-Glimmer-30B.toml       |   1 +
 .../flexai/Nemotron-3-Super-120B-A12B.toml    |   1 +
 .../edenai/models/flexai/Step-3.7-Flash.toml  |   1 +
 .../models/flexai/deepseek-v4-flash-0731.toml |   1 +
 .../edenai/models/flexai/gpt-oss-120b.toml    |   1 +
 .../edenai/models/flexai/gpt-oss-20b.toml     |   1 +
 .../models/google/gemini-flash-latest.toml    |   1 +
 .../models/google/gemini-pro-latest.toml      |   1 +
 .../models/groq/openai/gpt-oss-120b.toml      |   1 +
 .../models/groq/openai/gpt-oss-20b.toml       |   1 +
 .../meta-llama/Llama-3.3-70B-Instruct.toml    |   1 +
 .../models/ionos/openai/gpt-oss-120b.toml     |   1 +
 .../deepseek-ai/DeepSeek-V4-Flash-0731.toml   |   1 +
 .../meta-llama/Llama-3.3-70B-Instruct.toml    |   1 +
 .../nvidia/Nemotron-3-Ultra-550b-a55b.toml    |   1 +
 .../nvidia/nemotron-3-super-120b-a12b.toml    |   1 +
 .../models/nebius/openai/gpt-oss-120b.toml    |   1 +
 .../edenai/models/openai/gpt-latest.toml      |   1 +
 .../edenai/models/openai/gpt-mini-latest.toml |   1 +
 .../edenai/models/openai/gpt-pro-latest.toml  |   1 +
 .../edenai/models/ovhcloud/gpt-oss-120b.toml  |   1 +
 .../edenai/models/ovhcloud/gpt-oss-20b.toml   |   1 +
 .../models/qwen/deepseek-v4-flash-0731.toml   |   1 +
 .../models/qwen/deepseek-v4-pro-0813.toml     |   1 +
 .../scaleway/deepseek-v4-flash-0731.toml      |   1 +
 .../edenai/models/scaleway/gpt-oss-120b.toml  |   1 +
 .../scaleway/llama-3.3-70b-instruct.toml      |   1 +
 .../deepseek/deepseek-v4-flash-0731.toml      |   1 +
 .../deepseek/deepseek-v4-pro-0813.toml        |   1 +
 .../models/tensorx/moonshotai/kimi-k2.5.toml  |   1 +
 .../deepseek-ai/DeepSeek-V4-Flash-0731.toml   |   1 +
 .../deepseek-ai/DeepSeek-V4-Pro-0813.toml     |   1 +
 .../meta-models/Muse-Glimmer-30B.toml         |   1 +
 .../together_ai/openai/gpt-oss-120b.toml      |   1 +
 .../together_ai/openai/gpt-oss-20b.toml       |   1 +
 .../thinkingmachines/Inkling-Small.toml       |   1 +
 .../together_ai/thinkingmachines/Inkling.toml |   1 +
 .../models/vertex/gemini-2.5-flash-image.toml |   1 +
 .../models/vertex/gemini-3-flash-preview.toml |   1 +
 .../models/vertex/gemini-3-pro-image.toml     |   1 +
 .../models/vertex/gemini-3.1-flash-image.toml |   1 +
 .../vertex/gemini-3.1-flash-lite-image.toml   |   1 +
 .../models/vertex/gemini-3.1-flash-lite.toml  |   1 +
 .../vertex/gemini-3.1-flash-lite@eu.toml      |   2 +-
 .../vertex/gemini-3.1-flash-lite@us.toml      |   2 +-
 .../models/vertex/gemini-3.1-pro-preview.toml |   1 +
 .../models/vertex/gemini-3.5-flash-lite.toml  |   1 +
 .../vertex/gemini-3.5-flash-lite@eu.toml      |   2 +-
 .../vertex/gemini-3.5-flash-lite@us.toml      |   2 +-
 .../models/vertex/gemini-3.5-flash.toml       |   1 +
 .../models/vertex/gemini-3.5-flash@eu.toml    |   2 +-
 .../models/vertex/gemini-3.5-flash@us.toml    |   2 +-
 .../models/vertex/gemini-3.6-flash.toml       |   1 +
 .../models/vertex/gemini-3.6-flash@eu.toml    |   2 +-
 .../models/vertex/gemini-3.6-flash@us.toml    |   2 +-
 .../models/vertex/gemini-3.7-flash.toml       |   1 +
 .../models/vertex/gemini-3.7-flash@eu.toml    |   2 +-
 .../models/vertex/gemini-3.7-flash@us.toml    |   2 +-
 .../models/vertex/gemini-3.8-flash.toml       |   1 +
 .../models/vertex/gemini-3.8-flash@eu.toml    |   2 +-
 .../models/vertex/gemini-3.8-flash@us.toml    |   2 +-
 .../models/vertex/gemini-flash-latest.toml    |   1 +
 .../models/vertex/gemini-pro-latest.toml      |   1 +
 providers/edenai/models/xai/grok-latest.toml  |   1 +
 .../gemini-3.1-flash-image-preview.toml       |   2 +-
 sync.md                                       |   2 +
 124 files changed, 303 insertions(+), 23 deletions(-)

diff --git a/models/google/gemini-3-pro-image-preview.toml b/models/google/gemini-3-pro-image-preview.toml
index 664595945fb..f10fda90081 100644
--- a/models/google/gemini-3-pro-image-preview.toml
+++ b/models/google/gemini-3-pro-image-preview.toml
@@ -1,4 +1,4 @@
-name = "Nano Banana Pro"
+name = "Nano Banana Pro Preview"
 description = "Nano Banana Pro for higher-fidelity image generation and design-heavy edits"
 family = "gemini-pro"
 release_date = "2025-11-20"
diff --git a/models/google/gemini-3.1-flash-image-preview.toml b/models/google/gemini-3.1-flash-image-preview.toml
index 725a26006dd..9e0c07cd3b6 100644
--- a/models/google/gemini-3.1-flash-image-preview.toml
+++ b/models/google/gemini-3.1-flash-image-preview.toml
@@ -1,4 +1,4 @@
-name = "Nano Banana 2"
+name = "Nano Banana 2 Preview"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
 family = "gemini-flash"
 release_date = "2026-02-26"
diff --git a/packages/core/src/sync/providers/edenai.ts b/packages/core/src/sync/providers/edenai.ts
index bedd8be7743..dff562fbdf4 100644
--- a/packages/core/src/sync/providers/edenai.ts
+++ b/packages/core/src/sync/providers/edenai.ts
@@ -59,6 +59,8 @@ const MODALITY_BY_EDENAI: Record<
 };
 
 // Upstreams that are the lab's own API for models under that namespace.
+// The first entry is the unsuffixed display route when several first-party
+// hosts exist (Google AI Studio vs Vertex AI).
 const LAB_UPSTREAMS: Record<string, readonly string[]> = {
   alibaba: ["qwen"],
   amazon: ["amazon"],
@@ -76,6 +78,29 @@ const LAB_UPSTREAMS: Record<string, readonly string[]> = {
   zhipuai: ["zai"],
 };
 
+const ROUTE_LABELS: Record<string, string> = {
+  amazon: "Amazon Bedrock",
+  azure: "Azure",
+  cerebras: "Cerebras",
+  cloudflare: "Cloudflare",
+  compactifai: "CompactifAI",
+  databricks: "Databricks",
+  deepinfra: "Deep Infra",
+  fireworks_ai: "Fireworks AI",
+  flexai: "FlexAI",
+  groq: "Groq",
+  infomaniak: "Infomaniak",
+  ionos: "IONOS",
+  lilac: "Lilac",
+  nebius: "Nebius",
+  ovhcloud: "OVHcloud",
+  qwen: "Alibaba",
+  scaleway: "Scaleway",
+  tensorx: "TensorX",
+  together_ai: "Together AI",
+  vertex: "Vertex AI",
+};
+
 type ReasoningOption = NonNullable<
   SyncedFullModel["reasoning_options"]
 >[number];
@@ -217,13 +242,52 @@ function hasOutputLimit(baseModel: string) {
   );
 }
 
-function regionVariantName(model: EdenAIModel, baseModel: string) {
+function titleCaseSlug(slug: string) {
+  return slug
+    .split(/[-_]/)
+    .filter((word) => word.length > 0)
+    .map((word) =>
+      word.toLowerCase() === "gpt"
+        ? "GPT"
+        : word[0]!.toUpperCase() + word.slice(1).toLowerCase(),
+    )
+    .join(" ");
+}
+
+function isLatestAlias(model: EdenAIModel) {
+  if (model.alias_of == null) return false;
+  const id = model.id.replace(REGION_SUFFIX, "");
+  const target = model.alias_of.replace(REGION_SUFFIX, "");
+  if (id.toLowerCase() === target.toLowerCase()) return false;
+  const slug = id.split("/").at(-1) ?? id;
+  return /(?:^|-)latest$/i.test(slug);
+}
+
+function routeLabel(model: EdenAIModel, baseModel: string) {
+  const lab = baseModel.split("/")[0] ?? "";
+  const primary = LAB_UPSTREAMS[lab]?.[0];
+  if (model.owned_by === primary) return undefined;
+  return ROUTE_LABELS[model.owned_by] ?? titleCaseSlug(model.owned_by);
+}
+
+function displayName(model: EdenAIModel, baseModel: string) {
   const region = REGION_SUFFIX.exec(model.id)?.[0].slice(1);
-  if (region === undefined) return undefined;
+  const latest = isLatestAlias(model);
+  const route = routeLabel(model, baseModel);
+  if (region === undefined && !latest && route === undefined) return undefined;
 
   const canonical = canonicalModelName(baseModel);
   if (canonical === undefined) return undefined;
-  return `${canonical} (${region.toUpperCase()})`;
+
+  const head = latest
+    ? titleCaseSlug(model.id.replace(REGION_SUFFIX, "").split("/").at(-1) ?? "")
+    : canonical;
+  const details = [
+    ...(latest ? [canonical] : []),
+    ...(route !== undefined ? [route] : []),
+    ...(region !== undefined ? [region.toUpperCase()] : []),
+  ];
+  return `${head} (${details.join(", ")})`;
 }
 
 // ========================================
@@ -462,7 +526,7 @@ export function buildEdenAIModel(
   return factorBaseModel(
     baseModel,
     {
-      name: regionVariantName(model, baseModel),
+      name: displayName(model, baseModel),
       modalities,
       attachment: input?.some((value) => value !== "text"),
       reasoning_options: reasoningOptions,
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index c190f650eac..0e51e85cfe1 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2613,7 +2613,112 @@ test("names Eden AI regional deployments after the canonical model", () => {
     ),
   ).toMatchObject({
     base_model: "anthropic/claude-opus-5",
-    name: "Claude Opus 5 (EU)",
+    name: "Claude Opus 5 (Amazon Bedrock, EU)",
+  });
+});
+
+test("names Eden AI latest aliases as Latest plus the current target", () => {
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "anthropic/claude-fable-latest",
+        model_name: "claude-fable-5-1",
+        owned_by: "anthropic",
+        alias_of: "anthropic/claude-fable-5-1",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "anthropic/claude-fable-5-1",
+    name: "Claude Fable Latest (Claude Fable 5.1)",
+  });
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "openai/gpt-latest",
+        model_name: "gpt-6-astra",
+        owned_by: "openai",
+        alias_of: "openai/gpt-6-astra",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "openai/gpt-6-astra",
+    name: "GPT Latest (GPT-6 Astra)",
+  });
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "vertex/gemini-flash-latest@us",
+        model_name: "gemini-3.8-flash",
+        owned_by: "vertex",
+        alias_of: "vertex/gemini-3.8-flash",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "google/gemini-3.8-flash",
+    name: "Gemini Flash Latest (Gemini 3.8 Flash, Vertex AI, US)",
+  });
+});
+
+test("names Eden AI non-primary hosts distinctly from the lab route", () => {
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "google/gemini-3.8-flash",
+        model_name: "gemini-3.8-flash",
+        owned_by: "google",
+      }),
+    ),
+  ).not.toHaveProperty("name");
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "vertex/gemini-3.8-flash",
+        model_name: "gemini-3.8-flash",
+        owned_by: "vertex",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "google/gemini-3.8-flash",
+    name: "Gemini 3.8 Flash (Vertex AI)",
+  });
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "vertex/gemini-3.8-flash@us",
+        model_name: "gemini-3.8-flash",
+        owned_by: "vertex",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "google/gemini-3.8-flash",
+    name: "Gemini 3.8 Flash (Vertex AI, US)",
+  });
+  expect(
+    buildEdenAIModel(
+      edenAIModel({
+        id: "deepinfra/openai/gpt-oss-120b",
+        model_name: "openai/gpt-oss-120b",
+        owned_by: "deepinfra",
+      }),
+    ),
+  ).toMatchObject({
+    base_model: "openai/gpt-oss-120b",
+    name: "GPT OSS 120B (Deep Infra)",
+  });
+});
+
+test("does not treat Eden AI case-only aliases as latest pointers", () => {
+  const built = buildEdenAIModel(
+    edenAIModel({
+      id: "flexai/deepseek-v4-flash-0731",
+      model_name: "DeepSeek-V4-Flash-0731",
+      owned_by: "flexai",
+      alias_of: "flexai/DeepSeek-V4-Flash-0731",
+    }),
+  );
+  expect(built).toMatchObject({
+    base_model: "deepseek/deepseek-v4-flash-0731",
+    name: "DeepSeek V4 Flash 0731 (FlexAI)",
   });
 });
 
@@ -2675,9 +2780,15 @@ test("keeps every Eden AI route for models with no first-party relay", () => {
 
   const firstParty = collectFirstPartyBaseModels(models);
   expect(firstParty.size).toBe(0);
+  const names = {
+    deepinfra: "GPT OSS 120B (Deep Infra)",
+    groq: "GPT OSS 120B (Groq)",
+    cerebras: "GPT OSS 120B (Cerebras)",
+  };
   for (const model of models) {
     expect(buildEdenAIModel(model, undefined, firstParty)).toMatchObject({
       base_model: "openai/gpt-oss-120b",
+      name: names[model.owned_by as keyof typeof names],
     });
   }
 });
diff --git a/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml b/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
index 0d40d801a1a..1fdc4e3a58c 100644
--- a/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
+++ b/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
@@ -1,4 +1,5 @@
 base_model = "moonshotai/kimi-k2-thinking"
+name = "Kimi K2 Thinking (Amazon Bedrock)"
 tool_call = false
 structured_output = false
 reasoning_options = []
diff --git a/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml b/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
index aa7b37f014a..45d5d0a17cb 100644
--- a/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
+++ b/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
@@ -1,4 +1,5 @@
 base_model = "moonshotai/kimi-k2.5"
+name = "Kimi K2.5 (Amazon Bedrock)"
 structured_output = false
 reasoning_options = []
 
diff --git a/providers/edenai/models/amazon/zai.glm-4.7-flash.toml b/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
index 3799e7654b7..2a31f0a6612 100644
--- a/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
+++ b/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
@@ -1,4 +1,5 @@
 base_model = "zhipuai/glm-4.7-flash"
+name = "GLM-4.7-Flash (Amazon Bedrock)"
 structured_output = false
 reasoning_options = []
 
diff --git a/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml b/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
index 284900db350..ec6e4df6db1 100644
--- a/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
+++ b/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
@@ -1,5 +1,5 @@
 base_model = "zhipuai/glm-4.7-flash"
-name = "GLM-4.7-Flash (US)"
+name = "GLM-4.7-Flash (Amazon Bedrock, US)"
 structured_output = false
 reasoning_options = []
 
diff --git a/providers/edenai/models/anthropic/claude-fable-latest.toml b/providers/edenai/models/anthropic/claude-fable-latest.toml
index 6f1a3fda3d9..fe892ae3296 100644
--- a/providers/edenai/models/anthropic/claude-fable-latest.toml
+++ b/providers/edenai/models/anthropic/claude-fable-latest.toml
@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-fable-5-1"
+name = "Claude Fable Latest (Claude Fable 5.1)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/anthropic/claude-opus-latest.toml b/providers/edenai/models/anthropic/claude-opus-latest.toml
index 0a58cfec23a..d336938f9f1 100644
--- a/providers/edenai/models/anthropic/claude-opus-latest.toml
+++ b/providers/edenai/models/anthropic/claude-opus-latest.toml
@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-5"
+name = "Claude Opus Latest (Claude Opus 5)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/anthropic/claude-sonnet-latest.toml b/providers/edenai/models/anthropic/claude-sonnet-latest.toml
index 9db87aa596e..84e5fce82ae 100644
--- a/providers/edenai/models/anthropic/claude-sonnet-latest.toml
+++ b/providers/edenai/models/anthropic/claude-sonnet-latest.toml
@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-sonnet-5"
+name = "Claude Sonnet Latest (Claude Sonnet 5)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/azure/gpt-5.1-codex-max.toml b/providers/edenai/models/azure/gpt-5.1-codex-max.toml
index ac6346aa6ac..4fd9416e551 100644
--- a/providers/edenai/models/azure/gpt-5.1-codex-max.toml
+++ b/providers/edenai/models/azure/gpt-5.1-codex-max.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.1-codex-max"
 base_model_omit = ["limit.input"]
+name = "GPT-5.1 Codex Max (Azure)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/azure/gpt-5.1-codex-mini.toml b/providers/edenai/models/azure/gpt-5.1-codex-mini.toml
index eed8b111d93..f039c4791b2 100644
--- a/providers/edenai/models/azure/gpt-5.1-codex-mini.toml
+++ b/providers/edenai/models/azure/gpt-5.1-codex-mini.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.1-codex-mini"
 base_model_omit = ["limit.input"]
+name = "GPT-5.1 Codex mini (Azure)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/azure/gpt-5.1-codex.toml b/providers/edenai/models/azure/gpt-5.1-codex.toml
index 89e3613f55e..aa4847da453 100644
--- a/providers/edenai/models/azure/gpt-5.1-codex.toml
+++ b/providers/edenai/models/azure/gpt-5.1-codex.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.1-codex"
 base_model_omit = ["limit.input"]
+name = "GPT-5.1 Codex (Azure)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/azure/gpt-5.2-codex.toml b/providers/edenai/models/azure/gpt-5.2-codex.toml
index cc88366f47e..054a25112fd 100644
--- a/providers/edenai/models/azure/gpt-5.2-codex.toml
+++ b/providers/edenai/models/azure/gpt-5.2-codex.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.2-codex"
 base_model_omit = ["limit.input"]
+name = "GPT-5.2 Codex (Azure)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/cerebras/gpt-oss-120b.toml b/providers/edenai/models/cerebras/gpt-oss-120b.toml
index fc6d3786a69..ec562201273 100644
--- a/providers/edenai/models/cerebras/gpt-oss-120b.toml
+++ b/providers/edenai/models/cerebras/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Cerebras)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/cloudflare/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml b/providers/edenai/models/cloudflare/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
index 06e14f09dab..d323efca4cb 100644
--- a/providers/edenai/models/cloudflare/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
+++ b/providers/edenai/models/cloudflare/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
@@ -1,4 +1,5 @@
 base_model = "aisingapore/gemma-sea-lion-v4-27b-it"
+name = "Gemma-SEA-LION-v4-27B-IT (Cloudflare)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-flash-0731.toml b/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
index caded09fb6b..f82d3e094da 100644
--- a/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Cloudflare)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-pro-0813.toml b/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
index 2bac9436ad1..707f926f6bd 100644
--- a/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/cloudflare/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Cloudflare)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/cloudflare/@cf/meta/llama-guard-3-8b.toml b/providers/edenai/models/cloudflare/@cf/meta/llama-guard-3-8b.toml
index b18940b04a4..c9be7ae7e00 100644
--- a/providers/edenai/models/cloudflare/@cf/meta/llama-guard-3-8b.toml
+++ b/providers/edenai/models/cloudflare/@cf/meta/llama-guard-3-8b.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-guard-3-8b"
+name = "Llama-Guard-3-8B (Cloudflare)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-120b.toml b/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-120b.toml
index 7e8312cd34c..f82bcaa155a 100644
--- a/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Cloudflare)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-20b.toml b/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-20b.toml
index c43fc3001c7..c0000eece82 100644
--- a/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-20b.toml
+++ b/providers/edenai/models/cloudflare/@cf/openai/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Cloudflare)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml b/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
index 93cb5d3bd2e..33e52d4c7a4 100644
--- a/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
+++ b/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
@@ -1,4 +1,5 @@
 base_model = "alibaba/qwen2.5-coder-32b-instruct"
+name = "Qwen2.5-Coder-32B-Instruct (Cloudflare)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/cloudflare/@cf/zai-org/glm-4.7-flash.toml b/providers/edenai/models/cloudflare/@cf/zai-org/glm-4.7-flash.toml
index 1abd0ea732a..a2adc939783 100644
--- a/providers/edenai/models/cloudflare/@cf/zai-org/glm-4.7-flash.toml
+++ b/providers/edenai/models/cloudflare/@cf/zai-org/glm-4.7-flash.toml
@@ -1,4 +1,5 @@
 base_model = "zhipuai/glm-4.7-flash"
+name = "GLM-4.7-Flash (Cloudflare)"
 structured_output = true
 reasoning_options = []
 
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
index baee0759016..d4d7d35c658 100644
--- a/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Databricks)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
index b5283d22349..a25d6b3e95b 100644
--- a/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Databricks)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml b/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
index ba07e71a376..d937f340ccb 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Databricks)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
index 37a595e6865..a6fc70753e5 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-120b@eu.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-oss-120b"
-name = "GPT OSS 120B (EU)"
+name = "GPT OSS 120B (Databricks, EU)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml b/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
index 4e3ee6c0c1b..88c8b5e2847 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Databricks)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
index 6cee3a77f8b..456e8fff78b 100644
--- a/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
+++ b/providers/edenai/models/databricks/databricks-gpt-oss-20b@eu.toml
@@ -1,5 +1,5 @@
 base_model = "openai/gpt-oss-20b"
-name = "GPT OSS 20B (EU)"
+name = "GPT OSS 20B (Databricks, EU)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
index 6a4e99b9ea7..bc15dd4e96c 100644
--- a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
+++ b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-code.toml
@@ -1,4 +1,5 @@
 base_model = "bytedance-seed/seed-2.0-code"
+name = "Seed 2.0 Code (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
index 9d7aa010e62..c50bdb655e0 100644
--- a/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
+++ b/providers/edenai/models/deepinfra/ByteDance/Seed-2.0-mini.toml
@@ -1,4 +1,5 @@
 base_model = "bytedance-seed/seed-2.0-mini"
+name = "Seed 2.0 Mini (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-R1.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-R1.toml
index 6ba25593e54..78d2e12defe 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-R1.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-R1.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-r1"
+name = "DeepSeek-R1 (Deep Infra)"
 structured_output = false
 reasoning_options = []
 
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3-0324.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3-0324.toml
index e18c6527ba1..b3f50f4e8da 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3-0324.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3-0324.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v3-0324"
+name = "DeepSeek V3 0324 (Deep Infra)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3.toml
index a6d680ebbe5..85b700dcad1 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V3.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v3"
+name = "DeepSeek-V3 (Deep Infra)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index ac478b62e04..3df69ac9ec4 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Pro-0813.toml
index 8e11ef85e8c..165791c08dd 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Pro-0813.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct.toml b/providers/edenai/models/deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct.toml
index 1bfc897a035..706cee36ebb 100644
--- a/providers/edenai/models/deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct.toml
+++ b/providers/edenai/models/deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.2-11b-vision-instruct"
+name = "Llama-3.2-11B-Vision-Instruct (Deep Infra)"
 tool_call = false
 structured_output = true
 
diff --git a/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
index 2638ae0d04c..b55af51c0bb 100644
--- a/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
+name = "Llama-3.3-70B-Instruct (Deep Infra)"
 attachment = false
 structured_output = true
 
diff --git a/providers/edenai/models/deepinfra/meta-llama/Llama-Guard-3-8B.toml b/providers/edenai/models/deepinfra/meta-llama/Llama-Guard-3-8B.toml
index 7e26572ec9d..57e863aac4e 100644
--- a/providers/edenai/models/deepinfra/meta-llama/Llama-Guard-3-8B.toml
+++ b/providers/edenai/models/deepinfra/meta-llama/Llama-Guard-3-8B.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-guard-3-8b"
+name = "Llama-Guard-3-8B (Deep Infra)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/deepinfra/meta-models/Muse-Glimmer-30B.toml b/providers/edenai/models/deepinfra/meta-models/Muse-Glimmer-30B.toml
index c01aa281174..02739958be0 100644
--- a/providers/edenai/models/deepinfra/meta-models/Muse-Glimmer-30B.toml
+++ b/providers/edenai/models/deepinfra/meta-models/Muse-Glimmer-30B.toml
@@ -1,4 +1,5 @@
 base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/moonshotai/Kimi-K2.5.toml b/providers/edenai/models/deepinfra/moonshotai/Kimi-K2.5.toml
index 56ab5b2b572..3495335028d 100644
--- a/providers/edenai/models/deepinfra/moonshotai/Kimi-K2.5.toml
+++ b/providers/edenai/models/deepinfra/moonshotai/Kimi-K2.5.toml
@@ -1,4 +1,5 @@
 base_model = "moonshotai/kimi-k2.5"
+name = "Kimi K2.5 (Deep Infra)"
 reasoning_options = []
 
 [cost]
diff --git a/providers/edenai/models/deepinfra/nemotron-3-ultra-550b-a55b.toml b/providers/edenai/models/deepinfra/nemotron-3-ultra-550b-a55b.toml
index db9ab214604..51ea14aca4b 100644
--- a/providers/edenai/models/deepinfra/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/edenai/models/deepinfra/nemotron-3-ultra-550b-a55b.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/nemotron-3-ultra-550b-a55b"
+name = "Nemotron 3 Ultra 550B A55B (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct.toml b/providers/edenai/models/deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct.toml
index 1b1be74a6c9..ac91c06b77d 100644
--- a/providers/edenai/models/deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct.toml
+++ b/providers/edenai/models/deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/llama-3.1-nemotron-70b-instruct"
+name = "Llama 3.1 Nemotron 70B Instruct (Deep Infra)"
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/deepinfra/nvidia/Nemotron-3-Nano-30B-A3B.toml b/providers/edenai/models/deepinfra/nvidia/Nemotron-3-Nano-30B-A3B.toml
index aaff49826e1..8b32282deec 100644
--- a/providers/edenai/models/deepinfra/nvidia/Nemotron-3-Nano-30B-A3B.toml
+++ b/providers/edenai/models/deepinfra/nvidia/Nemotron-3-Nano-30B-A3B.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/nemotron-3-nano-30b-a3b"
+name = "Nemotron 3 Nano 30B A3B (Deep Infra)"
 structured_output = true
 reasoning_options = []
 
diff --git a/providers/edenai/models/deepinfra/openai/gpt-oss-120b.toml b/providers/edenai/models/deepinfra/openai/gpt-oss-120b.toml
index e0ec3903a08..644ba82cfd3 100644
--- a/providers/edenai/models/deepinfra/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/deepinfra/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/openai/gpt-oss-20b.toml b/providers/edenai/models/deepinfra/openai/gpt-oss-20b.toml
index b37413d1850..18cba074ce6 100644
--- a/providers/edenai/models/deepinfra/openai/gpt-oss-20b.toml
+++ b/providers/edenai/models/deepinfra/openai/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Deep Infra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.5-Flash.toml b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.5-Flash.toml
index d11a94a99d9..d6ad2dc31a7 100644
--- a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.5-Flash.toml
+++ b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.5-Flash.toml
@@ -1,5 +1,6 @@
 base_model = "stepfun/step-3.5-flash"
 base_model_omit = ["limit.input"]
+name = "Step 3.5 Flash (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
index d395a094601..6cc4f58727b 100644
--- a/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
+++ b/providers/edenai/models/deepinfra/stepfun-ai/Step-3.7-Flash.toml
@@ -1,5 +1,6 @@
 base_model = "stepfun/step-3.7-flash"
 base_model_omit = ["limit.input"]
+name = "Step 3.7 Flash (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/tencent/Hy3.toml b/providers/edenai/models/deepinfra/tencent/Hy3.toml
index ae1d1975c90..c08482653b4 100644
--- a/providers/edenai/models/deepinfra/tencent/Hy3.toml
+++ b/providers/edenai/models/deepinfra/tencent/Hy3.toml
@@ -1,5 +1,6 @@
 base_model = "tencent/hy3"
 base_model_omit = ["limit.input"]
+name = "Hy3 (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/thinkingmachines/Inkling-Small.toml b/providers/edenai/models/deepinfra/thinkingmachines/Inkling-Small.toml
index a0470066d6b..aab2e4ab14e 100644
--- a/providers/edenai/models/deepinfra/thinkingmachines/Inkling-Small.toml
+++ b/providers/edenai/models/deepinfra/thinkingmachines/Inkling-Small.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling-small"
+name = "Inkling Small (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/thinkingmachines/Inkling.toml b/providers/edenai/models/deepinfra/thinkingmachines/Inkling.toml
index 5353ced4450..f73f8b80c7f 100644
--- a/providers/edenai/models/deepinfra/thinkingmachines/Inkling.toml
+++ b/providers/edenai/models/deepinfra/thinkingmachines/Inkling.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling"
+name = "Inkling (Deep Infra)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/deepinfra/zai-org/GLM-4.7-Flash.toml b/providers/edenai/models/deepinfra/zai-org/GLM-4.7-Flash.toml
index 51a1a790b52..a5eba99fa24 100644
--- a/providers/edenai/models/deepinfra/zai-org/GLM-4.7-Flash.toml
+++ b/providers/edenai/models/deepinfra/zai-org/GLM-4.7-Flash.toml
@@ -1,4 +1,5 @@
 base_model = "zhipuai/glm-4.7-flash"
+name = "GLM-4.7-Flash (Deep Infra)"
 structured_output = true
 reasoning_options = []
 
diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-flash-0731.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-flash-0731.toml
index 7f25a297a2b..43b37cf7cbd 100644
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Fireworks AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-pro-0813.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-pro-0813.toml
index 2bac9436ad1..2213244487c 100644
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/deepseek-v4-pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Fireworks AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml
index add2a32eee5..7989fb0fe01 100644
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml
+++ b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/inkling.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling"
+name = "Inkling (Fireworks AI)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/muse-glimmer-30b.toml b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/muse-glimmer-30b.toml
index 2812d8df497..d7924809f4a 100644
--- a/providers/edenai/models/fireworks_ai/accounts/fireworks/models/muse-glimmer-30b.toml
+++ b/providers/edenai/models/fireworks_ai/accounts/fireworks/models/muse-glimmer-30b.toml
@@ -1,4 +1,5 @@
 base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (Fireworks AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/fireworks_ai/gpt-oss-120b.toml b/providers/edenai/models/fireworks_ai/gpt-oss-120b.toml
index 8a5eae63d21..992ec3575a6 100644
--- a/providers/edenai/models/fireworks_ai/gpt-oss-120b.toml
+++ b/providers/edenai/models/fireworks_ai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Fireworks AI)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
index df75a46ae61..2585cc19725 100644
--- a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
+++ b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
@@ -1,4 +1,5 @@
 base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml b/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
index 3aa985d31f1..310bb953a9d 100644
--- a/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
+++ b/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/nemotron-3-super-120b-a12b"
+name = "Nemotron 3 Super 120B A12B (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/flexai/Step-3.7-Flash.toml b/providers/edenai/models/flexai/Step-3.7-Flash.toml
index fb725b7c60f..85009ab8b76 100644
--- a/providers/edenai/models/flexai/Step-3.7-Flash.toml
+++ b/providers/edenai/models/flexai/Step-3.7-Flash.toml
@@ -1,5 +1,6 @@
 base_model = "stepfun/step-3.7-flash"
 base_model_omit = ["limit.input"]
+name = "Step 3.7 Flash (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml b/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
index 7a18dfdfacb..e25ac9dd061 100644
--- a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/flexai/gpt-oss-120b.toml b/providers/edenai/models/flexai/gpt-oss-120b.toml
index c3cccca018f..74066689255 100644
--- a/providers/edenai/models/flexai/gpt-oss-120b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/flexai/gpt-oss-20b.toml b/providers/edenai/models/flexai/gpt-oss-20b.toml
index 64a0c7acc13..a9511f536f9 100644
--- a/providers/edenai/models/flexai/gpt-oss-20b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (FlexAI)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/google/gemini-flash-latest.toml b/providers/edenai/models/google/gemini-flash-latest.toml
index e1d0abd66e5..e3ccde409d2 100644
--- a/providers/edenai/models/google/gemini-flash-latest.toml
+++ b/providers/edenai/models/google/gemini-flash-latest.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.8-flash"
+name = "Gemini Flash Latest (Gemini 3.8 Flash)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/google/gemini-pro-latest.toml b/providers/edenai/models/google/gemini-pro-latest.toml
index 38907bc3ce0..ae9e93123b3 100644
--- a/providers/edenai/models/google/gemini-pro-latest.toml
+++ b/providers/edenai/models/google/gemini-pro-latest.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-pro-preview"
+name = "Gemini Pro Latest (Gemini 3.1 Pro Preview)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/groq/openai/gpt-oss-120b.toml b/providers/edenai/models/groq/openai/gpt-oss-120b.toml
index e66f11b7da6..bb02ded6a84 100644
--- a/providers/edenai/models/groq/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/groq/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Groq)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/groq/openai/gpt-oss-20b.toml b/providers/edenai/models/groq/openai/gpt-oss-20b.toml
index 7fb706235b9..255f5361da6 100644
--- a/providers/edenai/models/groq/openai/gpt-oss-20b.toml
+++ b/providers/edenai/models/groq/openai/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Groq)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 7bb385c666d..c8066cf5348 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
+name = "Llama-3.3-70B-Instruct (IONOS)"
 attachment = false
 tool_call = false
 structured_output = false
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 0079b03c739..2f0c218b10b 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (IONOS)"
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index 957b9dd7a3d..fa8a3e7e345 100644
--- a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Nebius)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
index 0881ce87b06..9485ca66aaf 100644
--- a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
+name = "Llama-3.3-70B-Instruct (Nebius)"
 attachment = false
 structured_output = true
 
diff --git a/providers/edenai/models/nebius/nvidia/Nemotron-3-Ultra-550b-a55b.toml b/providers/edenai/models/nebius/nvidia/Nemotron-3-Ultra-550b-a55b.toml
index 5ba56b54de3..f2436a7ba86 100644
--- a/providers/edenai/models/nebius/nvidia/Nemotron-3-Ultra-550b-a55b.toml
+++ b/providers/edenai/models/nebius/nvidia/Nemotron-3-Ultra-550b-a55b.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/nemotron-3-ultra-550b-a55b"
+name = "Nemotron 3 Ultra 550B A55B (Nebius)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/nebius/nvidia/nemotron-3-super-120b-a12b.toml b/providers/edenai/models/nebius/nvidia/nemotron-3-super-120b-a12b.toml
index c3806a4c850..5405a7eeccd 100644
--- a/providers/edenai/models/nebius/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/edenai/models/nebius/nvidia/nemotron-3-super-120b-a12b.toml
@@ -1,4 +1,5 @@
 base_model = "nvidia/nemotron-3-super-120b-a12b"
+name = "Nemotron 3 Super 120B A12B (Nebius)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/nebius/openai/gpt-oss-120b.toml b/providers/edenai/models/nebius/openai/gpt-oss-120b.toml
index a7f3229d340..f6045b89cc3 100644
--- a/providers/edenai/models/nebius/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/nebius/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Nebius)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/openai/gpt-latest.toml b/providers/edenai/models/openai/gpt-latest.toml
index d3cae944f01..70646ca0a1e 100644
--- a/providers/edenai/models/openai/gpt-latest.toml
+++ b/providers/edenai/models/openai/gpt-latest.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-6-astra"
+name = "GPT Latest (GPT-6 Astra)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/openai/gpt-mini-latest.toml b/providers/edenai/models/openai/gpt-mini-latest.toml
index b71b67ad193..b53e46a9e2d 100644
--- a/providers/edenai/models/openai/gpt-mini-latest.toml
+++ b/providers/edenai/models/openai/gpt-mini-latest.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.4-mini"
+name = "GPT Mini Latest (GPT-5.4 mini)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/openai/gpt-pro-latest.toml b/providers/edenai/models/openai/gpt-pro-latest.toml
index eca4c05cfd4..31f2a4df936 100644
--- a/providers/edenai/models/openai/gpt-pro-latest.toml
+++ b/providers/edenai/models/openai/gpt-pro-latest.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-5.5-pro"
+name = "GPT Pro Latest (GPT-5.5 Pro)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/ovhcloud/gpt-oss-120b.toml b/providers/edenai/models/ovhcloud/gpt-oss-120b.toml
index c8f65159859..008c82e0cbf 100644
--- a/providers/edenai/models/ovhcloud/gpt-oss-120b.toml
+++ b/providers/edenai/models/ovhcloud/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (OVHcloud)"
 tool_call = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/ovhcloud/gpt-oss-20b.toml b/providers/edenai/models/ovhcloud/gpt-oss-20b.toml
index 616a18348e8..102125304ad 100644
--- a/providers/edenai/models/ovhcloud/gpt-oss-20b.toml
+++ b/providers/edenai/models/ovhcloud/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (OVHcloud)"
 tool_call = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 1f11f70d4ba..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Alibaba)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 7e31b605e19..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Alibaba)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 170b09c1bf1..5269ff6d619 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Scaleway)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 3bc7ab9b2af..6125d529272 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Scaleway)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 65b07937b29..a86f8391e98 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
+name = "Llama-3.3-70B-Instruct (Scaleway)"
 attachment = false
 structured_output = false
 
diff --git a/providers/edenai/models/tensorx/deepseek/deepseek-v4-flash-0731.toml b/providers/edenai/models/tensorx/deepseek/deepseek-v4-flash-0731.toml
index 8e0dd3fbcbf..b08e9335e76 100644
--- a/providers/edenai/models/tensorx/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/tensorx/deepseek/deepseek-v4-flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (TensorX)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml b/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml
index 74e9ef4da95..7676b089585 100644
--- a/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/tensorx/deepseek/deepseek-v4-pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (TensorX)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml b/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
index ea149e3c5f0..87d3cc945d7 100644
--- a/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
+++ b/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
@@ -1,4 +1,5 @@
 base_model = "moonshotai/kimi-k2.5"
+name = "Kimi K2.5 (TensorX)"
 structured_output = false
 reasoning_options = []
 
diff --git a/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Flash-0731.toml
index 76a721abc1d..7059ffb8bf9 100644
--- a/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (Together AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Pro-0813.toml
index 7ea403f9d79..a859c5f3781 100644
--- a/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Pro-0813.toml
+++ b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Together AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/together_ai/meta-models/Muse-Glimmer-30B.toml b/providers/edenai/models/together_ai/meta-models/Muse-Glimmer-30B.toml
index 2812d8df497..24eaec5a33d 100644
--- a/providers/edenai/models/together_ai/meta-models/Muse-Glimmer-30B.toml
+++ b/providers/edenai/models/together_ai/meta-models/Muse-Glimmer-30B.toml
@@ -1,4 +1,5 @@
 base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (Together AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/together_ai/openai/gpt-oss-120b.toml b/providers/edenai/models/together_ai/openai/gpt-oss-120b.toml
index ee19598d69f..4c7b1c1c1da 100644
--- a/providers/edenai/models/together_ai/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/together_ai/openai/gpt-oss-120b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-120b"
+name = "GPT OSS 120B (Together AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml b/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
index 75f0276365b..4ac31b02dbc 100644
--- a/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
+++ b/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
@@ -1,4 +1,5 @@
 base_model = "openai/gpt-oss-20b"
+name = "GPT OSS 20B (Together AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml b/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml
index 97f9be04094..9627b70053f 100644
--- a/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml
+++ b/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling-small"
+name = "Inkling Small (Together AI)"
 structured_output = false
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/together_ai/thinkingmachines/Inkling.toml b/providers/edenai/models/together_ai/thinkingmachines/Inkling.toml
index 4d654c74315..66bfe97992d 100644
--- a/providers/edenai/models/together_ai/thinkingmachines/Inkling.toml
+++ b/providers/edenai/models/together_ai/thinkingmachines/Inkling.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling"
+name = "Inkling (Together AI)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/vertex/gemini-2.5-flash-image.toml b/providers/edenai/models/vertex/gemini-2.5-flash-image.toml
index 02d2fae70e5..9494e8da758 100644
--- a/providers/edenai/models/vertex/gemini-2.5-flash-image.toml
+++ b/providers/edenai/models/vertex/gemini-2.5-flash-image.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-2.5-flash-image"
+name = "Nano Banana (Vertex AI)"
 structured_output = true
 reasoning_options = []
 
diff --git a/providers/edenai/models/vertex/gemini-3-flash-preview.toml b/providers/edenai/models/vertex/gemini-3-flash-preview.toml
index 9acd0dde8dc..1ff397be940 100644
--- a/providers/edenai/models/vertex/gemini-3-flash-preview.toml
+++ b/providers/edenai/models/vertex/gemini-3-flash-preview.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3-flash-preview"
+name = "Gemini 3 Flash Preview (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3-pro-image.toml b/providers/edenai/models/vertex/gemini-3-pro-image.toml
index dbeda02bac7..80b8191940a 100644
--- a/providers/edenai/models/vertex/gemini-3-pro-image.toml
+++ b/providers/edenai/models/vertex/gemini-3-pro-image.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3-pro-image"
+name = "Nano Banana Pro (Vertex AI)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-image.toml b/providers/edenai/models/vertex/gemini-3.1-flash-image.toml
index 124aed86d30..008f70de785 100644
--- a/providers/edenai/models/vertex/gemini-3.1-flash-image.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-image.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-flash-image"
+name = "Nano Banana 2 (Vertex AI)"
 structured_output = true
 
 [[reasoning_options]]
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite-image.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite-image.toml
index 9a410b9bf77..5fa34cea0b7 100644
--- a/providers/edenai/models/vertex/gemini-3.1-flash-lite-image.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite-image.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-flash-lite-image"
+name = "Nano Banana 2 Lite (Vertex AI)"
 tool_call = false
 structured_output = true
 
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
index c9c3c4295e6..72cced92785 100644
--- a/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-flash-lite"
+name = "Gemini 3.1 Flash Lite (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
index f5fa68cae40..83586e62f83 100644
--- a/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.1-flash-lite"
-name = "Gemini 3.1 Flash Lite (EU)"
+name = "Gemini 3.1 Flash Lite (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml b/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml
index 2c7d840b683..43b451572a6 100644
--- a/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-flash-lite@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.1-flash-lite"
-name = "Gemini 3.1 Flash Lite (US)"
+name = "Gemini 3.1 Flash Lite (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.1-pro-preview.toml b/providers/edenai/models/vertex/gemini-3.1-pro-preview.toml
index 8daae0c833e..b2a7736dd52 100644
--- a/providers/edenai/models/vertex/gemini-3.1-pro-preview.toml
+++ b/providers/edenai/models/vertex/gemini-3.1-pro-preview.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-pro-preview"
+name = "Gemini 3.1 Pro Preview (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash-lite.toml b/providers/edenai/models/vertex/gemini-3.5-flash-lite.toml
index 6b453fde7ea..bdf99faf08a 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash-lite.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash-lite.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.5-flash-lite"
+name = "Gemini 3.5 Flash Lite (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash-lite@eu.toml b/providers/edenai/models/vertex/gemini-3.5-flash-lite@eu.toml
index b20e5f62901..903fd416455 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash-lite@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash-lite@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.5-flash-lite"
-name = "Gemini 3.5 Flash Lite (EU)"
+name = "Gemini 3.5 Flash Lite (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash-lite@us.toml b/providers/edenai/models/vertex/gemini-3.5-flash-lite@us.toml
index 2208a8ca3e0..06e99f9ab6a 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash-lite@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash-lite@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.5-flash-lite"
-name = "Gemini 3.5 Flash Lite (US)"
+name = "Gemini 3.5 Flash Lite (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash.toml b/providers/edenai/models/vertex/gemini-3.5-flash.toml
index 1d261426563..95432f795b9 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.5-flash"
+name = "Gemini 3.5 Flash (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.5-flash@eu.toml
index c2c6ac845d5..10645d51a40 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.5-flash"
-name = "Gemini 3.5 Flash (EU)"
+name = "Gemini 3.5 Flash (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.5-flash@us.toml b/providers/edenai/models/vertex/gemini-3.5-flash@us.toml
index 0a18a5eab71..13f2dad5653 100644
--- a/providers/edenai/models/vertex/gemini-3.5-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.5-flash@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.5-flash"
-name = "Gemini 3.5 Flash (US)"
+name = "Gemini 3.5 Flash (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.6-flash.toml b/providers/edenai/models/vertex/gemini-3.6-flash.toml
index 230c56de2e5..c4c1b8a1eb7 100644
--- a/providers/edenai/models/vertex/gemini-3.6-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.6-flash.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.6-flash"
+name = "Gemini 3.6 Flash (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.6-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.6-flash@eu.toml
index a91902ab537..affe7bd94f3 100644
--- a/providers/edenai/models/vertex/gemini-3.6-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.6-flash@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.6-flash"
-name = "Gemini 3.6 Flash (EU)"
+name = "Gemini 3.6 Flash (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.6-flash@us.toml b/providers/edenai/models/vertex/gemini-3.6-flash@us.toml
index 315ad0ad0b0..7db2190999f 100644
--- a/providers/edenai/models/vertex/gemini-3.6-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.6-flash@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.6-flash"
-name = "Gemini 3.6 Flash (US)"
+name = "Gemini 3.6 Flash (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash.toml b/providers/edenai/models/vertex/gemini-3.7-flash.toml
index 1a64f9681ee..cec321b9ba5 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.7-flash"
+name = "Gemini 3.7 Flash (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
index f582538d697..41edb99ded2 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.7-flash"
-name = "Gemini 3.7 Flash (EU)"
+name = "Gemini 3.7 Flash (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.7-flash@us.toml b/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
index c422d1e56f7..5cfc072389a 100644
--- a/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.7-flash@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.7-flash"
-name = "Gemini 3.7 Flash (US)"
+name = "Gemini 3.7 Flash (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash.toml b/providers/edenai/models/vertex/gemini-3.8-flash.toml
index e1d0abd66e5..678f54bb0aa 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.8-flash"
+name = "Gemini 3.8 Flash (Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
index 3fbeb6b0ae9..9426cf35d75 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@eu.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.8-flash"
-name = "Gemini 3.8 Flash (EU)"
+name = "Gemini 3.8 Flash (Vertex AI, EU)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-3.8-flash@us.toml b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
index b0a19cbb5df..1971cc01e88 100644
--- a/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
+++ b/providers/edenai/models/vertex/gemini-3.8-flash@us.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemini-3.8-flash"
-name = "Gemini 3.8 Flash (US)"
+name = "Gemini 3.8 Flash (Vertex AI, US)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-flash-latest.toml b/providers/edenai/models/vertex/gemini-flash-latest.toml
index e1d0abd66e5..3df1c883ad6 100644
--- a/providers/edenai/models/vertex/gemini-flash-latest.toml
+++ b/providers/edenai/models/vertex/gemini-flash-latest.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.8-flash"
+name = "Gemini Flash Latest (Gemini 3.8 Flash, Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/vertex/gemini-pro-latest.toml b/providers/edenai/models/vertex/gemini-pro-latest.toml
index 8daae0c833e..f0c0fcb8cbf 100644
--- a/providers/edenai/models/vertex/gemini-pro-latest.toml
+++ b/providers/edenai/models/vertex/gemini-pro-latest.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-pro-preview"
+name = "Gemini Pro Latest (Gemini 3.1 Pro Preview, Vertex AI)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/xai/grok-latest.toml b/providers/edenai/models/xai/grok-latest.toml
index cbcf623d7cd..eece4afa8de 100644
--- a/providers/edenai/models/xai/grok-latest.toml
+++ b/providers/edenai/models/xai/grok-latest.toml
@@ -1,4 +1,5 @@
 base_model = "xai/grok-4.6"
+name = "Grok Latest (Grok 4.6)"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/google/models/gemini-3.1-flash-image-preview.toml b/providers/google/models/gemini-3.1-flash-image-preview.toml
index e12ea22a95b..ca8bd4dd2ff 100644
--- a/providers/google/models/gemini-3.1-flash-image-preview.toml
+++ b/providers/google/models/gemini-3.1-flash-image-preview.toml
@@ -1,4 +1,4 @@
-name = "Nano Banana 2"
+name = "Nano Banana 2 Preview"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
 family = "gemini-flash"
 release_date = "2026-02-26"
diff --git a/sync.md b/sync.md
index 3e8b9f684ba..7595a924dc4 100644
--- a/sync.md
+++ b/sync.md
@@ -144,6 +144,8 @@ Actions are pinned by commit SHA. Keep new workflow actions pinned the same way.
 ## Eden AI Notes
 
 - Source endpoint: `https://api.edenai.run/v3/models`; no authentication required.
+- Latest aliases (`alias_of` plus an ID ending in `-latest`) get a distinct display name such as `Claude Fable Latest (Claude Fable 5.1)` so they do not collide with the versioned target in UIs that key on `name`. Case-only `alias_of` duplicates are not treated as latest aliases.
+- Extra labels (current target, host, region) share one parenthetical, e.g. `Gemini Flash Latest (Gemini 3.8 Flash, Vertex AI)` and `GPT OSS 120B (Deep Infra)`. The lab's own API keeps the unsuffixed canonical name; other hosts (Vertex AI, Deep Infra, Groq, Together AI, …) are named.
 - Reasoning effort options are derived from the lab's provider entry or OpenRouter. A toggle-only or budget-only control is not an effort list; do not invent effort levels.
 - When the effort mapper cannot resolve controls, preserve the existing route's authored `reasoning_options` while syncing other authoritative fields. Do not replace authored toggle, effort, or budget controls with `[]`.
 - New reasoning models with neither a resolved mapping nor authored controls remain skipped for manual authoring. No empty placeholder is generated, so the normal auto-merge policy remains unchanged; legitimate always-on `[]` entries are not blanket-blocked.

From ebd194c02bf1974bad08c0d0750f0abeff3cf8e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 17:20:40 +0000
Subject: [PATCH 0772/2012] chore(sync): update Eden AI model catalog (#6430)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/databricks/databricks-inkling.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/edenai/models/databricks/databricks-inkling.toml b/providers/edenai/models/databricks/databricks-inkling.toml
index e32923d89a4..177f6bd3f2d 100644
--- a/providers/edenai/models/databricks/databricks-inkling.toml
+++ b/providers/edenai/models/databricks/databricks-inkling.toml
@@ -1,4 +1,5 @@
 base_model = "thinkingmachines/inkling"
+name = "Inkling (Databricks)"
 structured_output = false
 
 [[reasoning_options]]

From 93d4c3337e57eeafa391cee6e2f730ed8b210cc8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 17:20:42 +0000
Subject: [PATCH 0773/2012] chore(sync): update Google model catalog (#6431)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/google/models/gemini-3-pro-image-preview.toml     | 1 +
 providers/google/models/gemini-3.1-flash-image-preview.toml | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/google/models/gemini-3-pro-image-preview.toml b/providers/google/models/gemini-3-pro-image-preview.toml
index 64807b961ff..6ac9306cf6a 100644
--- a/providers/google/models/gemini-3-pro-image-preview.toml
+++ b/providers/google/models/gemini-3-pro-image-preview.toml
@@ -1,4 +1,5 @@
 base_model = "google/gemini-3-pro-image-preview"
+name = "Nano Banana Pro"
 reasoning_options = []
 
 [cost]
diff --git a/providers/google/models/gemini-3.1-flash-image-preview.toml b/providers/google/models/gemini-3.1-flash-image-preview.toml
index ca8bd4dd2ff..e12ea22a95b 100644
--- a/providers/google/models/gemini-3.1-flash-image-preview.toml
+++ b/providers/google/models/gemini-3.1-flash-image-preview.toml
@@ -1,4 +1,4 @@
-name = "Nano Banana 2 Preview"
+name = "Nano Banana 2"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
 family = "gemini-flash"
 release_date = "2026-02-26"

From dc0cb7b75bd625f62508a626b50edd49b63545ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 17:20:44 +0000
Subject: [PATCH 0774/2012] chore(sync): update OpenRouter model catalog
 (#6432)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c19d2ff4995..404f8d66b38 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07994
-output = 0.15988
-cache_read = 0.015988
+input = 0.08022
+output = 0.16044
+cache_read = 0.016044
 
 [limit]
 context = 1_048_576

From ffd5a98b30b2dbe6d6d2351550af5e2065201dee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 18:26:21 +0000
Subject: [PATCH 0775/2012] chore(sync): update OpenRouter model catalog
 (#6435)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml   | 8 ++------
 3 files changed, 8 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 404f8d66b38..c19d2ff4995 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08022
-output = 0.16044
-cache_read = 0.016044
+input = 0.07994
+output = 0.15988
+cache_read = 0.015988
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 96c9123e736..74ef47494d8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.657198
-output = 1.314396
-cache_read = 0.054767
+input = 0.647106
+output = 1.294212
+cache_read = 0.053926
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
index a5c5b5bf91f..1adcf16cd0e 100644
--- a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
@@ -6,12 +6,8 @@ base_model = "alibaba/qwen3.5-397b-a17b"
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 3.5
-cache_read = 0.225
-
-[limit]
-output = 235_929
+input = 0.39
+output = 2.34
 
 [modalities]
 input = ["text", "image", "video"]

From 7cd97ea00bf703562839be3bbd00b5ec5eb24ab9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 18:26:29 +0000
Subject: [PATCH 0776/2012] chore(sync): update Kilo model catalog (#6436)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.5-397b-a17b.toml | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
index 3d76879dbef..d0e349f341a 100644
--- a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
@@ -9,8 +9,5 @@ values = ["none", "high"]
 input = 0.39
 output = 2.34
 
-[limit]
-output = 235_929
-
 [modalities]
 input = ["text", "image", "video"]

From db83107636fa89df070151de753f36577e5d76db Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 19:20:04 +0000
Subject: [PATCH 0777/2012] chore(sync): update OpenRouter model catalog
 (#6437)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c19d2ff4995..620219f924c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07994
-output = 0.15988
-cache_read = 0.015988
+input = 0.07966
+output = 0.15932
+cache_read = 0.015932
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 74ef47494d8..6af7edcf708 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.647106
-output = 1.294212
-cache_read = 0.053926
+input = 0.63684
+output = 1.27368
+cache_read = 0.05307
 
 [limit]
 context = 1_048_576

From 4a4332cd1ff9cf20d84003acdcbc122929359ded Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 20:24:12 +0000
Subject: [PATCH 0778/2012] chore(sync): update OpenRouter model catalog
 (#6438)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 6af7edcf708..b78d26ea76d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.63684
-output = 1.27368
-cache_read = 0.05307
+input = 0.626748
+output = 1.253496
+cache_read = 0.052229
 
 [limit]
 context = 1_048_576

From 9f5e4c1b34d4850e3121bc6cc20ccec8c5554258 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 21:21:33 +0000
Subject: [PATCH 0779/2012] chore(sync): update OpenRouter model catalog
 (#6440)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 620219f924c..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07966
-output = 0.15932
-cache_read = 0.015932
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576

From 4e11a917dc98efff54bffb21fa7e9362a20efdb3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 22:22:54 +0000
Subject: [PATCH 0780/2012] chore(sync): update OpenRouter model catalog
 (#6441)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..2bdde0e35b4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.0791
+output = 0.1582
+cache_read = 0.01582
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b78d26ea76d..f5feb49efa2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.626748
-output = 1.253496
-cache_read = 0.052229
+input = 0.616656
+output = 1.233312
+cache_read = 0.051388
 
 [limit]
 context = 1_048_576

From 0f314b6326f2cc7b35cc544a0a60c236fed5b943 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 6 Sep 2026 23:21:53 +0000
Subject: [PATCH 0781/2012] chore(sync): update OpenRouter model catalog
 (#6442)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 2bdde0e35b4..aa6d2a9c87f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0791
-output = 0.1582
-cache_read = 0.01582
+input = 0.07882
+output = 0.15764
+cache_read = 0.015764
 
 [limit]
 context = 1_048_576

From edb84fec55da8f72cf152525d10b644c634b4b98 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 01:04:16 +0000
Subject: [PATCH 0782/2012] chore(sync): update NanoGPT model catalog (#6444)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 4fbcd3c588979661cbd6b7155df63a87859d6ed0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 01:04:25 +0000
Subject: [PATCH 0783/2012] chore(sync): update OpenRouter model catalog
 (#6443)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml          | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 .../models/qwen/qwen3-next-80b-a3b-thinking.toml         | 1 +
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 ++++++---
 providers/openrouter/models/z-ai/glm-5.3.toml            | 4 ++--
 8 files changed, 24 insertions(+), 20 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 26fedac913a..97aeb796e4c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04998
-output = 0.09996
-cache_read = 0.009996
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index aa6d2a9c87f..57d98266566 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07882
-output = 0.15764
-cache_read = 0.015764
+input = 0.089866
+output = 0.179732
+cache_read = 0.017973
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..09d98bc8130 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 1.12068
+output = 3.36204
+cache_read = 0.037356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f5feb49efa2..0efd13162d1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.616656
-output = 1.233312
-cache_read = 0.051388
+input = 1.0353
+output = 2.0706
+cache_read = 0.086275
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 14e222d894d..35f13bee686 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -8,3 +8,4 @@ output = 1.2
 
 [limit]
 context = 262_144
+output = 235_929
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 6e636f26b9c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 0269e6a5aba..d939b2691d5 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -8,8 +8,8 @@ values = ["low", "high", "max"]
 [cost]
 input = 1.4
 output = 4.4
-cache_read = 0.14
+cache_read = 0.26
 
 [limit]
 context = 1_310_720
-output = 262_144
+output = 943_718

From 40d743643425e7d490278958124a728d8c354611 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 01:04:27 +0000
Subject: [PATCH 0784/2012] chore(sync): update LLM Gateway model catalog
 (#6445)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/kimi-k2.5.toml           | 19 -------------------
 1 file changed, 19 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/deepinfra/kimi-k2.5.toml

diff --git a/providers/llmgateway-providers/models/deepinfra/kimi-k2.5.toml b/providers/llmgateway-providers/models/deepinfra/kimi-k2.5.toml
deleted file mode 100644
index 3e524f9954f..00000000000
--- a/providers/llmgateway-providers/models/deepinfra/kimi-k2.5.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "moonshotai/kimi-k2.5"
-name = "Kimi K2.5 (DeepInfra)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.45
-output = 2.25
-cache_read = 0.07
-
-[limit]
-context = 256_000
-output = 131_072

From 42c43a60c035e7e5d9383b6d94d6dcefc12e6f25 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 01:04:56 +0000
Subject: [PATCH 0785/2012] chore(sync): update Kilo model catalog (#6446)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml  |  4 ++++
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                | 10 +++++++---
 providers/kilo/models/z-ai/glm-5.3.toml                |  2 +-
 4 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 7c76fb082af..5d67b671b7a 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -9,3 +9,7 @@ values = ["high"]
 [cost]
 input = 0.15
 output = 1.2
+
+[limit]
+context = 262_144
+output = 235_929
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fee21b1f2d8..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,6 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+context = 198_000
+output = 16_384
diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index 9e9823ba910..959b6c62561 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -12,4 +12,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 262_144
+output = 943_718

From 9156b622baad134a43c6da2f0bf535ac9302c5ac Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 01:39:34 +0000
Subject: [PATCH 0786/2012] chore(sync): update OpenRouter model catalog
 (#6448)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 182a8affd9f213acc83c782d31e2b64d05b8e119 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 03:32:14 +0000
Subject: [PATCH 0787/2012] chore(sync): update OpenRouter model catalog
 (#6451)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 57d98266566..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.089866
-output = 0.179732
-cache_read = 0.017973
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index a6efa176c68..27d5c88cb9e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.045
-output = 0.09
-cache_read = 0.009
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]

From 7a5375f3b3f4af10ad79ddbe337cef4c84b1edd3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 03:32:26 +0000
Subject: [PATCH 0788/2012] chore(sync): update Kilo model catalog (#6450)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 67b240e58e2..80dc12b4ce8 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.045
-output = 0.09
-cache_read = 0.009
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]

From 3c26ab2c79995ef7b3691290d8ab863d78448341 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Sun, 6 Sep 2026 22:55:07 -0500
Subject: [PATCH 0789/2012] feat(amazon-bedrock): add Grok 4.6 inference
 profiles (#6452)

---
 .../amazon-bedrock/models/global.xai.grok-4.6.toml    | 11 +++++++++++
 providers/amazon-bedrock/models/us.xai.grok-4.6.toml  | 11 +++++++++++
 2 files changed, 22 insertions(+)
 create mode 100644 providers/amazon-bedrock/models/global.xai.grok-4.6.toml
 create mode 100644 providers/amazon-bedrock/models/us.xai.grok-4.6.toml

diff --git a/providers/amazon-bedrock/models/global.xai.grok-4.6.toml b/providers/amazon-bedrock/models/global.xai.grok-4.6.toml
new file mode 100644
index 00000000000..552f860bcc3
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.xai.grok-4.6.toml
@@ -0,0 +1,11 @@
+# Source: AWS Bedrock Grok 4.6 model card — Runtime profiles, pricing, and capabilities.
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+name = "Grok 4.6 (Global)"
+last_updated = "2026-08-18"
+structured_output = false
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.50
diff --git a/providers/amazon-bedrock/models/us.xai.grok-4.6.toml b/providers/amazon-bedrock/models/us.xai.grok-4.6.toml
new file mode 100644
index 00000000000..44fb743ba04
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.xai.grok-4.6.toml
@@ -0,0 +1,11 @@
+# Source: AWS Bedrock Grok 4.6 model card — Runtime profiles, pricing, and capabilities.
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+name = "Grok 4.6 (US)"
+last_updated = "2026-08-18"
+structured_output = false
+
+[cost]
+input = 2.20
+output = 6.60
+cache_read = 0.55

From 6641168de72aab4a33e1aab4d6468972c4bcf4ce Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Mon, 7 Sep 2026 11:55:17 +0800
Subject: [PATCH 0790/2012] feat(vivgrid): add gpt-6-astra model (#6447)

---
 providers/vivgrid/models/gpt-6-astra.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/vivgrid/models/gpt-6-astra.toml

diff --git a/providers/vivgrid/models/gpt-6-astra.toml b/providers/vivgrid/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..afdc38e7f28
--- /dev/null
+++ b/providers/vivgrid/models/gpt-6-astra.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
\ No newline at end of file

From f904b5683234b4656ef8377a91bb711c4ee56d83 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 04:31:32 +0000
Subject: [PATCH 0791/2012] chore(sync): update NanoGPT model catalog (#6454)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From cf912f89ad267f512bb70b3ad64bab19c4fb8e5f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 04:31:47 +0000
Subject: [PATCH 0792/2012] chore(sync): update OpenRouter model catalog
 (#6453)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 09d98bc8130..c0fb8edf24d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12068
-output = 3.36204
-cache_read = 0.037356
+input = 1.1154
+output = 3.3462
+cache_read = 0.03718
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 0efd13162d1..66443ade4c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.0353
-output = 2.0706
-cache_read = 0.086275
+input = 1.04226
+output = 2.08452
+cache_read = 0.086855
 
 [limit]
 context = 1_048_576

From 1f800ba082a98cb76ff611669dc191e7938ff0c6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 04:31:51 +0000
Subject: [PATCH 0793/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6455)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/minimax/minimax-m2.7-free.toml | 11 -----------
 .../vercel/models/minimax/minimax-m3-free.toml   | 16 ----------------
 2 files changed, 27 deletions(-)
 delete mode 100644 providers/vercel/models/minimax/minimax-m2.7-free.toml
 delete mode 100644 providers/vercel/models/minimax/minimax-m3-free.toml

diff --git a/providers/vercel/models/minimax/minimax-m2.7-free.toml b/providers/vercel/models/minimax/minimax-m2.7-free.toml
deleted file mode 100644
index d47133c050a..00000000000
--- a/providers/vercel/models/minimax/minimax-m2.7-free.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "minimax/MiniMax-M2.7"
-name = "Minimax M2.7 (Free)"
-reasoning_options = []
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 196_608
-output = 196_608
diff --git a/providers/vercel/models/minimax/minimax-m3-free.toml b/providers/vercel/models/minimax/minimax-m3-free.toml
deleted file mode 100644
index c277befa85d..00000000000
--- a/providers/vercel/models/minimax/minimax-m3-free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "minimax/MiniMax-M3"
-name = "MiniMax M3 (Free)"
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[limit]
-output = 1_048_576
-
-[modalities]
-input = ["text", "image"]

From c5a3d800cf5540ee323e964642473e3bacb0208a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 05:27:24 +0000
Subject: [PATCH 0794/2012] chore(sync): update OpenRouter model catalog
 (#6456)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..8119fb87874 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 16ae7fcba35c2746b5c0e84fbbbf023aec0fb049 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 05:27:29 +0000
Subject: [PATCH 0795/2012] chore(sync): update Eden AI model catalog (#6457)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From b30527fe491926cb487151da9e9c77490f736afc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 06:50:20 +0000
Subject: [PATCH 0796/2012] chore(sync): update NanoGPT model catalog (#6461)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 7dc788b657435c6b26b1d5f80ee7d45070c0c961 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 06:50:23 +0000
Subject: [PATCH 0797/2012] chore(sync): update OpenRouter model catalog
 (#6462)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From c12d832b40d93d84b5f84633be0372bd2962551e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 06:50:37 +0000
Subject: [PATCH 0798/2012] chore(sync): update Kilo model catalog (#6460)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..7b848afc9df 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 1b1108add293255437dcb94aae4a1b12469368d3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 07:37:03 +0000
Subject: [PATCH 0799/2012] chore(sync): update Kilo model catalog (#6463)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml      |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index a88240b8a86..61e4718fa7b 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.17
-output = 3.96
-cache_read = 0.234
+input = 1.12
+output = 3.52
+cache_read = 0.208
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From b5385daa57186b3ae73b009d738141cd4f25e382 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 08:38:02 +0000
Subject: [PATCH 0800/2012] chore(sync): update Cortecs model catalog (#6465)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/cosmos3-super-reasoner.toml        | 23 -------------------
 .../llama-3.1-nemotron-ultra-253b-v1.toml     | 23 -------------------
 .../models/nvidia-nemotron-3-nano-omni.toml   | 23 -------------------
 .../models/qwen3-next-80b-a3b-thinking.toml   | 12 ----------
 4 files changed, 81 deletions(-)
 delete mode 100644 providers/cortecs/models/cosmos3-super-reasoner.toml
 delete mode 100644 providers/cortecs/models/llama-3.1-nemotron-ultra-253b-v1.toml
 delete mode 100644 providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
 delete mode 100644 providers/cortecs/models/qwen3-next-80b-a3b-thinking.toml

diff --git a/providers/cortecs/models/cosmos3-super-reasoner.toml b/providers/cortecs/models/cosmos3-super-reasoner.toml
deleted file mode 100644
index 61bc01e31da..00000000000
--- a/providers/cortecs/models/cosmos3-super-reasoner.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "cosmos3-super-reasoner"
-description = "Cosmos3 Super Reasoner is a high-capacity reasoning model designed for complex multi-agent tasks and advanced physical AI understanding."
-release_date = "2026-06-02"
-last_updated = "2026-06-02"
-attachment = false
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.099
-output = 0.296
-
-[limit]
-context = 256_000
-output = 256_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cortecs/models/llama-3.1-nemotron-ultra-253b-v1.toml b/providers/cortecs/models/llama-3.1-nemotron-ultra-253b-v1.toml
deleted file mode 100644
index 7b81f53f5b3..00000000000
--- a/providers/cortecs/models/llama-3.1-nemotron-ultra-253b-v1.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "llama-3.1-nemotron-ultra-253b-v1"
-description = "A reasoning-optimized LLM based on Llama 3.1, Nemotron Ultra 253B delivers strong performance in tasks like RAG and tool use, with high efficiency and reduced latency."
-release_date = "2025-04-07"
-last_updated = "2025-04-07"
-attachment = false
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.598
-output = 1.794
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml b/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
deleted file mode 100644
index 8dbfc72233f..00000000000
--- a/providers/cortecs/models/nvidia-nemotron-3-nano-omni.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "nvidia-nemotron-3-nano-omni"
-description = "Nemotron-3-Nano-Omni is an open, efficient omni-modal reasoning model that unifies text, image, audio, and video for agentic AI workflows."
-release_date = "2026-04-29"
-last_updated = "2026-04-29"
-attachment = false
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.059
-output = 0.237
-
-[limit]
-context = 300_000
-output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cortecs/models/qwen3-next-80b-a3b-thinking.toml b/providers/cortecs/models/qwen3-next-80b-a3b-thinking.toml
deleted file mode 100644
index d0add520f99..00000000000
--- a/providers/cortecs/models/qwen3-next-80b-a3b-thinking.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "alibaba/qwen3-next-80b-a3b-thinking"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 0.149
-output = 1.195
-
-[limit]
-context = 128_000
-output = 128_000

From 872d94d546231e442ecb5b353acda85ce88aeabd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 08:38:07 +0000
Subject: [PATCH 0801/2012] chore(sync): update OpenRouter model catalog
 (#6464)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c0fb8edf24d..5bc4e6af877 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.1154
-output = 3.3462
-cache_read = 0.03718
+input = 1.0494
+output = 3.1482
+cache_read = 0.03498
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 66443ade4c4..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.04226
-output = 2.08452
-cache_read = 0.086855
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 2eab1890c5a..1b1d7eb0fda 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.17
-output = 3.96
-cache_read = 0.234
+input = 1.12
+output = 3.52
+cache_read = 0.208
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From e151078e09bfa1ddeb093ab81c9dd1dce414015b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 10:27:59 +0000
Subject: [PATCH 0802/2012] chore(sync): update OpenRouter model catalog
 (#6475)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 00c8960400a4de02332e04f7c21863e724af8eb3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 10:28:33 +0000
Subject: [PATCH 0803/2012] chore(sync): update NanoGPT model catalog (#6473)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/qwen3-8b.toml   | 27 -------------------
 ...epseek-v4-flash-vision-exp-uncensored.toml | 25 +++++++++++++++++
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 3 files changed, 28 insertions(+), 30 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/qwen3-8b.toml
 create mode 100644 providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml

diff --git a/providers/nano-gpt/models/TEE/qwen3-8b.toml b/providers/nano-gpt/models/TEE/qwen3-8b.toml
deleted file mode 100644
index cf35e5e6ed6..00000000000
--- a/providers/nano-gpt/models/TEE/qwen3-8b.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Qwen3 8B TEE"
-description = "Qwen3 8B is an open-weight dense language model from Alibaba's Qwen team for efficient dialogue, reasoning, mathematics, coding, and tool use. Running inside a TEE (Trusted Execution Environment), with provider attestation support."
-release_date = "2026-09-04"
-last_updated = "2026-09-04"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 0.11
-output = 0.45
-cache_read = 0.11
-
-[limit]
-context = 40_960
-input = 40_960
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml
new file mode 100644
index 00000000000..4e196584f84
--- /dev/null
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml
@@ -0,0 +1,25 @@
+name = "DeepSeek V4 Flash Vision Exp Uncensored"
+description = "An uncensored variant of the experimental vision-enabled DeepSeek V4 Flash model for chat, image understanding, reasoning, coding, and tool use, with a 524K context window."
+family = "deepseek"
+release_date = "2026-09-07"
+last_updated = "2026-09-07"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.44
+output = 1.6
+cache_read = 0.22
+
+[limit]
+context = 524_288
+input = 524_288
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From c12e91757aa9a652a506c3ed73d4462a810b5b67 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 11:25:28 +0000
Subject: [PATCH 0804/2012] chore(sync): update Kilo model catalog (#6495)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-5.6-sol-discounted.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml b/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml
index d29366792b0..6008af017c3 100644
--- a/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml
+++ b/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml
@@ -15,11 +15,11 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.5
-output = 15
+input = 2
+output = 10
 reasoning = 0
-cache_read = 0.25
-cache_write = 3.125
+cache_read = 0.2
+cache_write = 2.5
 
 [limit]
 context = 1_050_000

From 46f9c00718704efeca18898bed932ab1deeb4617 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 15:25:47 +0000
Subject: [PATCH 0805/2012] chore(sync): update Cortecs model catalog (#6501)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/glm-5.3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/cortecs/models/glm-5.3.toml b/providers/cortecs/models/glm-5.3.toml
index d2aacc59483..d6f49c7e113 100644
--- a/providers/cortecs/models/glm-5.3.toml
+++ b/providers/cortecs/models/glm-5.3.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.399
-cache_read = 0.26
+input = 1.114
+output = 3.899
+cache_read = 0.279
 
 [limit]
 context = 1_048_576

From ada94c523bcd4edfb1b9454b54e05de7d27a1743 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 15:25:58 +0000
Subject: [PATCH 0806/2012] chore(sync): update NanoGPT model catalog (#6504)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/glm-5.2.toml          | 4 ++--
 providers/nano-gpt/models/TEE/glm-5.2:thinking.toml | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/nano-gpt/models/TEE/glm-5.2.toml b/providers/nano-gpt/models/TEE/glm-5.2.toml
index e47df3b05ad..d13661c5591 100644
--- a/providers/nano-gpt/models/TEE/glm-5.2.toml
+++ b/providers/nano-gpt/models/TEE/glm-5.2.toml
@@ -4,8 +4,8 @@ reasoning_options = []
 
 [cost]
 input = 1.4
-output = 4.6
-cache_read = 0.5
+output = 4.4
+cache_read = 0.7
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/TEE/glm-5.2:thinking.toml b/providers/nano-gpt/models/TEE/glm-5.2:thinking.toml
index 5e70a8d829d..25c9de217ca 100644
--- a/providers/nano-gpt/models/TEE/glm-5.2:thinking.toml
+++ b/providers/nano-gpt/models/TEE/glm-5.2:thinking.toml
@@ -7,8 +7,8 @@ values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.4
-output = 4.6
-cache_read = 0.5
+output = 4.4
+cache_read = 0.7
 
 [limit]
 context = 1_048_576

From 1253bb6c05076b5bf0f49ad05b55fce018d1cf26 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 15:26:00 +0000
Subject: [PATCH 0807/2012] chore(sync): update OpenRouter model catalog
 (#6503)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5bc4e6af877..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0494
-output = 3.1482
-cache_read = 0.03498
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From e276afc41174ec63b87d1584df4ed050d3ecd931 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 15:26:05 +0000
Subject: [PATCH 0808/2012] chore(sync): update Eden AI model catalog (#6502)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 ...epseek-v4-flash-0731.toml => DeepSeek-V4-Flash-0731.toml} | 0
 providers/edenai/models/mistral/magistral-medium-latest.toml | 5 +++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml     | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml       | 4 ++--
 4 files changed, 7 insertions(+), 6 deletions(-)
 rename providers/edenai/models/flexai/{deepseek-v4-flash-0731.toml => DeepSeek-V4-Flash-0731.toml} (100%)

diff --git a/providers/edenai/models/flexai/deepseek-v4-flash-0731.toml b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
similarity index 100%
rename from providers/edenai/models/flexai/deepseek-v4-flash-0731.toml
rename to providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
diff --git a/providers/edenai/models/mistral/magistral-medium-latest.toml b/providers/edenai/models/mistral/magistral-medium-latest.toml
index 4e681bedade..cc8503b4e41 100644
--- a/providers/edenai/models/mistral/magistral-medium-latest.toml
+++ b/providers/edenai/models/mistral/magistral-medium-latest.toml
@@ -4,8 +4,9 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 2
-output = 5
+input = 1.5
+output = 7.5
+cache_read = 0.15
 
 [limit]
 context = 262_144
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424

From 91037a5f5be26214ebc4190cbf91728dad765330 Mon Sep 17 00:00:00 2001
From: Vladimir Glafirov <vglafirov@gitlab.com>
Date: Mon, 7 Sep 2026 17:52:27 +0200
Subject: [PATCH 0809/2012] feat: add gitlab duo-chat-gpt-6-astra model (#6499)

---
 providers/gitlab/models/duo-chat-gpt-6-astra.toml | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/gitlab/models/duo-chat-gpt-6-astra.toml

diff --git a/providers/gitlab/models/duo-chat-gpt-6-astra.toml b/providers/gitlab/models/duo-chat-gpt-6-astra.toml
new file mode 100644
index 00000000000..55ba068c60b
--- /dev/null
+++ b/providers/gitlab/models/duo-chat-gpt-6-astra.toml
@@ -0,0 +1,9 @@
+base_model = "openai/gpt-6-astra"
+name = "Agentic Chat (GPT-6 Astra)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 5b7d265da4ec9c2ae175003e8317548c4fe977ee Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Mon, 7 Sep 2026 23:53:03 +0800
Subject: [PATCH 0810/2012] feat(vivgrid): add claude-fable-5-1,
 claude-fable-5, gemini-3.8-flash, deepseek-v4-pro-0813 (#6496)

Adds four newly launched Vivgrid models using the base_model override pattern.
Pricing, context windows and max output come from https://vivgrid.com/docs/models.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 .../vivgrid/models/claude-fable-5-1.toml      | 13 +++++++++++
 providers/vivgrid/models/claude-fable-5.toml  | 13 +++++++++++
 .../vivgrid/models/deepseek-v4-pro-0813.toml  | 22 +++++++++++++++++++
 .../vivgrid/models/gemini-3.8-flash.toml      | 14 ++++++++++++
 4 files changed, 62 insertions(+)
 create mode 100644 providers/vivgrid/models/claude-fable-5-1.toml
 create mode 100644 providers/vivgrid/models/claude-fable-5.toml
 create mode 100644 providers/vivgrid/models/deepseek-v4-pro-0813.toml
 create mode 100644 providers/vivgrid/models/gemini-3.8-flash.toml

diff --git a/providers/vivgrid/models/claude-fable-5-1.toml b/providers/vivgrid/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..93b3075879a
--- /dev/null
+++ b/providers/vivgrid/models/claude-fable-5-1.toml
@@ -0,0 +1,13 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/claude-fable-5-1 (accessed 2026-09-07)
+# Cached input is $0.50/MTok. Vivgrid publishes no cache-write rate and states
+# pricing matches the original provider, so Anthropic's $12.50/MTok applies.
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.5
+cache_write = 12.5
diff --git a/providers/vivgrid/models/claude-fable-5.toml b/providers/vivgrid/models/claude-fable-5.toml
new file mode 100644
index 00000000000..239c250c0d4
--- /dev/null
+++ b/providers/vivgrid/models/claude-fable-5.toml
@@ -0,0 +1,13 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/claude-fable-5 (accessed 2026-09-07)
+# Cached input is $1.25/MTok. Vivgrid publishes no cache-write rate and states
+# pricing matches the original provider, so Anthropic's $12.50/MTok applies.
+base_model = "anthropic/claude-fable-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1.25
+cache_write = 12.5
diff --git a/providers/vivgrid/models/deepseek-v4-pro-0813.toml b/providers/vivgrid/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..be0105b9639
--- /dev/null
+++ b/providers/vivgrid/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,22 @@
+# Pricing and capabilities:
+# https://docs.vivgrid.com/models/deepseek-v4-pro-0813 (accessed 2026-09-07)
+# Reasoning tokens bill at the output rate (no separate CoT price).
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = high|max
+base_model = "deepseek/deepseek-v4-pro-0813"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.35
+output = 3.00
+reasoning = 3.00
+cache_read = 0.05
diff --git a/providers/vivgrid/models/gemini-3.8-flash.toml b/providers/vivgrid/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..32db1ca1276
--- /dev/null
+++ b/providers/vivgrid/models/gemini-3.8-flash.toml
@@ -0,0 +1,14 @@
+# Pricing and capabilities:
+# https://docs.vivgrid.com/models/gemini-3.8-flash (accessed 2026-09-07)
+# Vivgrid documents 128K max output for this model, double the 64K the Gemini
+# Flash line normally serves.
+base_model = "google/gemini-3.8-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.15
+
+[limit]
+output = 128_000

From fc8f9d8d768422da7ca6c0e6c480db3d411a5578 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 16:27:22 +0000
Subject: [PATCH 0811/2012] chore(sync): update Kilo model catalog (#6506)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From d95c3d7d18f66433d3e63a37a4d5fd236407b6af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 16:27:28 +0000
Subject: [PATCH 0812/2012] chore(sync): update OpenRouter model catalog
 (#6505)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 1c7eb6cc0dcbf3a303c16179fd13513e4a540435 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 18:29:52 +0000
Subject: [PATCH 0813/2012] chore(sync): update OpenRouter model catalog
 (#6512)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/minimax/minimax-m2.7:free.toml        | 13 -------------
 .../models/minimax/minimax-m3:free.toml          | 16 ----------------
 2 files changed, 29 deletions(-)
 delete mode 100644 providers/openrouter/models/minimax/minimax-m2.7:free.toml
 delete mode 100644 providers/openrouter/models/minimax/minimax-m3:free.toml

diff --git a/providers/openrouter/models/minimax/minimax-m2.7:free.toml b/providers/openrouter/models/minimax/minimax-m2.7:free.toml
deleted file mode 100644
index a7661fb85f8..00000000000
--- a/providers/openrouter/models/minimax/minimax-m2.7:free.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "minimax/MiniMax-M2.7"
-name = "MiniMax M2.7 (free)"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-structured_output = false
-reasoning_options = []
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 196_608
-output = 176_947
diff --git a/providers/openrouter/models/minimax/minimax-m3:free.toml b/providers/openrouter/models/minimax/minimax-m3:free.toml
deleted file mode 100644
index 961bae5aa8e..00000000000
--- a/providers/openrouter/models/minimax/minimax-m3:free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-base_model = "minimax/MiniMax-M3"
-name = "MiniMax M3 (free)"
-description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
-structured_output = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-output = 943_718

From effd4f8ca3ea503c65f980c850139653fda73890 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 18:30:02 +0000
Subject: [PATCH 0814/2012] chore(sync): update LLM Gateway model catalog
 (#6510)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/Qwen3.8-27B.toml                 | 4 ++--
 .../models/consensusprotocol/glm-5.3-flash.toml               | 4 ----
 2 files changed, 2 insertions(+), 6 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
index b5674abce63..7d725a4921e 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
@@ -2,7 +2,7 @@ name = "Qwen3.8 27B (Consensus Protocol)"
 description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
 release_date = "2026-09-02"
 last_updated = "2026-09-02"
-attachment = false
+attachment = true
 reasoning = true
 temperature = true
 tool_call = true
@@ -23,5 +23,5 @@ context = 32_768
 output = 32_768
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]
diff --git a/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
index bbc353ac4a5..d53e927044b 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
@@ -1,6 +1,5 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM-5.3 Flash (Consensus Protocol)"
-attachment = false
 structured_output = false
 
 [[reasoning_options]]
@@ -14,6 +13,3 @@ cache_read = 0.02
 
 [limit]
 context = 1_048_576
-
-[modalities]
-input = ["text"]

From a234b2f5b74b0c9314e60f6a14eaae7c4c6fb55e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 18:30:07 +0000
Subject: [PATCH 0815/2012] chore(sync): update Kilo model catalog (#6511)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/minimax/minimax-m2.7:free.toml   | 16 ----------------
 .../kilo/models/minimax/minimax-m3:free.toml     | 15 ---------------
 2 files changed, 31 deletions(-)
 delete mode 100644 providers/kilo/models/minimax/minimax-m2.7:free.toml
 delete mode 100644 providers/kilo/models/minimax/minimax-m3:free.toml

diff --git a/providers/kilo/models/minimax/minimax-m2.7:free.toml b/providers/kilo/models/minimax/minimax-m2.7:free.toml
deleted file mode 100644
index a0258eb4533..00000000000
--- a/providers/kilo/models/minimax/minimax-m2.7:free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "minimax/MiniMax-M2.7"
-name = "MiniMax: MiniMax M2.7 (free)"
-description = "MiniMax-M2.7 is a next-generation large language model designed for autonomous, real-world productivity and continuous improvement. Built to actively participate in its own evolution, M2.7 integrates advanced agentic capabilities through multi-agent..."
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["high"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 196_608
-output = 176_947
diff --git a/providers/kilo/models/minimax/minimax-m3:free.toml b/providers/kilo/models/minimax/minimax-m3:free.toml
deleted file mode 100644
index d922c671141..00000000000
--- a/providers/kilo/models/minimax/minimax-m3:free.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "minimax/MiniMax-M3"
-name = "MiniMax: MiniMax M3 (free)"
-description = "MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding,..."
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-output = 943_718

From 01a4140b892f8bdfbaceb903c5d13b40c394fcec Mon Sep 17 00:00:00 2001
From: Ryan Madsen <97546348+ryanmadsenn@users.noreply.github.com>
Date: Mon, 7 Sep 2026 13:19:38 -0600
Subject: [PATCH 0816/2012] fix(fireworks-ai): correct GLM 5.3 / Flash serving
 limits (#6508)

* fix(fireworks-ai): correct GLM 5.3 / Flash serving limits

* fix(fireworks-ai): correct max_tokens validation note in GLM 5.3 comments
---
 .../accounts/fireworks/models/glm-5p3-flash.toml  | 11 ++++++++++-
 .../models/accounts/fireworks/models/glm-5p3.toml | 15 ++++++++++++++-
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
index 8a1203370b8..06b9cb00646 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
@@ -5,9 +5,15 @@
 # empirically 2026-09-04 — same prompt yields ~9/17/559 reasoning tokens for
 # low/high/max. max/xhigh select max.
 # https://docs.fireworks.ai/api-reference/post-chatcompletions
+# Serving limits verified 2026-09-07: GET /v1/models reports context_length
+# 1_048_576; an over-limit probe (1_200_012 prompt tokens) was rejected with
+# "model maximum context length: 1048573" — the enforced prompt cap, recorded
+# here (the lab default is 1M context). Output stays at the inherited
+# 131_072: Fireworks performs no pre-flight max_tokens validation (up to
+# 2_000_000 accepted), and Baseten documents 131_072 for GLM-5.3-Flash.
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
-last_updated = "2026-09-04"
+last_updated = "2026-09-07"
 
 [[reasoning_options]]
 type = "effort"
@@ -20,3 +26,6 @@ field = "reasoning_content"
 input = 0.15
 output = 0.50
 cache_read = 0.03
+
+[limit]
+context = 1_048_573
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
index fa2c2de0072..6ba5c186726 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
@@ -6,9 +6,18 @@
 # verified empirically 2026-09-04 — same prompt yields ~10/25/198 reasoning
 # tokens for low/high/max. max/xhigh select max.
 # https://docs.fireworks.ai/api-reference/post-chatcompletions
+# Serving limits verified 2026-09-07: GET /v1/models reports context_length
+# 1_048_576; an over-limit probe (1_200_012 prompt tokens) was rejected with
+# "model maximum context length: 1048573" — the enforced prompt cap, recorded
+# here. Fireworks performs no pre-flight max_tokens validation at all
+# (262_144, 524_288, and 2_000_000 all accepted), so a request-side output
+# ceiling is untestable by probes; output = 262_144 per the Baseten-
+# documented generation ceiling for the same weights. The Z.ai lab default
+# (1M / 128K, https://docs.z.ai/guides/llm/glm-5.3) understates Fireworks
+# serving.
 base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
-last_updated = "2026-09-04"
+last_updated = "2026-09-07"
 
 [[reasoning_options]]
 type = "effort"
@@ -21,3 +30,7 @@ field = "reasoning_content"
 input = 1.40
 output = 4.40
 cache_read = 0.26
+
+[limit]
+context = 1_048_573
+output = 262_144

From c5b41d639debb60f969de7925e75bb1550b77c88 Mon Sep 17 00:00:00 2001
From: Andre Landgraf <andre.timo.landgraf@gmail.com>
Date: Mon, 7 Sep 2026 12:20:06 -0700
Subject: [PATCH 0817/2012] Mark GLM-5.3-Flash as open weights. (#6509)

The checkpoint is public on Hugging Face (zai-org/GLM-5.3-Flash, MIT, safetensors).
---
 models/zhipuai/glm-5.3-flash.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/models/zhipuai/glm-5.3-flash.toml b/models/zhipuai/glm-5.3-flash.toml
index d77dee2034f..d080a79f86f 100644
--- a/models/zhipuai/glm-5.3-flash.toml
+++ b/models/zhipuai/glm-5.3-flash.toml
@@ -1,3 +1,4 @@
+# Open weights: https://huggingface.co/zai-org/GLM-5.3-Flash (MIT, safetensors)
 name = "GLM-5.3-Flash"
 description = "Native multimodal GLM model for efficient coding and long-horizon agent tasks"
 family = "glm"
@@ -8,7 +9,7 @@ reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_000_000

From 61920df12a1c9d71ac33fe034a4a66c486da5e0f Mon Sep 17 00:00:00 2001
From: Ryan Madsen <97546348+ryanmadsenn@users.noreply.github.com>
Date: Mon, 7 Sep 2026 13:20:15 -0600
Subject: [PATCH 0818/2012] feat(fireworks-ai): add GLM 5.3 Fast router (#6507)

* feat(fireworks-ai): add GLM 5.3 Fast router

* feat(fireworks-ai): verify GLM 5.3 Fast limits and release date

* feat(fireworks-ai): record enforced context cap for glm-5p3-fast

* fix(fireworks-ai): correct max_tokens validation note in glm-5p3-fast comment
---
 .../fireworks/routers/glm-5p3-fast.toml       | 40 +++++++++++++++++++
 1 file changed, 40 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-fast.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-fast.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-fast.toml
new file mode 100644
index 00000000000..c65dd42dfce
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-fast.toml
@@ -0,0 +1,40 @@
+# Fast serving-path router for GLM 5.3 — same model, faster serving (the
+# chat-completions response echoes the underlying
+# accounts/fireworks/models/glm-5p3):
+# https://docs.fireworks.ai/serverless/serving-paths
+# Router is live on Fireworks accounts but not yet listed on the public
+# pricing page (which currently lists GLM 5.2 Fast only). Pricing confirmed
+# from a Fireworks account rate card (2026-09-07): $2.10 / $0.39 / $6.60
+# per 1M (input / cached input / output) — a 1.5x premium over GLM 5.3
+# standard, matching the GLM 5.2 Fast premium over GLM 5.2.
+# Limits and release verified 2026-09-07: GET /v1/models reports
+# context_length 1_048_576 and a created timestamp of 2026-08-28 (identical
+# to glm-5p3 — router launched with the base model); an over-limit probe
+# (1_200_012 prompt tokens) was rejected
+# with "model maximum context length: 1048572" — the enforced prompt cap
+# sits 4 tokens below the advertised context_length, so that value is
+# recorded here. Fireworks performs no pre-flight max_tokens validation
+# (2_000_000 accepted); output = 262_144 per the Baseten/Vercel-documented
+# generation ceiling for the same weights.
+# Thinking-only, effort tiers identical to the sibling glm-5p3 entry:
+# low/high/max (max/xhigh select max).
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 Fast"
+release_date = "2026-08-28"
+last_updated = "2026-09-07"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.10
+output = 6.60
+cache_read = 0.39
+
+[limit]
+context = 1_048_572
+output = 262_144

From 711589ce737dbacfc72f204f68d3ef9e8ab09359 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 19:22:22 +0000
Subject: [PATCH 0819/2012] chore(sync): update Venice model catalog (#6514)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/openai-gpt-56-luna-pro.toml  | 2 +-
 providers/venice/models/openai-gpt-56-terra-pro.toml | 2 +-
 providers/venice/models/z-ai-glm-5-3-flash.toml      | 1 -
 3 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/venice/models/openai-gpt-56-luna-pro.toml b/providers/venice/models/openai-gpt-56-luna-pro.toml
index 0ecf54ab9e3..8e0a117d967 100644
--- a/providers/venice/models/openai-gpt-56-luna-pro.toml
+++ b/providers/venice/models/openai-gpt-56-luna-pro.toml
@@ -5,7 +5,7 @@ description = "Frontier GPT model for professional reasoning, coding, and multim
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.25
diff --git a/providers/venice/models/openai-gpt-56-terra-pro.toml b/providers/venice/models/openai-gpt-56-terra-pro.toml
index 945c54a580e..313a829c386 100644
--- a/providers/venice/models/openai-gpt-56-terra-pro.toml
+++ b/providers/venice/models/openai-gpt-56-terra-pro.toml
@@ -5,7 +5,7 @@ description = "Frontier GPT model for professional reasoning, coding, and multim
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 2.5
diff --git a/providers/venice/models/z-ai-glm-5-3-flash.toml b/providers/venice/models/z-ai-glm-5-3-flash.toml
index 9c56abf44f5..fd7d0779d73 100644
--- a/providers/venice/models/z-ai-glm-5-3-flash.toml
+++ b/providers/venice/models/z-ai-glm-5-3-flash.toml
@@ -2,7 +2,6 @@ base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
 release_date = "2026-08-21"
-open_weights = true
 
 [[reasoning_options]]
 type = "effort"

From 1a84fdd72ad6c7f507af96aafbcc59a2f818f9fd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 19:22:25 +0000
Subject: [PATCH 0820/2012] chore(sync): update Weights & Biases model catalog
 (#6513)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/zai-org/GLM-5.3-Flash.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
index c3b6661bb43..91d7e42d00a 100644
--- a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
+++ b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
@@ -3,7 +3,6 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
-open_weights = true
 reasoning_options = []
 
 [cost]

From 33d28d4500cd490657b6bf4fd5c07a20f08d6de8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 20:25:33 +0000
Subject: [PATCH 0821/2012] chore(sync): update Kilo model catalog (#6519)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml | 5 +++--
 providers/kilo/models/~moonshotai/kimi-latest.toml        | 6 +++---
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
index 764c5575e7c..e46b521c61f 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
@@ -11,8 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.25
-output = 1
+input = 0.29
+output = 1.14
+cache_read = 0.11
 
 [limit]
 context = 163_840
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index de31079e2dd..6481e4e1993 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.5
+output = 14
+cache_read = 0.29
 
 [limit]
 context = 1_048_576

From 144c75394246872c660978ada7fc3b354fdaf254 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 20:25:36 +0000
Subject: [PATCH 0822/2012] chore(sync): update NanoGPT model catalog (#6518)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml      | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 2d0019c9056..7278af0463a 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high"]
 
 [cost]
-input = 0.35
-output = 1.4
-cache_read = 0.175
+input = 0.07
+output = 0.21
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 5a2f0f1c97d7ac783f13b283f92afd79f7577efd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 20:25:55 +0000
Subject: [PATCH 0823/2012] chore(sync): update OpenRouter model catalog
 (#6520)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-chat-v3-0324.toml   | 5 +++--
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
index 905726abf7a..d8a154006c7 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
@@ -12,8 +12,9 @@ knowledge = "2024-07-31"
 open_weights = true
 
 [cost]
-input = 0.25
-output = 1
+input = 0.29
+output = 1.14
+cache_read = 0.11
 
 [limit]
 context = 163_840
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 70251750002..052fb05abd3 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.5
+output = 14
+cache_read = 0.29
 
 [limit]
 context = 1_048_576

From 24fd77d4665e984453e633855079d9f1f35c5935 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 21:23:44 +0000
Subject: [PATCH 0824/2012] chore(sync): update Kilo model catalog (#6522)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
index e46b521c61f..764c5575e7c 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
@@ -11,9 +11,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.29
-output = 1.14
-cache_read = 0.11
+input = 0.25
+output = 1
 
 [limit]
 context = 163_840

From 255c4fc8cd50b68b3ebb6ccdf402f9816d7749a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 21:23:51 +0000
Subject: [PATCH 0825/2012] chore(sync): update OpenRouter model catalog
 (#6521)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-chat-v3-0324.toml    | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
index d8a154006c7..905726abf7a 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
@@ -12,9 +12,8 @@ knowledge = "2024-07-31"
 open_weights = true
 
 [cost]
-input = 0.29
-output = 1.14
-cache_read = 0.11
+input = 0.25
+output = 1
 
 [limit]
 context = 163_840

From fb2e542be5a54873977ce81de4505a47854b1707 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 22:23:40 +0000
Subject: [PATCH 0826/2012] chore(sync): update Charm Hyper model catalog
 (#6524)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml | 6 +++---
 providers/hyper/models/glm-5.1.toml            | 6 +++---
 providers/hyper/models/kimi-k2.5.toml          | 6 +++---
 providers/hyper/models/minimax-m2.7.toml       | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index c98d649e236..443e26fe9ac 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.106
-output = 0.368
-cache_write = 0.053
+input = 0.11
+output = 0.408
+cache_write = 0.055
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 532a9b2b8f7..a1b201fde20 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.262
-output = 4.132
-cache_write = 0.631
+input = 1.332
+output = 4.312
+cache_write = 0.666
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index 6629d8c2afd..b8552721ba3 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5144
-output = 2.755
-cache_write = 0.2572
+input = 0.5584
+output = 2.935
+cache_write = 0.2792
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 4ca78e9dbd1..f49f9b449be 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.484
-output = 1.852
-cache_write = 0.242
+input = 0.47
+output = 1.76
+cache_write = 0.235
 
 [limit]
 context = 262_100

From f87dee5da29db9df20050c88ea096275cf34817b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 22:23:44 +0000
Subject: [PATCH 0827/2012] chore(sync): update NanoGPT model catalog (#6523)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/poolside/laguna-xs-2.1.toml | 10 ++++++++++
 1 file changed, 10 insertions(+)
 create mode 100644 providers/nano-gpt/models/poolside/laguna-xs-2.1.toml

diff --git a/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml b/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml
new file mode 100644
index 00000000000..3f0e815a50b
--- /dev/null
+++ b/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml
@@ -0,0 +1,10 @@
+base_model = "poolside/laguna-xs-2.1"
+reasoning_options = []
+
+[cost]
+input = 0.063
+output = 0.126
+cache_read = 0.0315
+
+[limit]
+input = 262_144

From 2aa825fb3f47783b454c70083411ed62b4fe6f97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:00:21 +0000
Subject: [PATCH 0828/2012] chore(sync): update Kilo model catalog (#6525)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-chat-v3-0324.toml    |  5 +++--
 providers/kilo/models/deepseek/deepseek-chat-v3.1.toml |  4 ++--
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml |  2 +-
 providers/kilo/models/qwen/qwen3-30b-a3b.toml          |  2 +-
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                | 10 +++-------
 providers/kilo/models/z-ai/glm-4.7-flash.toml          |  7 +++----
 7 files changed, 16 insertions(+), 20 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
index 764c5575e7c..e46b521c61f 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
@@ -11,8 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.25
-output = 1
+input = 0.29
+output = 1.14
+cache_read = 0.11
 
 [limit]
 context = 163_840
diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
index 7717353dce8..4269d8ff1fa 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3.1.toml
@@ -20,8 +20,8 @@ output = 1
 cache_read = 0.135
 
 [limit]
-context = 161_000
-output = 144_900
+context = 163_840
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 75e607a6e02..3e83e7378b1 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b.toml b/providers/kilo/models/qwen/qwen3-30b-a3b.toml
index a4d4e484294..a8c6cef56d6 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b.toml
@@ -1,6 +1,6 @@
 base_model = "alibaba/qwen3-30b-a3b"
 description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fee21b1f2d8 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,6 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-context = 198_000
-output = 16_384
+input = 0.55
+output = 2.2
+cache_read = 0.11
diff --git a/providers/kilo/models/z-ai/glm-4.7-flash.toml b/providers/kilo/models/z-ai/glm-4.7-flash.toml
index 96d11325c93..1c40468998f 100644
--- a/providers/kilo/models/z-ai/glm-4.7-flash.toml
+++ b/providers/kilo/models/z-ai/glm-4.7-flash.toml
@@ -7,10 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.06
+input = 0.0605
 output = 0.4
-cache_read = 0.01
 
 [limit]
-context = 202_752
-output = 16_384
+context = 131_072
+output = 117_964

From be5d544ca8eab60e390f91e6113dc8950776b829 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:00:30 +0000
Subject: [PATCH 0829/2012] chore(sync): update OpenRouter model catalog
 (#6527)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-chat-v3-0324.toml           | 5 +++--
 .../openrouter/models/deepseek/deepseek-chat-v3.1.toml   | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 6 +++---
 .../openrouter/models/moonshotai/kimi-k2-thinking.toml   | 3 +--
 .../openrouter/models/moonshotai/kimi-k2.7-code.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3-30b-a3b.toml      | 4 +++-
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 +++------
 providers/openrouter/models/z-ai/glm-4.7-flash.toml      | 5 ++---
 9 files changed, 25 insertions(+), 27 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
index 905726abf7a..d8a154006c7 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
@@ -12,8 +12,9 @@ knowledge = "2024-07-31"
 open_weights = true
 
 [cost]
-input = 0.25
-output = 1
+input = 0.29
+output = 1.14
+cache_read = 0.11
 
 [limit]
 context = 163_840
diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
index d60c40e2af2..24302ed5e58 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3.1.toml
@@ -17,13 +17,13 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 1.65
-cache_read = 0.55
+input = 0.25
+output = 0.95
+cache_read = 0.13
 
 [limit]
 context = 163_840
-output = 144_900
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..5bc4e6af877 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.0494
+output = 3.1482
+cache_read = 0.03498
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 91ad22ed77f..3ccd5a4f3ef 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,7 +8,6 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
-cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
index 09933092d9a..0006cbd4e10 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
@@ -3,9 +3,9 @@ temperature = true
 reasoning_options = []
 
 [cost]
-input = 0.66
-output = 3.4
-cache_read = 0.18
+input = 0.71
+output = 3.5
+cache_read = 0.15
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b.toml
index 6c07fb64c63..dca225b27ec 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b.toml
@@ -1,6 +1,8 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3-30b-a3b"
 description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..6e636f26b9c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.55
+output = 2.2
+cache_read = 0.11
diff --git a/providers/openrouter/models/z-ai/glm-4.7-flash.toml b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
index d2109236f90..1745eeaa253 100644
--- a/providers/openrouter/models/z-ai/glm-4.7-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
@@ -10,10 +10,9 @@ field = "reasoning_details"
 type = "toggle"
 
 [cost]
-input = 0.06
+input = 0.0605
 output = 0.4
-cache_read = 0.01
 
 [limit]
 context = 202_752
-output = 16_384
+output = 117_964

From 182fd58d50d4c305a7c70272d5e8d13f46650ff8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:00:40 +0000
Subject: [PATCH 0830/2012] chore(sync): update Deep Infra model catalog
 (#6526)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepinfra/models/moonshotai/Kimi-K2.5.toml      | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
index 597479e9f1e..83e9e8835d0 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
@@ -5,12 +5,18 @@ release_date = "2026-01-27"
 last_updated = "2026-01-27"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
-structured_output = true
 temperature = true
 tool_call = true
+structured_output = true
 knowledge = "2025-01"
 open_weights = true
+status = "deprecated"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.45
@@ -24,6 +30,3 @@ output = 32_768
 [modalities]
 input = ["text", "image", "video"]
 output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

From 3479581fe9f4b5f527a027466006cb115c1e13e0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:35:06 +0000
Subject: [PATCH 0831/2012] chore(sync): update NanoGPT model catalog (#6528)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From d10c119525e7e8e144cf2c59cd6455640d7c3279 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:35:09 +0000
Subject: [PATCH 0832/2012] chore(sync): update OpenRouter model catalog
 (#6530)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/openrouter/models/minimax/minimax-m1.toml         | 2 +-
 providers/openrouter/models/qwen/qwen3-14b.toml             | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 178789497a8..063613f9d24 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -17,7 +17,7 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.55
+input = 0.4
 output = 2.2
 
 [limit]
diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index b10030292e3..b6ce24c73e0 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -17,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.12
-output = 0.24
+input = 0.2275
+output = 0.91
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
 
 [modalities]
 input = ["text"]

From ed0a75d822f54e61a25737b4051cffb704ac0f84 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 01:35:18 +0000
Subject: [PATCH 0833/2012] chore(sync): update Kilo model catalog (#6529)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-14b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 2ce8bc46042..67828967f41 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 40_960
-output = 16_384
+context = 131_072
+output = 8_192
 
 [modalities]
 input = ["text"]

From de925e265d88fea45692eeba6b335df9305733a3 Mon Sep 17 00:00:00 2001
From: Martin Mose Facondini <me@martinmose.com>
Date: Tue, 8 Sep 2026 05:41:31 +0200
Subject: [PATCH 0834/2012] fix(zeldoc): update zdev config (#6517)

---
 providers/zeldoc/models/zdev.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/zeldoc/models/zdev.toml b/providers/zeldoc/models/zdev.toml
index 22294cd7219..2d6b42d72f5 100644
--- a/providers/zeldoc/models/zdev.toml
+++ b/providers/zeldoc/models/zdev.toml
@@ -2,11 +2,10 @@
 # https://docs.zeldoc.ai/ (accessed 2026-06-25)
 name = "ZDev"
 description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-release_date = "2026-04-15"
-last_updated = "2026-04-15"
+release_date = "2026-08-26"
+last_updated = "2026-08-26"
 attachment = true
 reasoning = true
-knowledge = "2025-01"
 temperature = true
 tool_call = true
 structured_output = true
@@ -14,7 +13,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"

From 83102d58fe5ebd48b2d3504cde618d3968f135b0 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 7 Sep 2026 22:42:04 -0500
Subject: [PATCH 0835/2012] fix: [privatemode] sync supported models (#6014)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/privatemode-ai/models/glm-5.3.toml  | 21 ++++++++++++++++++
 .../privatemode-ai/models/glm-latest.toml     | 22 +++++++++++++++++++
 2 files changed, 43 insertions(+)
 create mode 100644 providers/privatemode-ai/models/glm-5.3.toml
 create mode 100644 providers/privatemode-ai/models/glm-latest.toml

diff --git a/providers/privatemode-ai/models/glm-5.3.toml b/providers/privatemode-ai/models/glm-5.3.toml
new file mode 100644
index 00000000000..dfa4c3d6952
--- /dev/null
+++ b/providers/privatemode-ai/models/glm-5.3.toml
@@ -0,0 +1,21 @@
+# Sources (accessed 2026-09-01):
+# Model overview (preview): https://docs.privatemode.ai/models/overview/#glm-53
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Added in v1.55.0 (upgraded from GLM-5.2): https://docs.privatemode.ai/release/#v1550
+# Upstream: https://huggingface.co/zai-org/GLM-5.3
+# Effort: reasoning_effort = low|high|max (default max; always on).
+# Any other value including none is silently mapped to max; reasoning cannot be off.
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10),
+# matching other Privatemode catalog entries with the same EUR list prices.
+# EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
+base_model = "zhipuai/glm-5.3"
+status = "beta"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 1.7910
+output = 8.9436
+cache_read = 0.1733
+
+[limit]
+context = 256_000
diff --git a/providers/privatemode-ai/models/glm-latest.toml b/providers/privatemode-ai/models/glm-latest.toml
new file mode 100644
index 00000000000..dd1cbd2bcc9
--- /dev/null
+++ b/providers/privatemode-ai/models/glm-latest.toml
@@ -0,0 +1,22 @@
+# Sources (accessed 2026-09-01):
+# Alias routes to the latest available GLM model (currently GLM-5.3):
+# https://docs.privatemode.ai/models/overview/#glm-53
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Release: https://docs.privatemode.ai/release/#v1550
+# Effort: reasoning_effort = low|high|max (default max; always on).
+# Any other value including none is silently mapped to max; reasoning cannot be off.
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10),
+# matching other Privatemode catalog entries with the same EUR list prices.
+# EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
+base_model = "zhipuai/glm-5.3"
+name = "GLM (latest)"
+status = "beta"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 1.7910
+output = 8.9436
+cache_read = 0.1733
+
+[limit]
+context = 256_000

From c2f56eb5b44d82b8c195817dd7a8c53847cdcf50 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 04:29:33 +0000
Subject: [PATCH 0836/2012] chore(sync): update OpenRouter model catalog
 (#6532)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From a51b9546289e30b12030cd99db5e36991058d556 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 04:29:41 +0000
Subject: [PATCH 0837/2012] chore(sync): update NanoGPT model catalog (#6533)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 68b935a5a8bc1ce87639884194364466be4e2a80 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 05:26:43 +0000
Subject: [PATCH 0838/2012] chore(sync): update Kilo model catalog (#6536)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
index 7cb4fe7a46b..9f316a31c44 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-0731.toml
@@ -12,4 +12,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718

From 9cd50ca5e838c6686b23fe1c6b35b7fca495767c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 05:26:55 +0000
Subject: [PATCH 0839/2012] chore(sync): update OpenRouter model catalog
 (#6534)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 97aeb796e4c..d06dd2cd8d1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,10 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
+input = 0.065
+output = 0.18
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718

From ec924ec6b65d5001cbda71aeaebb8e279134cd77 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 05:27:38 +0000
Subject: [PATCH 0840/2012] chore(sync): update Eden AI model catalog (#6535)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From 0c482163b5b9e4bb149485a6b83cb4236a3fae6b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 06:40:01 +0000
Subject: [PATCH 0841/2012] chore(sync): update OpenRouter model catalog
 (#6538)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..0b9f0295d73 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 9d1e6bae238e9acedec66fa55d58ce9b9ac4fe0a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 06:40:10 +0000
Subject: [PATCH 0842/2012] chore(sync): update NanoGPT model catalog (#6539)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From fe0f84b1c8197b52148148d61e49ca6f02c8d78a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 06:40:14 +0000
Subject: [PATCH 0843/2012] chore(sync): update Kilo model catalog (#6537)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index da7301fa993..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 1.32
 output = 3.96
-cache_read = 0.132
+cache_read = 0.044
 
 [limit]
 context = 1_024_000

From 2c88ba794887e1bb186da26f4ff44722ce98677d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 08:34:19 +0000
Subject: [PATCH 0844/2012] chore(sync): update NanoGPT model catalog (#6540)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml      | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 7278af0463a..50bc9fa38b4 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high"]
 
 [cost]
-input = 0.07
-output = 0.21
-cache_read = 0.014
+input = 0.25
+output = 1
+cache_read = 0.125
 
 [limit]
 context = 1_048_576

From 3188fb2a4ba8b9602a110c6e229e7c613cbd3ec7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 09:29:04 +0000
Subject: [PATCH 0845/2012] chore(sync): update NanoGPT model catalog (#6544)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml  | 28 +++++++++++++++++++
 .../models/poolside/laguna-xs-2.1.toml        | 10 -------
 providers/nano-gpt/models/yi-large.toml       | 24 ----------------
 providers/nano-gpt/models/yi-medium-200k.toml | 24 ----------------
 4 files changed, 28 insertions(+), 58 deletions(-)
 create mode 100644 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
 delete mode 100644 providers/nano-gpt/models/poolside/laguna-xs-2.1.toml
 delete mode 100644 providers/nano-gpt/models/yi-large.toml
 delete mode 100644 providers/nano-gpt/models/yi-medium-200k.toml

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..40304b5bd8e
--- /dev/null
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek V4.1 Flash"
+description = "DeepSeek V4.1 Flash supports text and image input, reasoning, tool calling, and structured output with a 1M-token context window. This is a rate-limited beta with limited capacity, intended for testing rather than production use. Assume prompts and responses are logged by the provider and may be used for model training or service improvement. Do not send sensitive or confidential data."
+family = "deepseek"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.1562
+output = 0.3124
+cache_read = 0.03124
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml b/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml
deleted file mode 100644
index 3f0e815a50b..00000000000
--- a/providers/nano-gpt/models/poolside/laguna-xs-2.1.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "poolside/laguna-xs-2.1"
-reasoning_options = []
-
-[cost]
-input = 0.063
-output = 0.126
-cache_read = 0.0315
-
-[limit]
-input = 262_144
diff --git a/providers/nano-gpt/models/yi-large.toml b/providers/nano-gpt/models/yi-large.toml
deleted file mode 100644
index 5529660165a..00000000000
--- a/providers/nano-gpt/models/yi-large.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Yi Large"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "yi"
-release_date = "2024-01-01"
-last_updated = "2024-05-13"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 3.196
-output = 3.196
-cache_read = 1.598
-
-[limit]
-context = 32_000
-input = 32_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/yi-medium-200k.toml b/providers/nano-gpt/models/yi-medium-200k.toml
deleted file mode 100644
index e9b63e28d6a..00000000000
--- a/providers/nano-gpt/models/yi-medium-200k.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Yi Medium 200k"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "yi"
-release_date = "2024-01-01"
-last_updated = "2024-03-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.499
-output = 2.499
-cache_read = 1.2495
-
-[limit]
-context = 200_000
-input = 200_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 26703fa74cc2a990d4a095d5dce6d79dd2202524 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 09:29:36 +0000
Subject: [PATCH 0846/2012] chore(sync): update Requesty model catalog (#6545)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/deepseek-v4-flash-0731.toml     | 4 ++--
 providers/requesty/models/deepseek-v4-flash-0731@eu.toml  | 4 ++--
 providers/requesty/models/deepseek-v4-flash.toml          | 4 ++--
 providers/requesty/models/glm-5.3-flash.toml              | 2 +-
 providers/requesty/models/glm-5.3-flash@eu.toml           | 2 +-
 providers/requesty/models/kimi-k3.toml                    | 6 +++---
 providers/requesty/models/kimi-k3@eu.toml                 | 6 +++---
 providers/requesty/models/thinkingcap-qwen3.6-27b.toml    | 4 ++--
 providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml | 4 ++--
 9 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/providers/requesty/models/deepseek-v4-flash-0731.toml b/providers/requesty/models/deepseek-v4-flash-0731.toml
index d73d297b44e..ddf98793307 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731.toml
@@ -8,8 +8,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.14
-output = 0.28
+input = 0.28
+output = 0.56
 cache_read = 0.07
 
 [limit]
diff --git a/providers/requesty/models/deepseek-v4-flash-0731@eu.toml b/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
index e33044e4663..aa3265185b8 100644
--- a/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
+++ b/providers/requesty/models/deepseek-v4-flash-0731@eu.toml
@@ -9,8 +9,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.14
-output = 0.28
+input = 0.28
+output = 0.56
 cache_read = 0.07
 
 [limit]
diff --git a/providers/requesty/models/deepseek-v4-flash.toml b/providers/requesty/models/deepseek-v4-flash.toml
index 415669c9c5e..a0e563ee006 100644
--- a/providers/requesty/models/deepseek-v4-flash.toml
+++ b/providers/requesty/models/deepseek-v4-flash.toml
@@ -8,8 +8,8 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.14
-output = 0.28
+input = 0.28
+output = 0.56
 cache_read = 0.07
 
 [limit]
diff --git a/providers/requesty/models/glm-5.3-flash.toml b/providers/requesty/models/glm-5.3-flash.toml
index 45ac44632bb..de8892ee202 100644
--- a/providers/requesty/models/glm-5.3-flash.toml
+++ b/providers/requesty/models/glm-5.3-flash.toml
@@ -9,7 +9,7 @@ type = "budget_tokens"
 
 [cost]
 input = 0.2
-output = 0.5
+output = 0.6
 cache_read = 0.07
 
 [limit]
diff --git a/providers/requesty/models/glm-5.3-flash@eu.toml b/providers/requesty/models/glm-5.3-flash@eu.toml
index d58a76089fe..e31d9a04405 100644
--- a/providers/requesty/models/glm-5.3-flash@eu.toml
+++ b/providers/requesty/models/glm-5.3-flash@eu.toml
@@ -10,7 +10,7 @@ type = "budget_tokens"
 
 [cost]
 input = 0.2
-output = 0.5
+output = 0.6
 cache_read = 0.07
 
 [limit]
diff --git a/providers/requesty/models/kimi-k3.toml b/providers/requesty/models/kimi-k3.toml
index b1999cda9f0..c2452effd5c 100644
--- a/providers/requesty/models/kimi-k3.toml
+++ b/providers/requesty/models/kimi-k3.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.25
-output = 11.25
-cache_read = 0.225
+input = 3
+output = 15
+cache_read = 0.45
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/kimi-k3@eu.toml b/providers/requesty/models/kimi-k3@eu.toml
index bb3b723880d..021b0f05373 100644
--- a/providers/requesty/models/kimi-k3@eu.toml
+++ b/providers/requesty/models/kimi-k3@eu.toml
@@ -9,9 +9,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 2.25
-output = 11.25
-cache_read = 0.225
+input = 3
+output = 15
+cache_read = 0.45
 
 [limit]
 output = 262_144
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
index b34410d395c..6b202e90585 100644
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
+++ b/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
@@ -18,8 +18,8 @@ type = "budget_tokens"
 
 [cost]
 input = 0.4
-output = 3
-cache_read = 0.26
+output = 2.6
+cache_read = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
index ba21875fd54..d72b7f5786e 100644
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
+++ b/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
@@ -18,8 +18,8 @@ type = "budget_tokens"
 
 [cost]
 input = 0.4
-output = 3
-cache_read = 0.26
+output = 2.6
+cache_read = 0.05
 
 [limit]
 context = 262_144

From e36b2f6620b4dc3712539ced88782797942420dc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 10:27:21 +0000
Subject: [PATCH 0847/2012] chore(sync): update NanoGPT model catalog (#6546)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/qwen3-8b.toml   | 27 +++++++++++++++++++
 .../deepseek-v4-flash-vision-exp.toml         |  6 ++---
 2 files changed, 30 insertions(+), 3 deletions(-)
 create mode 100644 providers/nano-gpt/models/TEE/qwen3-8b.toml

diff --git a/providers/nano-gpt/models/TEE/qwen3-8b.toml b/providers/nano-gpt/models/TEE/qwen3-8b.toml
new file mode 100644
index 00000000000..cf35e5e6ed6
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/qwen3-8b.toml
@@ -0,0 +1,27 @@
+name = "Qwen3 8B TEE"
+description = "Qwen3 8B is an open-weight dense language model from Alibaba's Qwen team for efficient dialogue, reasoning, mathematics, coding, and tool use. Running inside a TEE (Trusted Execution Environment), with provider attestation support."
+release_date = "2026-09-04"
+last_updated = "2026-09-04"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.11
+output = 0.45
+cache_read = 0.11
+
+[limit]
+context = 40_960
+input = 40_960
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 49c3a9d06168ceed2ba087a231fc8ddacbdfa24d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 10:27:37 +0000
Subject: [PATCH 0848/2012] chore(sync): update OpenRouter model catalog
 (#6547)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0b9f0295d73..9e898f11a05 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 2a0d7c0b2b01dc854f32bf11d6792da29d2d8a4e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 11:24:46 +0000
Subject: [PATCH 0849/2012] chore(sync): update Kilo model catalog (#6550)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.5-397b-a17b.toml | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
index d0e349f341a..3d76879dbef 100644
--- a/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-397b-a17b.toml
@@ -9,5 +9,8 @@ values = ["none", "high"]
 input = 0.39
 output = 2.34
 
+[limit]
+output = 235_929
+
 [modalities]
 input = ["text", "image", "video"]

From 13a4840e427fa049eef6545cb17fadbf9cd21282 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 11:25:01 +0000
Subject: [PATCH 0850/2012] chore(sync): update OpenRouter model catalog
 (#6548)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
index 1adcf16cd0e..a5c5b5bf91f 100644
--- a/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-397b-a17b.toml
@@ -6,8 +6,12 @@ base_model = "alibaba/qwen3.5-397b-a17b"
 type = "toggle"
 
 [cost]
-input = 0.39
-output = 2.34
+input = 0.55
+output = 3.5
+cache_read = 0.225
+
+[limit]
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From 3f48b2f81c94abd5531fbc460772a4941c2a3046 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 11:25:17 +0000
Subject: [PATCH 0851/2012] chore(sync): update CrossModel model catalog
 (#6549)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/moonshot/kimi-k2.5.toml      | 14 --------------
 1 file changed, 14 deletions(-)
 delete mode 100644 providers/crossmodel/models/moonshot/kimi-k2.5.toml

diff --git a/providers/crossmodel/models/moonshot/kimi-k2.5.toml b/providers/crossmodel/models/moonshot/kimi-k2.5.toml
deleted file mode 100644
index 43dd71f19b9..00000000000
--- a/providers/crossmodel/models/moonshot/kimi-k2.5.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "moonshotai/kimi-k2.5"
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.62
-output = 3.3
-cache_read = 0.11
-cache_write = 0.62
-
-[limit]
-context = 262_000
-output = 262_000

From ce4959fdfea1a613034a81acea448d1d2bbf4d65 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 13:28:14 +0000
Subject: [PATCH 0852/2012] chore(sync): update OpenRouter model catalog
 (#6554)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nex-agi/nex-n2-mini.toml           | 29 -------------------
 .../openrouter/models/nex-agi/nex-n2-pro.toml | 29 -------------------
 .../openrouter/models/~z-ai/glm-latest.toml   |  8 ++---
 3 files changed, 4 insertions(+), 62 deletions(-)
 delete mode 100644 providers/openrouter/models/nex-agi/nex-n2-mini.toml
 delete mode 100644 providers/openrouter/models/nex-agi/nex-n2-pro.toml

diff --git a/providers/openrouter/models/nex-agi/nex-n2-mini.toml b/providers/openrouter/models/nex-agi/nex-n2-mini.toml
deleted file mode 100644
index ccad25ac0f1..00000000000
--- a/providers/openrouter/models/nex-agi/nex-n2-mini.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Nex-N2-Mini"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "agi"
-release_date = "2026-06-24"
-last_updated = "2026-06-24"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.025
-output = 0.1
-cache_read = 0.0025
-
-[limit]
-context = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/nex-agi/nex-n2-pro.toml b/providers/openrouter/models/nex-agi/nex-n2-pro.toml
deleted file mode 100644
index df1064ba567..00000000000
--- a/providers/openrouter/models/nex-agi/nex-n2-pro.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Nex-N2-Pro"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "agi"
-release_date = "2026-06-08"
-last_updated = "2026-06-08"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.25
-output = 1
-cache_read = 0.025
-
-[limit]
-context = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 1b1d7eb0fda..91a5845fc7b 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12
-output = 3.52
-cache_read = 0.208
+input = 1.113
+output = 3.498
+cache_read = 0.2067
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 128_000
 
 [modalities]
 input = ["text"]

From 331d59c817e50bf90bec3110d750c8b7bc2803f4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 13:28:33 +0000
Subject: [PATCH 0853/2012] chore(sync): update Kilo model catalog (#6552)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/nex-agi/nex-n2-mini.toml      | 28 -------------------
 providers/kilo/models/nex-agi/nex-n2-pro.toml | 28 -------------------
 providers/kilo/models/~z-ai/glm-latest.toml   | 10 +++----
 3 files changed, 5 insertions(+), 61 deletions(-)
 delete mode 100644 providers/kilo/models/nex-agi/nex-n2-mini.toml
 delete mode 100644 providers/kilo/models/nex-agi/nex-n2-pro.toml

diff --git a/providers/kilo/models/nex-agi/nex-n2-mini.toml b/providers/kilo/models/nex-agi/nex-n2-mini.toml
deleted file mode 100644
index b4d27604099..00000000000
--- a/providers/kilo/models/nex-agi/nex-n2-mini.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Nex AGI: Nex-N2-Mini (retires Sep 8)"
-description = "Nex-N2-Mini is an open-source agentic mixture-of-experts model from Nex AGI, the smaller sibling in the Nex-N2 series. It accepts text and image input and is built for coding, tool use,..."
-family = "agi"
-release_date = "2026-06-24"
-last_updated = "2026-06-24"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.025
-output = 0.1
-cache_read = 0.0025
-
-[limit]
-context = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/nex-agi/nex-n2-pro.toml b/providers/kilo/models/nex-agi/nex-n2-pro.toml
deleted file mode 100644
index b15895650f1..00000000000
--- a/providers/kilo/models/nex-agi/nex-n2-pro.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Nex AGI: Nex-N2-Pro (retires Sep 8)"
-description = "Nex-N2-Pro is an agentic mixture-of-experts model from Nex AGI, with 17B active parameters out of 397B total. Built on the Qwen3.5 architecture, it accepts text and image input and produces..."
-family = "agi"
-release_date = "2026-06-08"
-last_updated = "2026-06-08"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.25
-output = 1
-cache_read = 0.025
-
-[limit]
-context = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 61e4718fa7b..81909334ecc 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.12
-output = 3.52
-cache_read = 0.208
+input = 1.113
+output = 3.498
+cache_read = 0.2067
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
+output = 128_000
 
 [modalities]
 input = ["text"]

From e3a74a7aeb4994e9254781809eebfcc29123bf15 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 14:29:43 +0000
Subject: [PATCH 0854/2012] chore(sync): update OpenRouter model catalog
 (#6556)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/minimax/minimax-m1.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 063613f9d24..178789497a8 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -17,7 +17,7 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.4
+input = 0.55
 output = 2.2
 
 [limit]

From e56d4f72e9a86c7980144fd801311ac10336481c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 14:29:52 +0000
Subject: [PATCH 0855/2012] chore(sync): update Charm Hyper model catalog
 (#6555)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml | 6 +++---
 providers/hyper/models/glm-5.1.toml            | 6 +++---
 providers/hyper/models/gpt-oss-120b.toml       | 6 +++---
 providers/hyper/models/minimax-m2.7.toml       | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 443e26fe9ac..c98d649e236 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.11
-output = 0.408
-cache_write = 0.055
+input = 0.106
+output = 0.368
+cache_write = 0.053
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index a1b201fde20..b38d94327c2 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.332
-output = 4.312
-cache_write = 0.666
+input = 1.358
+output = 4.268
+cache_write = 0.679
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index ad08deed2d7..b8abd9a4a66 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.188
-output = 0.7
-cache_write = 0.094
+input = 0.178
+output = 0.68
+cache_write = 0.089
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index f49f9b449be..6ec098e5a6b 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.47
-output = 1.76
-cache_write = 0.235
+input = 0.458
+output = 1.712
+cache_write = 0.229
 
 [limit]
 context = 262_100

From 96ed2484605b7ced00d91bd363da28372ad0fe49 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:28:36 +0000
Subject: [PATCH 0856/2012] chore(sync): update Eden AI model catalog (#6559)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml    | 4 ++--
 providers/edenai/models/flexai/gpt-oss-120b.toml              | 4 ++--
 providers/edenai/models/flexai/gpt-oss-20b.toml               | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 10 files changed, 20 insertions(+), 20 deletions(-)

diff --git a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
index e25ac9dd061..8a3564408ab 100644
--- a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
@@ -8,8 +8,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.1
+input = 0.065
+output = 0.18
 
 [limit]
 context = 786_432
diff --git a/providers/edenai/models/flexai/gpt-oss-120b.toml b/providers/edenai/models/flexai/gpt-oss-120b.toml
index 74066689255..63d2eff1fa3 100644
--- a/providers/edenai/models/flexai/gpt-oss-120b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.039
-output = 0.1
+input = 0.037
+output = 0.17
diff --git a/providers/edenai/models/flexai/gpt-oss-20b.toml b/providers/edenai/models/flexai/gpt-oss-20b.toml
index a9511f536f9..cf8291af93e 100644
--- a/providers/edenai/models/flexai/gpt-oss-20b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-20b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.02
-output = 0.1
+input = 0.03
+output = 0.13
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index c8066cf5348..ae74bc38aee 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -5,5 +5,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75543
-output = 0.75543
+input = 0.75491
+output = 0.75491
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 2f0c218b10b..80db68aa28e 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17433
-output = 0.75543
+input = 0.17421
+output = 0.75491
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 5269ff6d619..e8d614f3c78 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46488
-output = 0.92976
+input = 0.46456
+output = 0.92912
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 6125d529272..50eaa6eb24f 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17433
-output = 0.69732
+input = 0.17421
+output = 0.69684
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index a86f8391e98..bb4a9388c3a 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04598
-output = 1.04598
+input = 1.04526
+output = 1.04526

From 198272be88f2d54bc8f29e81f573fa5d7854414c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:28:38 +0000
Subject: [PATCH 0857/2012] chore(sync): update OpenRouter model catalog
 (#6560)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5bc4e6af877..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0494
-output = 3.1482
-cache_read = 0.03498
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8119fb87874..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From 241e195fa9c9c3334dab51870bcd93ccf92a4128 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:29:08 +0000
Subject: [PATCH 0858/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6562)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/kimi-k2.6.toml | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/llmgateway/models/kimi-k2.6.toml b/providers/llmgateway/models/kimi-k2.6.toml
index 9d7d9a4940b..7e12ebbe10f 100644
--- a/providers/llmgateway/models/kimi-k2.6.toml
+++ b/providers/llmgateway/models/kimi-k2.6.toml
@@ -4,14 +4,14 @@
 # https://api.llmgateway.io/v1/models providers[].reasoning_efforts (accessed 2026-07-23)
 base_model = "moonshotai/kimi-k2.6"
 
-[[reasoning_options]]
-type = "effort" # API: {"reasoning_effort": "<v>"} or {"reasoning": {"effort": "<v>"}}
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
 [cost]
-input = 0.22
-output = 1.137
-cache_read = 0.048
+input = 0.6
+output = 3.05
+cache_read = 0.13

From 04e4c9f2a395af4f7666bb4719e7ee6b236829b0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:29:22 +0000
Subject: [PATCH 0859/2012] chore(sync): update LLM Gateway model catalog
 (#6561)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gonka24/kimi-k2.6.toml             | 22 -------------------
 1 file changed, 22 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/gonka24/kimi-k2.6.toml

diff --git a/providers/llmgateway-providers/models/gonka24/kimi-k2.6.toml b/providers/llmgateway-providers/models/gonka24/kimi-k2.6.toml
deleted file mode 100644
index c3ed4034327..00000000000
--- a/providers/llmgateway-providers/models/gonka24/kimi-k2.6.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-base_model = "moonshotai/kimi-k2.6"
-name = "Kimi K2.6 (Gonka24)"
-attachment = false
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.22
-output = 1.137
-cache_read = 0.048
-
-[limit]
-output = 98_304
-
-[modalities]
-input = ["text"]

From a63e41371cc203eb1b9fe8065e4d190df71d2cfe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:29:33 +0000
Subject: [PATCH 0860/2012] chore(sync): update Kilo model catalog (#6558)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 7b848afc9df..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From 04603485c85d12a2631127fcaa16cbad68c10dc5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 16:30:36 +0000
Subject: [PATCH 0861/2012] chore(sync): update NanoGPT model catalog (#6563)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/baseten/Kimi-K2-Instruct-FP4.toml  |  2 +-
 .../models/kimi-k2-instruct-fast.toml         |  2 +-
 .../nano-gpt/models/nex-agi/nex-n2-mini.toml  | 28 -------------------
 .../nano-gpt/models/nex-agi/nex-n2-pro.toml   | 28 -------------------
 ...emotron-3-nano-omni-30b-a3b-reasoning.toml | 16 -----------
 providers/nano-gpt/models/sarvam-105b.toml    |  6 ++--
 6 files changed, 5 insertions(+), 77 deletions(-)
 delete mode 100644 providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
 delete mode 100644 providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
 delete mode 100644 providers/nano-gpt/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning.toml

diff --git a/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml b/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
index c68686084a4..f23a348c265 100644
--- a/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
+++ b/providers/nano-gpt/models/baseten/Kimi-K2-Instruct-FP4.toml
@@ -5,7 +5,7 @@ release_date = "2024-01-01"
 last_updated = "2025-07-11"
 attachment = false
 reasoning = false
-tool_call = true
+tool_call = false
 structured_output = false
 open_weights = true
 
diff --git a/providers/nano-gpt/models/kimi-k2-instruct-fast.toml b/providers/nano-gpt/models/kimi-k2-instruct-fast.toml
index 93e059310df..d7999aabd8b 100644
--- a/providers/nano-gpt/models/kimi-k2-instruct-fast.toml
+++ b/providers/nano-gpt/models/kimi-k2-instruct-fast.toml
@@ -5,7 +5,7 @@ release_date = "2025-12-15"
 last_updated = "2025-07-15"
 attachment = true
 reasoning = false
-tool_call = true
+tool_call = false
 structured_output = false
 open_weights = true
 
diff --git a/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml b/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
deleted file mode 100644
index 0f217c50f68..00000000000
--- a/providers/nano-gpt/models/nex-agi/nex-n2-mini.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Nex N2 Mini"
-description = "Nex AGI's open-source agentic mixture-of-experts model in the Nex N2 family. It accepts text and image input and is built for coding, tool use, structured outputs, and optional reasoning with a 256K context window."
-family = "agi"
-release_date = "2026-06-24"
-last_updated = "2026-06-24"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
-
-[cost]
-input = 0.025
-output = 0.1
-cache_read = 0.0025
-
-[limit]
-context = 262_144
-input = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml b/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
deleted file mode 100644
index 4cb8ac34370..00000000000
--- a/providers/nano-gpt/models/nex-agi/nex-n2-pro.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "Nex N2 Pro"
-description = "Nex AGI's open-source agentic reasoning model, post-trained on Qwen3.5-397B-A17B. It is built for agentic coding, software engineering, deep research, tool use, and long-horizon tasks with a 256K context window."
-family = "agi"
-release_date = "2026-06-04"
-last_updated = "2026-06-04"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
-
-[cost]
-input = 0.5
-output = 2.5
-cache_read = 0.25
-
-[limit]
-context = 262_144
-input = 262_144
-output = 235_929
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning.toml b/providers/nano-gpt/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning.toml
deleted file mode 100644
index 092ddf387b9..00000000000
--- a/providers/nano-gpt/models/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-# Included in subscription
-base_model = "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning"
-name = "Nvidia Nemotron 3 Nano Omni"
-structured_output = false
-reasoning_options = []
-
-[cost]
-input = 0.105
-output = 0.42
-cache_read = 0.0525
-
-[limit]
-input = 256_000
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/sarvam-105b.toml b/providers/nano-gpt/models/sarvam-105b.toml
index 6731fdd9fce..75c1aa86b8a 100644
--- a/providers/nano-gpt/models/sarvam-105b.toml
+++ b/providers/nano-gpt/models/sarvam-105b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.045
-output = 0.177
-cache_read = 0.028
+input = 0.054
+output = 0.2124
+cache_read = 0.0336
 
 [limit]
 input = 131_072

From 5e11a84bd6550e6d9febaa2177b0ce3f4d839a25 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 16:30:39 +0000
Subject: [PATCH 0862/2012] chore(sync): update OpenRouter model catalog
 (#6564)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 ++++---
 providers/openrouter/models/tencent/hy3.toml               | 6 +++---
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..660f0f992c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.57948
+output = 1.73844
+cache_read = 0.057948
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 2ab476c47cfc935aebb24e65360ea8ce36f8e50d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 16:30:43 +0000
Subject: [PATCH 0863/2012] chore(sync): update Kilo model catalog (#6566)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 ++-
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 5ba5edc45ea13dc8d46c18bde7813bcde62896e6 Mon Sep 17 00:00:00 2001
From: JI4JUN <ji4jun0097@gmail.com>
Date: Wed, 9 Sep 2026 00:53:52 +0800
Subject: [PATCH 0864/2012] feat(302ai): sync models with latest featured
 catalog (#6497)

* feat(302ai): sync models with latest featured catalog

* fix(302ai): address review on featured catalog sync

- Remove image-generation SKU gemini-3.1-flash-lite-image (image rate unconfirmed)
- Drop unverified [interleaved] from MiniMax-M3
- Keep gemini-3.6-flash at 302.AI listed 1.5/7.5, cite doc.302.ai in-file

* docs(302ai): document GPT-5.6 Sol pricing vs OpenAI list

302.AI live catalog bills gpt-5.6-sol / gpt-5.6-sol-pro at $5/$30 per MTok
(GPT-5.5 rates, above OpenAI list $4/$20). Cite source in-file per review.

* fix(302ai): drop unverifiable SKUs, document verified reasoning controls

- Remove chat-latest (endpoint rejects plain completions; routing unverified)
- Remove mimo-v2-pro / mimo-v2-omni (IDs reject plain completions, -10003)
- provider.toml: rewrite relay comment distinguishing live-probed controls
  (OpenAI effort, Qwen toggle+budget, Kimi toggle) from lab-mapped ones
---
 providers/302ai/models/MiniMax-M2.1.toml      | 24 +++-----------
 providers/302ai/models/MiniMax-M2.5.toml      |  6 ++++
 .../302ai/models/MiniMax-M2.7-highspeed.toml  | 21 ------------
 providers/302ai/models/MiniMax-M2.7.toml      | 23 +++----------
 providers/302ai/models/MiniMax-M3.toml        |  7 ++++
 providers/302ai/models/chatgpt-4o-latest.toml | 23 -------------
 .../models/claude-3-5-haiku-20241022.toml     | 23 -------------
 .../302ai/models/claude-3-5-haiku-latest.toml | 23 -------------
 providers/302ai/models/claude-fable-5-1.toml  |  6 ++++
 providers/302ai/models/claude-fable-5.toml    |  6 ++++
 .../302ai/models/claude-opus-4-20250514.toml  | 24 --------------
 .../claude-opus-4-5-20251101-thinking.toml    | 23 -------------
 providers/302ai/models/claude-opus-4-5.toml   | 24 --------------
 .../models/claude-opus-4-6-thinking.toml      | 23 -------------
 providers/302ai/models/claude-opus-4-6.toml   | 24 --------------
 .../models/claude-opus-4-7-thinking.toml      |  7 ++++
 providers/302ai/models/claude-opus-4-8.toml   |  6 ++++
 .../302ai/models/claude-opus-5-thinking.toml  |  7 ++++
 providers/302ai/models/claude-opus-5.toml     |  6 ++++
 .../models/claude-sonnet-4-20250514.toml      | 24 --------------
 providers/302ai/models/claude-sonnet-4-5.toml | 24 --------------
 providers/302ai/models/claude-sonnet-5.toml   |  8 +++++
 providers/302ai/models/deepseek-chat.toml     | 23 -------------
 providers/302ai/models/deepseek-reasoner.toml | 24 --------------
 .../doubao-seed-code-preview-251028.toml      | 21 ------------
 .../models/gemini-3.1-flash-lite-preview.toml |  6 ++++
 .../302ai/models/gemini-3.1-flash-lite.toml   |  6 ++++
 .../302ai/models/gemini-3.1-pro-preview.toml  |  6 ++++
 .../302ai/models/gemini-3.5-flash-lite.toml   |  6 ++++
 .../models/gemini-3.5-flash-thinking.toml     |  7 ++++
 providers/302ai/models/gemini-3.5-flash.toml  |  6 ++++
 providers/302ai/models/gemini-3.6-flash.toml  |  7 ++++
 providers/302ai/models/gemini-3.7-flash.toml  |  6 ++++
 providers/302ai/models/gemini-3.8-flash.toml  |  6 ++++
 providers/302ai/models/glm-4.5-air.toml       | 24 --------------
 providers/302ai/models/glm-4.5-airx.toml      | 23 -------------
 providers/302ai/models/glm-4.5-x.toml         | 23 -------------
 providers/302ai/models/glm-4.7-flashx.toml    | 24 --------------
 providers/302ai/models/glm-5.1.toml           | 25 +++-----------
 providers/302ai/models/glm-5.2.toml           | 10 ++++++
 providers/302ai/models/glm-5.3-flash.toml     | 10 ++++++
 providers/302ai/models/glm-5.3.toml           | 10 ++++++
 providers/302ai/models/glm-for-coding.toml    | 23 -------------
 .../302ai/models/gpt-5.3-chat-latest.toml     |  5 +++
 .../302ai/models/gpt-5.4-mini-2026-03-17.toml | 26 ---------------
 .../302ai/models/gpt-5.4-nano-2026-03-17.toml | 26 ---------------
 providers/302ai/models/gpt-5.4-pro.toml       | 33 -------------------
 providers/302ai/models/gpt-5.5.toml           |  6 ++++
 providers/302ai/models/gpt-5.6-luna-pro.toml  |  8 +++++
 providers/302ai/models/gpt-5.6-luna.toml      |  6 ++++
 providers/302ai/models/gpt-5.6-sol-pro.toml   |  9 +++++
 providers/302ai/models/gpt-5.6-sol.toml       |  7 ++++
 providers/302ai/models/gpt-5.6-terra-pro.toml |  8 +++++
 providers/302ai/models/gpt-5.6-terra.toml     |  6 ++++
 providers/302ai/models/gpt-6-astra.toml       |  6 ++++
 .../models/grok-4-1-fast-non-reasoning.toml   | 22 -------------
 .../grok-4.20-beta-0309-non-reasoning.toml    | 21 ------------
 .../grok-4.20-multi-agent-beta-0309.toml      | 22 -------------
 providers/302ai/models/grok-4.3.toml          |  6 ++++
 providers/302ai/models/grok-4.5.toml          |  6 ++++
 providers/302ai/models/grok-4.6.toml          |  6 ++++
 .../302ai/models/kimi-k2-thinking-turbo.toml  | 23 -------------
 providers/302ai/models/kimi-k2.5.toml         | 10 ++++++
 providers/302ai/models/kimi-k2.6.toml         | 10 ++++++
 providers/302ai/models/kimi-k2.7-code.toml    |  9 +++++
 providers/302ai/models/kimi-k3.toml           | 11 +++++++
 providers/302ai/models/o3.toml                |  6 ++++
 providers/302ai/models/qwen-flash.toml        | 21 ------------
 providers/302ai/models/qwen-max-latest.toml   | 23 -------------
 providers/302ai/models/qwen-plus.toml         | 23 -------------
 providers/302ai/models/qwen3.5-35b-a3b.toml   |  8 +++++
 providers/302ai/models/qwen3.5-plus.toml      |  8 +++++
 providers/302ai/models/qwen3.6-35b-a3b.toml   |  8 +++++
 providers/302ai/models/qwen3.6-flash.toml     |  8 +++++
 providers/302ai/models/qwen3.6-plus.toml      |  8 +++++
 .../302ai/models/qwen3.7-max-2026-06-08.toml  |  9 +++++
 providers/302ai/models/qwen3.7-max.toml       |  8 +++++
 providers/302ai/models/qwen3.7-plus.toml      |  8 +++++
 providers/302ai/models/qwen3.8-flash.toml     | 16 +++++++++
 providers/302ai/models/qwen3.8-max.toml       | 16 +++++++++
 providers/302ai/provider.toml                 | 17 ++++++----
 81 files changed, 391 insertions(+), 746 deletions(-)
 create mode 100644 providers/302ai/models/MiniMax-M2.5.toml
 delete mode 100644 providers/302ai/models/MiniMax-M2.7-highspeed.toml
 create mode 100644 providers/302ai/models/MiniMax-M3.toml
 delete mode 100644 providers/302ai/models/chatgpt-4o-latest.toml
 delete mode 100644 providers/302ai/models/claude-3-5-haiku-20241022.toml
 delete mode 100644 providers/302ai/models/claude-3-5-haiku-latest.toml
 create mode 100644 providers/302ai/models/claude-fable-5-1.toml
 create mode 100644 providers/302ai/models/claude-fable-5.toml
 delete mode 100644 providers/302ai/models/claude-opus-4-20250514.toml
 delete mode 100644 providers/302ai/models/claude-opus-4-5-20251101-thinking.toml
 delete mode 100644 providers/302ai/models/claude-opus-4-5.toml
 delete mode 100644 providers/302ai/models/claude-opus-4-6-thinking.toml
 delete mode 100644 providers/302ai/models/claude-opus-4-6.toml
 create mode 100644 providers/302ai/models/claude-opus-4-7-thinking.toml
 create mode 100644 providers/302ai/models/claude-opus-4-8.toml
 create mode 100644 providers/302ai/models/claude-opus-5-thinking.toml
 create mode 100644 providers/302ai/models/claude-opus-5.toml
 delete mode 100644 providers/302ai/models/claude-sonnet-4-20250514.toml
 delete mode 100644 providers/302ai/models/claude-sonnet-4-5.toml
 create mode 100644 providers/302ai/models/claude-sonnet-5.toml
 delete mode 100644 providers/302ai/models/deepseek-chat.toml
 delete mode 100644 providers/302ai/models/deepseek-reasoner.toml
 delete mode 100644 providers/302ai/models/doubao-seed-code-preview-251028.toml
 create mode 100644 providers/302ai/models/gemini-3.1-flash-lite-preview.toml
 create mode 100644 providers/302ai/models/gemini-3.1-flash-lite.toml
 create mode 100644 providers/302ai/models/gemini-3.1-pro-preview.toml
 create mode 100644 providers/302ai/models/gemini-3.5-flash-lite.toml
 create mode 100644 providers/302ai/models/gemini-3.5-flash-thinking.toml
 create mode 100644 providers/302ai/models/gemini-3.5-flash.toml
 create mode 100644 providers/302ai/models/gemini-3.6-flash.toml
 create mode 100644 providers/302ai/models/gemini-3.7-flash.toml
 create mode 100644 providers/302ai/models/gemini-3.8-flash.toml
 delete mode 100644 providers/302ai/models/glm-4.5-air.toml
 delete mode 100644 providers/302ai/models/glm-4.5-airx.toml
 delete mode 100644 providers/302ai/models/glm-4.5-x.toml
 delete mode 100644 providers/302ai/models/glm-4.7-flashx.toml
 create mode 100644 providers/302ai/models/glm-5.2.toml
 create mode 100644 providers/302ai/models/glm-5.3-flash.toml
 create mode 100644 providers/302ai/models/glm-5.3.toml
 delete mode 100644 providers/302ai/models/glm-for-coding.toml
 create mode 100644 providers/302ai/models/gpt-5.3-chat-latest.toml
 delete mode 100644 providers/302ai/models/gpt-5.4-mini-2026-03-17.toml
 delete mode 100644 providers/302ai/models/gpt-5.4-nano-2026-03-17.toml
 delete mode 100644 providers/302ai/models/gpt-5.4-pro.toml
 create mode 100644 providers/302ai/models/gpt-5.5.toml
 create mode 100644 providers/302ai/models/gpt-5.6-luna-pro.toml
 create mode 100644 providers/302ai/models/gpt-5.6-luna.toml
 create mode 100644 providers/302ai/models/gpt-5.6-sol-pro.toml
 create mode 100644 providers/302ai/models/gpt-5.6-sol.toml
 create mode 100644 providers/302ai/models/gpt-5.6-terra-pro.toml
 create mode 100644 providers/302ai/models/gpt-5.6-terra.toml
 create mode 100644 providers/302ai/models/gpt-6-astra.toml
 delete mode 100644 providers/302ai/models/grok-4-1-fast-non-reasoning.toml
 delete mode 100644 providers/302ai/models/grok-4.20-beta-0309-non-reasoning.toml
 delete mode 100644 providers/302ai/models/grok-4.20-multi-agent-beta-0309.toml
 create mode 100644 providers/302ai/models/grok-4.3.toml
 create mode 100644 providers/302ai/models/grok-4.5.toml
 create mode 100644 providers/302ai/models/grok-4.6.toml
 delete mode 100644 providers/302ai/models/kimi-k2-thinking-turbo.toml
 create mode 100644 providers/302ai/models/kimi-k2.5.toml
 create mode 100644 providers/302ai/models/kimi-k2.6.toml
 create mode 100644 providers/302ai/models/kimi-k2.7-code.toml
 create mode 100644 providers/302ai/models/kimi-k3.toml
 create mode 100644 providers/302ai/models/o3.toml
 delete mode 100644 providers/302ai/models/qwen-flash.toml
 delete mode 100644 providers/302ai/models/qwen-max-latest.toml
 delete mode 100644 providers/302ai/models/qwen-plus.toml
 create mode 100644 providers/302ai/models/qwen3.5-35b-a3b.toml
 create mode 100644 providers/302ai/models/qwen3.5-plus.toml
 create mode 100644 providers/302ai/models/qwen3.6-35b-a3b.toml
 create mode 100644 providers/302ai/models/qwen3.6-flash.toml
 create mode 100644 providers/302ai/models/qwen3.6-plus.toml
 create mode 100644 providers/302ai/models/qwen3.7-max-2026-06-08.toml
 create mode 100644 providers/302ai/models/qwen3.7-max.toml
 create mode 100644 providers/302ai/models/qwen3.7-plus.toml
 create mode 100644 providers/302ai/models/qwen3.8-flash.toml
 create mode 100644 providers/302ai/models/qwen3.8-max.toml

diff --git a/providers/302ai/models/MiniMax-M2.1.toml b/providers/302ai/models/MiniMax-M2.1.toml
index 3c1fe4d34b9..bce1128bdc8 100644
--- a/providers/302ai/models/MiniMax-M2.1.toml
+++ b/providers/302ai/models/MiniMax-M2.1.toml
@@ -1,21 +1,7 @@
-name = "MiniMax-M2.1"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-release_date = "2025-12-19"
-last_updated = "2025-12-19"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
+# 302.AI API reports context_length = 204,800 for MiniMax-M2.1 (reconciled from prior 1M override)
+base_model = "minimax/MiniMax-M2.1"
+reasoning_options = []
 
 [cost]
-input = 0.300
-output = 1.200
-
-[limit]
-context = 1_000_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
+input = 0.3
+output = 1.2
diff --git a/providers/302ai/models/MiniMax-M2.5.toml b/providers/302ai/models/MiniMax-M2.5.toml
new file mode 100644
index 00000000000..67b638180ff
--- /dev/null
+++ b/providers/302ai/models/MiniMax-M2.5.toml
@@ -0,0 +1,6 @@
+base_model = "minimax/MiniMax-M2.5"
+reasoning_options = []
+
+[cost]
+input = 0.3
+output = 1.2
diff --git a/providers/302ai/models/MiniMax-M2.7-highspeed.toml b/providers/302ai/models/MiniMax-M2.7-highspeed.toml
deleted file mode 100644
index 80b702b6ad6..00000000000
--- a/providers/302ai/models/MiniMax-M2.7-highspeed.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-name = "MiniMax-M2.7-highspeed"
-description = "High-speed MiniMax model for low-latency coding and agent workflows"
-release_date = "2026-03-19"
-last_updated = "2026-03-19"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.600
-output = 4.800
-
-[limit]
-context = 204_800
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/MiniMax-M2.7.toml b/providers/302ai/models/MiniMax-M2.7.toml
index 3a53aaa154e..81563bf573e 100644
--- a/providers/302ai/models/MiniMax-M2.7.toml
+++ b/providers/302ai/models/MiniMax-M2.7.toml
@@ -1,21 +1,6 @@
-name = "MiniMax-M2.7"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-release_date = "2026-03-19"
-last_updated = "2026-03-19"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
+base_model = "minimax/MiniMax-M2.7"
+reasoning_options = []
 
 [cost]
-input = 0.300
-output = 1.200
-
-[limit]
-context = 204_800
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
+input = 0.3
+output = 1.2
diff --git a/providers/302ai/models/MiniMax-M3.toml b/providers/302ai/models/MiniMax-M3.toml
new file mode 100644
index 00000000000..be4210ae4ac
--- /dev/null
+++ b/providers/302ai/models/MiniMax-M3.toml
@@ -0,0 +1,7 @@
+# Toggle: thinking.type = enabled|disabled
+base_model = "minimax/MiniMax-M3"
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0.72
+output = 2.88
diff --git a/providers/302ai/models/chatgpt-4o-latest.toml b/providers/302ai/models/chatgpt-4o-latest.toml
deleted file mode 100644
index e1ed768d611..00000000000
--- a/providers/302ai/models/chatgpt-4o-latest.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "chatgpt-4o-latest"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2024-08-08"
-last_updated = "2024-08-08"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-open_weights = false
-knowledge = "2023-09"
-
-[cost]
-input = 5.000
-output = 15.000
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-3-5-haiku-20241022.toml b/providers/302ai/models/claude-3-5-haiku-20241022.toml
deleted file mode 100644
index 4187b860dda..00000000000
--- a/providers/302ai/models/claude-3-5-haiku-20241022.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "claude-3-5-haiku-20241022"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2024-10-22"
-last_updated = "2024-10-22"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-07-31"
-
-[cost]
-input = 0.800
-output = 4.000
-
-[limit]
-context = 200_000
-output = 8_192
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-3-5-haiku-latest.toml b/providers/302ai/models/claude-3-5-haiku-latest.toml
deleted file mode 100644
index 38e55c29625..00000000000
--- a/providers/302ai/models/claude-3-5-haiku-latest.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "claude-3-5-haiku-latest"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2024-10-22"
-last_updated = "2024-10-22"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-07-31"
-
-[cost]
-input = 0.800
-output = 4.000
-
-[limit]
-context = 200_000
-output = 8_192
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-fable-5-1.toml b/providers/302ai/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..8484f21020a
--- /dev/null
+++ b/providers/302ai/models/claude-fable-5-1.toml
@@ -0,0 +1,6 @@
+base_model = "anthropic/claude-fable-5-1"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.0
+output = 50.0
diff --git a/providers/302ai/models/claude-fable-5.toml b/providers/302ai/models/claude-fable-5.toml
new file mode 100644
index 00000000000..01a493e716f
--- /dev/null
+++ b/providers/302ai/models/claude-fable-5.toml
@@ -0,0 +1,6 @@
+base_model = "anthropic/claude-fable-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.0
+output = 50.0
diff --git a/providers/302ai/models/claude-opus-4-20250514.toml b/providers/302ai/models/claude-opus-4-20250514.toml
deleted file mode 100644
index 8a972a7f74a..00000000000
--- a/providers/302ai/models/claude-opus-4-20250514.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "claude-opus-4-20250514"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-05-22"
-last_updated = "2025-05-22"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 31999 }]
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-03-31"
-
-[cost]
-input = 15.000
-output = 75.000
-
-[limit]
-context = 200_000
-output = 32_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-opus-4-5-20251101-thinking.toml b/providers/302ai/models/claude-opus-4-5-20251101-thinking.toml
deleted file mode 100644
index 53ab746245c..00000000000
--- a/providers/302ai/models/claude-opus-4-5-20251101-thinking.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "claude-opus-4-5-20251101-thinking"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-release_date = "2025-11-25"
-last_updated = "2025-11-25"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-03"
-
-[cost]
-input = 5.000
-output = 25.000
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-opus-4-5.toml b/providers/302ai/models/claude-opus-4-5.toml
deleted file mode 100644
index 70bdeac12a1..00000000000
--- a/providers/302ai/models/claude-opus-4-5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "claude-opus-4-5"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-11-25"
-last_updated = "2025-11-25"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1024, max = 63999 }]
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-03-31"
-
-[cost]
-input = 5.000
-output = 25.000
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-opus-4-6-thinking.toml b/providers/302ai/models/claude-opus-4-6-thinking.toml
deleted file mode 100644
index 15d674f84b3..00000000000
--- a/providers/302ai/models/claude-opus-4-6-thinking.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "claude-opus-4-6-thinking"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-release_date = "2026-02-06"
-last_updated = "2026-03-13"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-05"
-
-[cost]
-input = 5.000
-output = 25.000
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-opus-4-6.toml b/providers/302ai/models/claude-opus-4-6.toml
deleted file mode 100644
index cb04746dff2..00000000000
--- a/providers/302ai/models/claude-opus-4-6.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "claude-opus-4-6"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-02-06"
-last_updated = "2026-03-13"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1024, max = 127999 }]
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-05-31"
-
-[cost]
-input = 5.000
-output = 25.000
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-opus-4-7-thinking.toml b/providers/302ai/models/claude-opus-4-7-thinking.toml
new file mode 100644
index 00000000000..c8490e371a8
--- /dev/null
+++ b/providers/302ai/models/claude-opus-4-7-thinking.toml
@@ -0,0 +1,7 @@
+base_model = "anthropic/claude-opus-4-7"
+name = "claude-opus-4-7-thinking"
+reasoning_options = []
+
+[cost]
+input = 5.0
+output = 25.0
diff --git a/providers/302ai/models/claude-opus-4-8.toml b/providers/302ai/models/claude-opus-4-8.toml
new file mode 100644
index 00000000000..aafe7fb3d26
--- /dev/null
+++ b/providers/302ai/models/claude-opus-4-8.toml
@@ -0,0 +1,6 @@
+base_model = "anthropic/claude-opus-4-8"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5.0
+output = 25.0
diff --git a/providers/302ai/models/claude-opus-5-thinking.toml b/providers/302ai/models/claude-opus-5-thinking.toml
new file mode 100644
index 00000000000..f1155ed03eb
--- /dev/null
+++ b/providers/302ai/models/claude-opus-5-thinking.toml
@@ -0,0 +1,7 @@
+base_model = "anthropic/claude-opus-5"
+name = "claude-opus-5-thinking"
+reasoning_options = []
+
+[cost]
+input = 5.0
+output = 25.0
diff --git a/providers/302ai/models/claude-opus-5.toml b/providers/302ai/models/claude-opus-5.toml
new file mode 100644
index 00000000000..31cd0633b58
--- /dev/null
+++ b/providers/302ai/models/claude-opus-5.toml
@@ -0,0 +1,6 @@
+base_model = "anthropic/claude-opus-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5.0
+output = 25.0
diff --git a/providers/302ai/models/claude-sonnet-4-20250514.toml b/providers/302ai/models/claude-sonnet-4-20250514.toml
deleted file mode 100644
index 0e53f6ef461..00000000000
--- a/providers/302ai/models/claude-sonnet-4-20250514.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "claude-sonnet-4-20250514"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-05-22"
-last_updated = "2025-05-22"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 63999 }]
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-03-31"
-
-[cost]
-input = 3.000
-output = 15.000
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-sonnet-4-5.toml b/providers/302ai/models/claude-sonnet-4-5.toml
deleted file mode 100644
index 9c5655b9568..00000000000
--- a/providers/302ai/models/claude-sonnet-4-5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "claude-sonnet-4-5"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-09-30"
-last_updated = "2025-09-30"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 63999 }]
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-07-31"
-
-[cost]
-input = 3.000
-output = 15.000
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/302ai/models/claude-sonnet-5.toml b/providers/302ai/models/claude-sonnet-5.toml
new file mode 100644
index 00000000000..7834d37a87f
--- /dev/null
+++ b/providers/302ai/models/claude-sonnet-5.toml
@@ -0,0 +1,8 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: output_config.effort = low|medium|high|xhigh|max
+base_model = "anthropic/claude-sonnet-5"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.0
+output = 10.0
diff --git a/providers/302ai/models/deepseek-chat.toml b/providers/302ai/models/deepseek-chat.toml
deleted file mode 100644
index 2b55c7e43a7..00000000000
--- a/providers/302ai/models/deepseek-chat.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Deepseek-Chat"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-family = "deepseek"
-release_date = "2024-11-29"
-last_updated = "2024-11-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-07"
-
-[cost]
-input = 0.290
-output = 0.430
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/deepseek-reasoner.toml b/providers/302ai/models/deepseek-reasoner.toml
deleted file mode 100644
index 448bf2bd4b7..00000000000
--- a/providers/302ai/models/deepseek-reasoner.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Deepseek-Reasoner"
-description = "DeepSeek reasoning model for multi-step analysis, math, coding, and tools"
-family = "deepseek-thinking"
-release_date = "2025-01-20"
-last_updated = "2025-01-20"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-07"
-
-[cost]
-input = 0.290
-output = 0.430
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/doubao-seed-code-preview-251028.toml b/providers/302ai/models/doubao-seed-code-preview-251028.toml
deleted file mode 100644
index 0d2128ff32c..00000000000
--- a/providers/302ai/models/doubao-seed-code-preview-251028.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-name = "doubao-seed-code-preview-251028"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-release_date = "2025-11-11"
-last_updated = "2025-11-11"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.170
-output = 1.140
-
-[limit]
-context = 256_000
-output = 32_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/gemini-3.1-flash-lite-preview.toml b/providers/302ai/models/gemini-3.1-flash-lite-preview.toml
new file mode 100644
index 00000000000..f96a23289ae
--- /dev/null
+++ b/providers/302ai/models/gemini-3.1-flash-lite-preview.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.1-flash-lite-preview"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.25
+output = 1.5
diff --git a/providers/302ai/models/gemini-3.1-flash-lite.toml b/providers/302ai/models/gemini-3.1-flash-lite.toml
new file mode 100644
index 00000000000..67ee9bbcaab
--- /dev/null
+++ b/providers/302ai/models/gemini-3.1-flash-lite.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.1-flash-lite"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.25
+output = 1.5
diff --git a/providers/302ai/models/gemini-3.1-pro-preview.toml b/providers/302ai/models/gemini-3.1-pro-preview.toml
new file mode 100644
index 00000000000..24eca3818c9
--- /dev/null
+++ b/providers/302ai/models/gemini-3.1-pro-preview.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.1-pro-preview"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2.0
+output = 12.0
diff --git a/providers/302ai/models/gemini-3.5-flash-lite.toml b/providers/302ai/models/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..43a2449ec76
--- /dev/null
+++ b/providers/302ai/models/gemini-3.5-flash-lite.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.5-flash-lite"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.3
+output = 2.5
diff --git a/providers/302ai/models/gemini-3.5-flash-thinking.toml b/providers/302ai/models/gemini-3.5-flash-thinking.toml
new file mode 100644
index 00000000000..4a42abeb130
--- /dev/null
+++ b/providers/302ai/models/gemini-3.5-flash-thinking.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemini-3.5-flash"
+name = "gemini-3.5-flash-thinking"
+reasoning_options = []
+
+[cost]
+input = 1.5
+output = 9.0
diff --git a/providers/302ai/models/gemini-3.5-flash.toml b/providers/302ai/models/gemini-3.5-flash.toml
new file mode 100644
index 00000000000..dd0bb8ad627
--- /dev/null
+++ b/providers/302ai/models/gemini-3.5-flash.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.5-flash"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 1.5
+output = 9.0
diff --git a/providers/302ai/models/gemini-3.6-flash.toml b/providers/302ai/models/gemini-3.6-flash.toml
new file mode 100644
index 00000000000..aa263aba2e2
--- /dev/null
+++ b/providers/302ai/models/gemini-3.6-flash.toml
@@ -0,0 +1,7 @@
+# Cost: 302.AI listed price, https://doc.302.ai/ (retrieved 2026-09-08)
+base_model = "google/gemini-3.6-flash"
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 1.5
+output = 7.5
diff --git a/providers/302ai/models/gemini-3.7-flash.toml b/providers/302ai/models/gemini-3.7-flash.toml
new file mode 100644
index 00000000000..0db1eb9f1de
--- /dev/null
+++ b/providers/302ai/models/gemini-3.7-flash.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.7-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.75
+output = 3.75
diff --git a/providers/302ai/models/gemini-3.8-flash.toml b/providers/302ai/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..256bd86ef41
--- /dev/null
+++ b/providers/302ai/models/gemini-3.8-flash.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemini-3.8-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.75
+output = 3.75
diff --git a/providers/302ai/models/glm-4.5-air.toml b/providers/302ai/models/glm-4.5-air.toml
deleted file mode 100644
index de558b88636..00000000000
--- a/providers/302ai/models/glm-4.5-air.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "glm-4.5-air"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm-air"
-release_date = "2025-07-29"
-last_updated = "2025-07-29"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-knowledge = "2025-04"
-
-[cost]
-input = 0.1143
-output = 0.286
-
-[limit]
-context = 131_072
-output = 98_304
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/glm-4.5-airx.toml b/providers/302ai/models/glm-4.5-airx.toml
deleted file mode 100644
index f547de2fa79..00000000000
--- a/providers/302ai/models/glm-4.5-airx.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "glm-4.5-airx"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2025-07-29"
-last_updated = "2025-07-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-04"
-
-[cost]
-input = 0.572
-output = 1.714
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/glm-4.5-x.toml b/providers/302ai/models/glm-4.5-x.toml
deleted file mode 100644
index 2564e3d0fe6..00000000000
--- a/providers/302ai/models/glm-4.5-x.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "glm-4.5-x"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2025-07-29"
-last_updated = "2025-07-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-04"
-
-[cost]
-input = 1.143
-output = 2.290
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/glm-4.7-flashx.toml b/providers/302ai/models/glm-4.7-flashx.toml
deleted file mode 100644
index 3218c5862b3..00000000000
--- a/providers/302ai/models/glm-4.7-flashx.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "glm-4.7-flashx"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm-flash"
-release_date = "2026-01-20"
-last_updated = "2026-01-20"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = true
-knowledge = "2025-04"
-
-[cost]
-input = 0.0715
-output = 0.429
-
-[limit]
-context = 200_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/glm-5.1.toml b/providers/302ai/models/glm-5.1.toml
index e5de89fd2a8..d9abb545ec2 100644
--- a/providers/302ai/models/glm-5.1.toml
+++ b/providers/302ai/models/glm-5.1.toml
@@ -1,27 +1,10 @@
-name = "glm-5.1"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2026-04-10"
-last_updated = "2026-04-10"
-attachment = false
-reasoning = true
+# Toggle: thinking.type = enabled|disabled
+base_model = "zhipuai/glm-5.1"
 reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.860
-output = 3.500
-
-[limit]
-context = 200_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
+input = 1.4
+output = 4.4
diff --git a/providers/302ai/models/glm-5.2.toml b/providers/302ai/models/glm-5.2.toml
new file mode 100644
index 00000000000..1bc0f7d91c8
--- /dev/null
+++ b/providers/302ai/models/glm-5.2.toml
@@ -0,0 +1,10 @@
+# Effort: reasoning_effort = high|max
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
diff --git a/providers/302ai/models/glm-5.3-flash.toml b/providers/302ai/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..ae7be522962
--- /dev/null
+++ b/providers/302ai/models/glm-5.3-flash.toml
@@ -0,0 +1,10 @@
+# Effort: reasoning_effort = low|high|max
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.075
+output = 0.25
diff --git a/providers/302ai/models/glm-5.3.toml b/providers/302ai/models/glm-5.3.toml
new file mode 100644
index 00000000000..d8ee5e7385a
--- /dev/null
+++ b/providers/302ai/models/glm-5.3.toml
@@ -0,0 +1,10 @@
+# Effort: reasoning_effort = low|high|max
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
diff --git a/providers/302ai/models/glm-for-coding.toml b/providers/302ai/models/glm-for-coding.toml
deleted file mode 100644
index 7ba4c169e01..00000000000
--- a/providers/302ai/models/glm-for-coding.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "glm-for-coding"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2025-09-30"
-last_updated = "2025-09-30"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.086
-output = 0.343
-
-[limit]
-context = 200_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/gpt-5.3-chat-latest.toml b/providers/302ai/models/gpt-5.3-chat-latest.toml
new file mode 100644
index 00000000000..c83996b77a9
--- /dev/null
+++ b/providers/302ai/models/gpt-5.3-chat-latest.toml
@@ -0,0 +1,5 @@
+base_model = "openai/gpt-5.3-chat-latest"
+
+[cost]
+input = 1.75
+output = 14.0
diff --git a/providers/302ai/models/gpt-5.4-mini-2026-03-17.toml b/providers/302ai/models/gpt-5.4-mini-2026-03-17.toml
deleted file mode 100644
index e45a9c5a5e3..00000000000
--- a/providers/302ai/models/gpt-5.4-mini-2026-03-17.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "gpt-5.4-mini-2026-03-17"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt-mini"
-release_date = "2026-03-19"
-last_updated = "2026-03-19"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-knowledge = "2025-08-31"
-
-[cost]
-input = 0.750
-output = 4.500
-
-[limit]
-context = 400_000
-input = 272_000
-output = 128_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/gpt-5.4-nano-2026-03-17.toml b/providers/302ai/models/gpt-5.4-nano-2026-03-17.toml
deleted file mode 100644
index eada76b28a2..00000000000
--- a/providers/302ai/models/gpt-5.4-nano-2026-03-17.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-name = "gpt-5.4-nano-2026-03-17"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt-nano"
-release_date = "2026-03-19"
-last_updated = "2026-03-19"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-knowledge = "2025-08-31"
-
-[cost]
-input = 0.200
-output = 1.250
-
-[limit]
-context = 400_000
-input = 272_000
-output = 128_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/gpt-5.4-pro.toml b/providers/302ai/models/gpt-5.4-pro.toml
deleted file mode 100644
index 4cd3385ee29..00000000000
--- a/providers/302ai/models/gpt-5.4-pro.toml
+++ /dev/null
@@ -1,33 +0,0 @@
-name = "gpt-5.4-pro"
-description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
-family = "gpt-pro"
-release_date = "2026-03-05"
-last_updated = "2026-03-05"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
-temperature = false
-tool_call = true
-structured_output = false
-open_weights = false
-knowledge = "2025-08-31"
-
-[cost]
-input = 30.000
-output = 180.000
-cache_read = 0
-cache_write = 0
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 60.000
-output = 270.000
-
-[limit]
-context = 1_050_000
-input = 922_000
-output = 128_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/gpt-5.5.toml b/providers/302ai/models/gpt-5.5.toml
new file mode 100644
index 00000000000..b4c4f7e0f22
--- /dev/null
+++ b/providers/302ai/models/gpt-5.5.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-5.5"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 5.0
+output = 30.0
diff --git a/providers/302ai/models/gpt-5.6-luna-pro.toml b/providers/302ai/models/gpt-5.6-luna-pro.toml
new file mode 100644
index 00000000000..0ab31b46ca9
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-luna-pro.toml
@@ -0,0 +1,8 @@
+# 302.AI model SKU routing to GPT-5.6 Pro reasoning tier (published at same token pricing in 302.AI API catalog)
+base_model = "openai/gpt-5.6-luna"
+name = "gpt-5.6-luna-pro"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.2
+output = 1.2
diff --git a/providers/302ai/models/gpt-5.6-luna.toml b/providers/302ai/models/gpt-5.6-luna.toml
new file mode 100644
index 00000000000..578fd4f5146
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-luna.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-5.6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.2
+output = 1.2
diff --git a/providers/302ai/models/gpt-5.6-sol-pro.toml b/providers/302ai/models/gpt-5.6-sol-pro.toml
new file mode 100644
index 00000000000..545e5b69ac5
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-sol-pro.toml
@@ -0,0 +1,9 @@
+# Cost: 302.AI bills Sol Pro at GPT-5.5 rates ($5.000/$30.000 per MTok), per live catalog https://api.302.ai/v1/models (retrieved 2026-09-08)
+# 302.AI model SKU routing to GPT-5.6 Pro reasoning tier (published at same token pricing in 302.AI API catalog)
+base_model = "openai/gpt-5.6-sol"
+name = "gpt-5.6-sol-pro"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5.0
+output = 30.0
diff --git a/providers/302ai/models/gpt-5.6-sol.toml b/providers/302ai/models/gpt-5.6-sol.toml
new file mode 100644
index 00000000000..20133f6c2f9
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-sol.toml
@@ -0,0 +1,7 @@
+# Cost: 302.AI bills Sol at GPT-5.5 rates ($5.000/$30.000 per MTok, intentionally above OpenAI list $4/$20), per live catalog https://api.302.ai/v1/models (retrieved 2026-09-08)
+base_model = "openai/gpt-5.6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5.0
+output = 30.0
diff --git a/providers/302ai/models/gpt-5.6-terra-pro.toml b/providers/302ai/models/gpt-5.6-terra-pro.toml
new file mode 100644
index 00000000000..2a80d5735cd
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-terra-pro.toml
@@ -0,0 +1,8 @@
+# 302.AI model SKU routing to GPT-5.6 Pro reasoning tier (published at same token pricing in 302.AI API catalog)
+base_model = "openai/gpt-5.6-terra"
+name = "gpt-5.6-terra-pro"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.0
+output = 12.0
diff --git a/providers/302ai/models/gpt-5.6-terra.toml b/providers/302ai/models/gpt-5.6-terra.toml
new file mode 100644
index 00000000000..7732618c97c
--- /dev/null
+++ b/providers/302ai/models/gpt-5.6-terra.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-5.6-terra"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.0
+output = 12.0
diff --git a/providers/302ai/models/gpt-6-astra.toml b/providers/302ai/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..015f6c3436f
--- /dev/null
+++ b/providers/302ai/models/gpt-6-astra.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10.0
+output = 50.0
diff --git a/providers/302ai/models/grok-4-1-fast-non-reasoning.toml b/providers/302ai/models/grok-4-1-fast-non-reasoning.toml
deleted file mode 100644
index cf2d24ec30b..00000000000
--- a/providers/302ai/models/grok-4-1-fast-non-reasoning.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "grok-4-1-fast-non-reasoning"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2025-11-20"
-last_updated = "2025-11-20"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-06"
-
-[cost]
-input = 0.200
-output = 0.500
-
-[limit]
-context = 2_000_000
-output = 30_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/grok-4.20-beta-0309-non-reasoning.toml b/providers/302ai/models/grok-4.20-beta-0309-non-reasoning.toml
deleted file mode 100644
index eb5149eb845..00000000000
--- a/providers/302ai/models/grok-4.20-beta-0309-non-reasoning.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-name = "grok-4.20-beta-0309-non-reasoning"
-description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-release_date = "2026-03-16"
-last_updated = "2026-03-16"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 2.000
-output = 6.000
-
-[limit]
-context = 2_000_000
-output = 30_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/grok-4.20-multi-agent-beta-0309.toml b/providers/302ai/models/grok-4.20-multi-agent-beta-0309.toml
deleted file mode 100644
index f70f1911f2e..00000000000
--- a/providers/302ai/models/grok-4.20-multi-agent-beta-0309.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "grok-4.20-multi-agent-beta-0309"
-description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-release_date = "2026-03-16"
-last_updated = "2026-03-16"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 2.000
-output = 6.000
-
-[limit]
-context = 2_000_000
-output = 30_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/302ai/models/grok-4.3.toml b/providers/302ai/models/grok-4.3.toml
new file mode 100644
index 00000000000..70000f8b6e7
--- /dev/null
+++ b/providers/302ai/models/grok-4.3.toml
@@ -0,0 +1,6 @@
+base_model = "xai/grok-4.3"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[cost]
+input = 1.25
+output = 2.5
diff --git a/providers/302ai/models/grok-4.5.toml b/providers/302ai/models/grok-4.5.toml
new file mode 100644
index 00000000000..bb8d4854b50
--- /dev/null
+++ b/providers/302ai/models/grok-4.5.toml
@@ -0,0 +1,6 @@
+base_model = "xai/grok-4.5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2.0
+output = 6.0
diff --git a/providers/302ai/models/grok-4.6.toml b/providers/302ai/models/grok-4.6.toml
new file mode 100644
index 00000000000..1189356448a
--- /dev/null
+++ b/providers/302ai/models/grok-4.6.toml
@@ -0,0 +1,6 @@
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2.0
+output = 6.0
diff --git a/providers/302ai/models/kimi-k2-thinking-turbo.toml b/providers/302ai/models/kimi-k2-thinking-turbo.toml
deleted file mode 100644
index 77bf3b84325..00000000000
--- a/providers/302ai/models/kimi-k2-thinking-turbo.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "kimi-k2-thinking-turbo"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
-release_date = "2025-09-05"
-last_updated = "2025-09-05"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2025-06"
-
-[cost]
-input = 1.265
-output = 9.119
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/kimi-k2.5.toml b/providers/302ai/models/kimi-k2.5.toml
new file mode 100644
index 00000000000..c482a507010
--- /dev/null
+++ b/providers/302ai/models/kimi-k2.5.toml
@@ -0,0 +1,10 @@
+# Toggle: thinking.type = enabled|disabled
+base_model = "moonshotai/kimi-k2.5"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.66
+output = 3.3
diff --git a/providers/302ai/models/kimi-k2.6.toml b/providers/302ai/models/kimi-k2.6.toml
new file mode 100644
index 00000000000..b528d6ec1ac
--- /dev/null
+++ b/providers/302ai/models/kimi-k2.6.toml
@@ -0,0 +1,10 @@
+# Toggle: thinking.type = enabled|disabled
+base_model = "moonshotai/kimi-k2.6"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.95
+output = 4.0
diff --git a/providers/302ai/models/kimi-k2.7-code.toml b/providers/302ai/models/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..10e5c19b665
--- /dev/null
+++ b/providers/302ai/models/kimi-k2.7-code.toml
@@ -0,0 +1,9 @@
+base_model = "moonshotai/kimi-k2.7-code"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.95
+output = 4.0
diff --git a/providers/302ai/models/kimi-k3.toml b/providers/302ai/models/kimi-k3.toml
new file mode 100644
index 00000000000..58faa62ebbd
--- /dev/null
+++ b/providers/302ai/models/kimi-k3.toml
@@ -0,0 +1,11 @@
+# Toggle: thinking.type = enabled|disabled
+# Effort: output_config.effort = low|high|max
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 3.0
+output = 15.0
diff --git a/providers/302ai/models/o3.toml b/providers/302ai/models/o3.toml
new file mode 100644
index 00000000000..9293fda41fe
--- /dev/null
+++ b/providers/302ai/models/o3.toml
@@ -0,0 +1,6 @@
+base_model = "openai/o3"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 2.0
+output = 8.0
diff --git a/providers/302ai/models/qwen-flash.toml b/providers/302ai/models/qwen-flash.toml
deleted file mode 100644
index fc1229a8733..00000000000
--- a/providers/302ai/models/qwen-flash.toml
+++ /dev/null
@@ -1,21 +0,0 @@
-name = "Qwen-Flash"
-description = "Efficient Qwen model for fast chat, extraction, and high-volume workloads"
-release_date = "2025-07-28"
-last_updated = "2025-07-28"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.022
-output = 0.220
-
-[limit]
-context = 1_000_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/qwen-max-latest.toml b/providers/302ai/models/qwen-max-latest.toml
deleted file mode 100644
index 24b7786be31..00000000000
--- a/providers/302ai/models/qwen-max-latest.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen-Max-Latest"
-description = "Flagship Qwen model for complex reasoning, coding, and agentic workflows"
-family = "qwen"
-release_date = "2024-04-03"
-last_updated = "2025-01-25"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-11"
-
-[cost]
-input = 0.343
-output = 1.372
-
-[limit]
-context = 131_072
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/qwen-plus.toml b/providers/302ai/models/qwen-plus.toml
deleted file mode 100644
index dd7d49fcc52..00000000000
--- a/providers/302ai/models/qwen-plus.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen-Plus"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-10"
-
-[cost]
-input = 0.120
-output = 1.200
-
-[limit]
-context = 1_000_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/302ai/models/qwen3.5-35b-a3b.toml b/providers/302ai/models/qwen3.5-35b-a3b.toml
new file mode 100644
index 00000000000..1b1a4828188
--- /dev/null
+++ b/providers/302ai/models/qwen3.5-35b-a3b.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.5-35b-a3b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.06
+output = 0.46
diff --git a/providers/302ai/models/qwen3.5-plus.toml b/providers/302ai/models/qwen3.5-plus.toml
new file mode 100644
index 00000000000..c1ad4c7c8b2
--- /dev/null
+++ b/providers/302ai/models/qwen3.5-plus.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.5-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.12
+output = 0.69
diff --git a/providers/302ai/models/qwen3.6-35b-a3b.toml b/providers/302ai/models/qwen3.6-35b-a3b.toml
new file mode 100644
index 00000000000..6c5d231a6d8
--- /dev/null
+++ b/providers/302ai/models/qwen3.6-35b-a3b.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.6-35b-a3b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.283
+output = 1.705
diff --git a/providers/302ai/models/qwen3.6-flash.toml b/providers/302ai/models/qwen3.6-flash.toml
new file mode 100644
index 00000000000..dd520724603
--- /dev/null
+++ b/providers/302ai/models/qwen3.6-flash.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.6-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.188
+output = 1.133
diff --git a/providers/302ai/models/qwen3.6-plus.toml b/providers/302ai/models/qwen3.6-plus.toml
new file mode 100644
index 00000000000..04cfea256f5
--- /dev/null
+++ b/providers/302ai/models/qwen3.6-plus.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.6-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.3
+output = 1.8
diff --git a/providers/302ai/models/qwen3.7-max-2026-06-08.toml b/providers/302ai/models/qwen3.7-max-2026-06-08.toml
new file mode 100644
index 00000000000..2c84633b1c2
--- /dev/null
+++ b/providers/302ai/models/qwen3.7-max-2026-06-08.toml
@@ -0,0 +1,9 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.7-max"
+name = "qwen3.7-max-2026-06-08"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 1.8
+output = 5.3
diff --git a/providers/302ai/models/qwen3.7-max.toml b/providers/302ai/models/qwen3.7-max.toml
new file mode 100644
index 00000000000..799e725392d
--- /dev/null
+++ b/providers/302ai/models/qwen3.7-max.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.7-max"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 1.8
+output = 5.3
diff --git a/providers/302ai/models/qwen3.7-plus.toml b/providers/302ai/models/qwen3.7-plus.toml
new file mode 100644
index 00000000000..a2e6f1b47b2
--- /dev/null
+++ b/providers/302ai/models/qwen3.7-plus.toml
@@ -0,0 +1,8 @@
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.7-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.285
+output = 1.15
diff --git a/providers/302ai/models/qwen3.8-flash.toml b/providers/302ai/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..490e46b5d0d
--- /dev/null
+++ b/providers/302ai/models/qwen3.8-flash.toml
@@ -0,0 +1,16 @@
+# Toggle: enable_thinking true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens" },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.18
+output = 0.564
diff --git a/providers/302ai/models/qwen3.8-max.toml b/providers/302ai/models/qwen3.8-max.toml
new file mode 100644
index 00000000000..bbc2ee40100
--- /dev/null
+++ b/providers/302ai/models/qwen3.8-max.toml
@@ -0,0 +1,16 @@
+# Toggle: enable_thinking true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# Budget: thinking_budget (integer reasoning tokens)
+base_model = "alibaba/qwen3.8-max"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens" },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.16
+output = 6.36
diff --git a/providers/302ai/provider.toml b/providers/302ai/provider.toml
index aa9acd503e1..9dc6df7b22f 100644
--- a/providers/302ai/provider.toml
+++ b/providers/302ai/provider.toml
@@ -1,11 +1,16 @@
 name = "302.AI"
 env = ["302AI_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
-# Reasoning HTTP format (accessed 2026-06-25):
-# Audited POST https://api.302.ai/v1/chat/completions. The provider's API guide
-# documents model/messages only; no reasoning toggle, effort, or numeric budget
-# request field is documented. Do not infer passthrough from upstream APIs.
-# Sources:
-# https://doc.302.ai/
+# Relay provider: 302.AI proxies multi-model lab endpoints (OpenAI, Anthropic, Google, GLM, Qwen, xAI, Moonshot, Xiaomi, MiniMax, etc.).
+# Reasoning controls verified live against POST https://api.302.ai/v1/chat/completions (2026-09-08):
+# - OpenAI / GPT: `reasoning_effort` (parsed upstream; invalid values rejected with 400)
+# - Qwen / Alibaba: `enable_thinking` true|false, `reasoning_effort`, `thinking_budget` (probed; `reasoning_content` returned)
+# - Moonshot / Kimi: `thinking.type = enabled|disabled` (probed; `reasoning_content` returned)
+# - GLM / Zhipu: `thinking.type = enabled|disabled` (same field as verified Kimi pattern)
+# - Anthropic / Claude: `output_config.effort` and optional `thinking.type` (lab-mapped, not probed)
+# - Google / Gemini, xAI / Grok: `reasoning_effort` (lab-mapped)
+# - MiniMax: `thinking.type = enabled|disabled` (M3, lab-mapped)
+# - Xiaomi / MiMo: entries removed 2026-09-08; IDs rejected even plain chat completions (-10003)
+# - Dedicated thinking IDs (e.g. *-thinking, MiniMax M2.x): []
 doc = "https://doc.302.ai"
 api = "https://api.302.ai/v1"

From 67bcb2c1465bb3b54e71e57f08c3c8eec1d375ba Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Tue, 8 Sep 2026 23:54:16 +0700
Subject: [PATCH 0865/2012] feat(orcarouter): add glm-5.3-flash-free, remove
 qwen3.8-27b-free (#6531)

* feat(orcarouter): add z-ai/glm-5.3-flash-free ( free tier)

* chore(orcarouter): remove qwen/qwen3.8-27b-free (no longer in pricing)
---
 .../orcarouter/models/qwen/qwen3.8-27b-free.toml  | 10 ----------
 .../models/z-ai/glm-5.3-flash-free.toml           | 15 +++++++++++++++
 2 files changed, 15 insertions(+), 10 deletions(-)
 delete mode 100644 providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
 create mode 100644 providers/orcarouter/models/z-ai/glm-5.3-flash-free.toml

diff --git a/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml b/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
deleted file mode 100644
index 72bf87a72bb..00000000000
--- a/providers/orcarouter/models/qwen/qwen3.8-27b-free.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "alibaba/qwen3.8-27b"
-name = "Qwen3.8 27B (free)"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 65_536
diff --git a/providers/orcarouter/models/z-ai/glm-5.3-flash-free.toml b/providers/orcarouter/models/z-ai/glm-5.3-flash-free.toml
new file mode 100644
index 00000000000..ef4e405112a
--- /dev/null
+++ b/providers/orcarouter/models/z-ai/glm-5.3-flash-free.toml
@@ -0,0 +1,15 @@
+# Source: https://api.orcarouter.ai/api/pricing (z-ai/glm-5.3-flash-free, is_free_tier=true, free_base_model=z-ai/glm-5.3-flash, 1M/128k)
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3-Flash (free)"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 1_000_000
+output = 128_000

From 54147a3a6fa68d4c587d754885ce021ba7e157ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 17:25:09 +0000
Subject: [PATCH 0866/2012] chore(sync): update NanoGPT model catalog (#6567)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml   | 4 ----
 providers/nano-gpt/models/gemini-2.5-flash-lite.toml          | 2 +-
 providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml    | 2 +-
 providers/nano-gpt/models/gemini-2.5-flash.toml               | 2 +-
 providers/nano-gpt/models/gemini-2.5-pro.toml                 | 2 +-
 .../models/google/gemini-3-flash-preview-thinking.toml        | 2 +-
 providers/nano-gpt/models/google/gemini-3-flash-preview.toml  | 2 +-
 .../models/google/gemini-3.1-pro-preview-customtools.toml     | 2 +-
 .../nano-gpt/models/google/gemini-3.1-pro-preview-high.toml   | 2 +-
 providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml  | 2 +-
 .../nano-gpt/models/google/gemini-3.5-flash-thinking.toml     | 2 +-
 providers/nano-gpt/models/google/gemini-3.5-flash.toml        | 2 +-
 providers/nano-gpt/models/google/gemini-3.7-flash.toml        | 2 +-
 providers/nano-gpt/models/google/gemini-3.8-flash.toml        | 2 +-
 providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml      | 3 +++
 .../nano-gpt/models/google/gemma-4-26b-a4b-it:thinking.toml   | 3 +++
 providers/nano-gpt/models/google/gemma-4-31b-it.toml          | 3 +++
 providers/nano-gpt/models/google/gemma-4-31b-it:thinking.toml | 3 +++
 providers/nano-gpt/models/meta/muse-spark-1.1.toml            | 2 +-
 providers/nano-gpt/models/qwen/qwen3.5-9b.toml                | 3 ---
 providers/nano-gpt/models/qwen3.8-27b.toml                    | 3 ---
 providers/nano-gpt/models/qwen3.8-27b:thinking.toml           | 3 ---
 providers/nano-gpt/models/thinkingmachines/Inkling-Small.toml | 3 ---
 .../models/thinkingmachines/Inkling-Small:thinking.toml       | 3 ---
 providers/nano-gpt/models/xiaomi/mimo-v2.5.toml               | 3 ---
 providers/nano-gpt/models/xiaomi/mimo-v2.5:thinking.toml      | 3 ---
 providers/nano-gpt/models/z-ai/glm-4.6v.toml                  | 3 ---
 providers/nano-gpt/models/z-ai/glm-5v-turbo.toml              | 2 +-
 providers/nano-gpt/models/z-ai/glm-5v-turbo:thinking.toml     | 2 +-
 29 files changed, 28 insertions(+), 44 deletions(-)

diff --git a/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml b/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
index 3b07182e39b..60e5d8ba554 100644
--- a/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
+++ b/providers/nano-gpt/models/bytedance-seed/seed-2.0-lite.toml
@@ -1,7 +1,6 @@
 # Not included in subscription
 base_model = "bytedance-seed/seed-2.0-lite"
 name = "ByteDance Seed 2.0 Lite"
-attachment = false
 reasoning = false
 tool_call = false
 
@@ -14,6 +13,3 @@ cache_read = 0.125
 context = 262_144
 input = 262_144
 output = 131_072
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-lite.toml b/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
index 8537dcb2c30..64903711c5e 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-lite.toml
@@ -15,4 +15,4 @@ cache_read = 0.01
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml b/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
index 772e6034dae..5c7e8b0ac77 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash-nothinking.toml
@@ -20,5 +20,5 @@ input = 1_048_576
 output = 65_536
 
 [modalities]
-input = ["text", "image", "audio", "pdf"]
+input = ["text", "image", "video", "audio", "pdf"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/gemini-2.5-flash.toml b/providers/nano-gpt/models/gemini-2.5-flash.toml
index 8cf079ce92f..b5ed5c50660 100644
--- a/providers/nano-gpt/models/gemini-2.5-flash.toml
+++ b/providers/nano-gpt/models/gemini-2.5-flash.toml
@@ -14,4 +14,4 @@ cache_read = 0.03
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/gemini-2.5-pro.toml b/providers/nano-gpt/models/gemini-2.5-pro.toml
index aac0d659349..e177a90eb72 100644
--- a/providers/nano-gpt/models/gemini-2.5-pro.toml
+++ b/providers/nano-gpt/models/gemini-2.5-pro.toml
@@ -15,4 +15,4 @@ cache_write = 0.375
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml b/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
index 8e849ee5ac4..0d08ad0d947 100644
--- a/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
+++ b/providers/nano-gpt/models/google/gemini-3-flash-preview-thinking.toml
@@ -16,4 +16,4 @@ cache_read = 0.05
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3-flash-preview.toml b/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
index dd42a4b85e1..9e6a8686ae5 100644
--- a/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
+++ b/providers/nano-gpt/models/google/gemini-3-flash-preview.toml
@@ -14,4 +14,4 @@ cache_read = 0.05
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
index ef98ef336f8..20b170e25a0 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-customtools.toml
@@ -16,4 +16,4 @@ cache_write = 0.375
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
index 253bc7b4555..d1cfd298546 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview-high.toml
@@ -23,5 +23,5 @@ input = 1_048_576
 output = 65_536
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
 output = ["text"]
diff --git a/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml b/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
index 264445e5b25..dc51f72a485 100644
--- a/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
+++ b/providers/nano-gpt/models/google/gemini-3.1-pro-preview.toml
@@ -16,4 +16,4 @@ cache_write = 0.375
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.5-flash-thinking.toml b/providers/nano-gpt/models/google/gemini-3.5-flash-thinking.toml
index 9f87ac4e534..07bb78bc52f 100644
--- a/providers/nano-gpt/models/google/gemini-3.5-flash-thinking.toml
+++ b/providers/nano-gpt/models/google/gemini-3.5-flash-thinking.toml
@@ -18,4 +18,4 @@ cache_write = 0.083333
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.5-flash.toml b/providers/nano-gpt/models/google/gemini-3.5-flash.toml
index 67649cb2131..893ac82319e 100644
--- a/providers/nano-gpt/models/google/gemini-3.5-flash.toml
+++ b/providers/nano-gpt/models/google/gemini-3.5-flash.toml
@@ -15,4 +15,4 @@ cache_write = 0.083333
 input = 1_048_576
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/google/gemini-3.7-flash.toml b/providers/nano-gpt/models/google/gemini-3.7-flash.toml
index 4ccc6179c15..007e8f6fd41 100644
--- a/providers/nano-gpt/models/google/gemini-3.7-flash.toml
+++ b/providers/nano-gpt/models/google/gemini-3.7-flash.toml
@@ -8,7 +8,7 @@ values = ["minimal", "low", "medium", "high"]
 input = 0.75
 output = 3.75
 cache_read = 0.075
-cache_write = 0.075
+cache_write = 0.041667
 
 [limit]
 input = 1_048_576
diff --git a/providers/nano-gpt/models/google/gemini-3.8-flash.toml b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
index 2ff854266fd..5123c75c06a 100644
--- a/providers/nano-gpt/models/google/gemini-3.8-flash.toml
+++ b/providers/nano-gpt/models/google/gemini-3.8-flash.toml
@@ -8,7 +8,7 @@ values = ["minimal", "low", "medium", "high"]
 input = 0.75
 output = 3.75
 cache_read = 0.075
-cache_write = 0.075
+cache_write = 0.041667
 
 [limit]
 input = 1_048_576
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
index 0603997c0b8..3e1fbd8ca47 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it.toml
@@ -13,3 +13,6 @@ cache_read = 0.06
 [limit]
 input = 262_144
 output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it:thinking.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it:thinking.toml
index 2c907734caa..08d90442d4f 100644
--- a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it:thinking.toml
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it:thinking.toml
@@ -11,3 +11,6 @@ cache_read = 0.065
 [limit]
 input = 262_144
 output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it.toml b/providers/nano-gpt/models/google/gemma-4-31b-it.toml
index e4866ce9cc0..ccc02c9a753 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it.toml
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it.toml
@@ -13,3 +13,6 @@ cache_read = 0.05
 [limit]
 input = 262_144
 output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/google/gemma-4-31b-it:thinking.toml b/providers/nano-gpt/models/google/gemma-4-31b-it:thinking.toml
index d59c3a42cf8..f7f76bab951 100644
--- a/providers/nano-gpt/models/google/gemma-4-31b-it:thinking.toml
+++ b/providers/nano-gpt/models/google/gemma-4-31b-it:thinking.toml
@@ -11,3 +11,6 @@ cache_read = 0.05
 [limit]
 input = 262_144
 output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/meta/muse-spark-1.1.toml b/providers/nano-gpt/models/meta/muse-spark-1.1.toml
index b5a1847c173..492ebfbd337 100644
--- a/providers/nano-gpt/models/meta/muse-spark-1.1.toml
+++ b/providers/nano-gpt/models/meta/muse-spark-1.1.toml
@@ -17,4 +17,4 @@ input = 1_000_000
 output = 65_536
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-9b.toml b/providers/nano-gpt/models/qwen/qwen3.5-9b.toml
index 3e8724687eb..0348c77e523 100644
--- a/providers/nano-gpt/models/qwen/qwen3.5-9b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-9b.toml
@@ -18,6 +18,3 @@ cache_read = 0.025
 [limit]
 context = 256_000
 input = 256_000
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/qwen3.8-27b.toml b/providers/nano-gpt/models/qwen3.8-27b.toml
index dc390a04804..7111eaf2f47 100644
--- a/providers/nano-gpt/models/qwen3.8-27b.toml
+++ b/providers/nano-gpt/models/qwen3.8-27b.toml
@@ -8,6 +8,3 @@ cache_read = 0.04
 
 [limit]
 input = 262_144
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/qwen3.8-27b:thinking.toml b/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
index fade9e57385..6ff4fbb2f69 100644
--- a/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
+++ b/providers/nano-gpt/models/qwen3.8-27b:thinking.toml
@@ -9,6 +9,3 @@ cache_read = 0.04
 
 [limit]
 input = 262_144
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/thinkingmachines/Inkling-Small.toml b/providers/nano-gpt/models/thinkingmachines/Inkling-Small.toml
index ed62fad0fb8..d769c5ef3e9 100644
--- a/providers/nano-gpt/models/thinkingmachines/Inkling-Small.toml
+++ b/providers/nano-gpt/models/thinkingmachines/Inkling-Small.toml
@@ -11,6 +11,3 @@ cache_read = 0.1
 context = 524_288
 input = 524_288
 output = 32_768
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/thinkingmachines/Inkling-Small:thinking.toml b/providers/nano-gpt/models/thinkingmachines/Inkling-Small:thinking.toml
index 6374bce6cad..10a8422f551 100644
--- a/providers/nano-gpt/models/thinkingmachines/Inkling-Small:thinking.toml
+++ b/providers/nano-gpt/models/thinkingmachines/Inkling-Small:thinking.toml
@@ -15,6 +15,3 @@ cache_read = 0.1
 context = 524_288
 input = 524_288
 output = 32_768
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.5.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.5.toml
index 6cfd5b6f080..9e12521fafe 100644
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.5.toml
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.5.toml
@@ -15,6 +15,3 @@ cache_write = 0
 
 [limit]
 input = 1_048_576
-
-[modalities]
-input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.5:thinking.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.5:thinking.toml
index c95a9878190..5649f067278 100644
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.5:thinking.toml
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.5:thinking.toml
@@ -14,6 +14,3 @@ cache_write = 0
 
 [limit]
 input = 1_048_576
-
-[modalities]
-input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/z-ai/glm-4.6v.toml b/providers/nano-gpt/models/z-ai/glm-4.6v.toml
index 24d900356a8..bf99f726025 100644
--- a/providers/nano-gpt/models/z-ai/glm-4.6v.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.6v.toml
@@ -12,6 +12,3 @@ cache_read = 0.15
 [limit]
 input = 128_000
 output = 24_000
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/nano-gpt/models/z-ai/glm-5v-turbo.toml b/providers/nano-gpt/models/z-ai/glm-5v-turbo.toml
index 18966c94ce2..02f43815cc6 100644
--- a/providers/nano-gpt/models/z-ai/glm-5v-turbo.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5v-turbo.toml
@@ -14,4 +14,4 @@ context = 202_800
 input = 202_800
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
diff --git a/providers/nano-gpt/models/z-ai/glm-5v-turbo:thinking.toml b/providers/nano-gpt/models/z-ai/glm-5v-turbo:thinking.toml
index a7d1b506c17..e7a34b04ca6 100644
--- a/providers/nano-gpt/models/z-ai/glm-5v-turbo:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5v-turbo:thinking.toml
@@ -14,4 +14,4 @@ context = 202_800
 input = 202_800
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]

From 640a3a21ff45c96abcc9056430de12369c07166b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 17:25:25 +0000
Subject: [PATCH 0867/2012] chore(sync): update OpenRouter model catalog
 (#6569)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..e57f7b0b838 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.953172
+output = 1.906344
+cache_read = 0.079431
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..8119fb87874 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 84eed2d7a418ada21595f7c0f4824c0924099daf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 17:25:30 +0000
Subject: [PATCH 0868/2012] chore(sync): update Kilo model catalog (#6568)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..7b848afc9df 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07125
+output = 0.2375
+cache_read = 0.01425
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 697777a602f714dd3df3fcd9bf7d82cfb5e5ba33 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 18:31:59 +0000
Subject: [PATCH 0869/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6570)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/Qwen3.8-27B.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/Qwen3.8-27B.toml b/providers/llmgateway/models/Qwen3.8-27B.toml
index 8e2a4937fda..b3afd48e7ed 100644
--- a/providers/llmgateway/models/Qwen3.8-27B.toml
+++ b/providers/llmgateway/models/Qwen3.8-27B.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.41
-output = 2.5
-cache_read = 0.08
+input = 0.2
+output = 2
+cache_read = 0.05
 
 [limit]
 context = 32_768

From 08f0d2ff8358a498ae2d62e2229dd2e3a5ef1804 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 18:32:09 +0000
Subject: [PATCH 0870/2012] chore(sync): update LLM Gateway model catalog
 (#6571)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/Qwen3.8-27B.toml               | 6 +++---
 .../models/consensusprotocol/gpt-oss-20b.toml               | 1 +
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
index 7d725a4921e..bbe9b9c35d1 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.41
-output = 2.5
-cache_read = 0.08
+input = 0.2
+output = 2
+cache_read = 0.05
 
 [limit]
 context = 32_768
diff --git a/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml b/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml
index 09259fddd1f..e855e9ed4ff 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/gpt-oss-20b.toml
@@ -1,5 +1,6 @@
 base_model = "openai/gpt-oss-20b"
 name = "GPT OSS 20B (Consensus Protocol)"
+attachment = true
 structured_output = false
 
 [[reasoning_options]]

From a4ff5584564ae1b84d50ee37ccc59318094fdb1f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 18:32:11 +0000
Subject: [PATCH 0871/2012] chore(sync): update OpenRouter model catalog
 (#6573)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++---
 .../models/inception/mercury-2.5-preview.toml |  6 ++---
 .../models/nex-agi/nex-n2.5-mini:free.toml    | 27 +++++++++++++++++++
 .../models/nex-agi/nex-n2.5-pro:free.toml     | 27 +++++++++++++++++++
 .../models/~z-ai/glm-flash-latest.toml        |  8 +++---
 5 files changed, 64 insertions(+), 10 deletions(-)
 create mode 100644 providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
 create mode 100644 providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e57f7b0b838..fa979a51830 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.953172
-output = 1.906344
-cache_read = 0.079431
+input = 0.949692
+output = 1.899384
+cache_read = 0.079141
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inception/mercury-2.5-preview.toml b/providers/openrouter/models/inception/mercury-2.5-preview.toml
index df3e6f4de7c..41b35c78f10 100644
--- a/providers/openrouter/models/inception/mercury-2.5-preview.toml
+++ b/providers/openrouter/models/inception/mercury-2.5-preview.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.04
-output = 0.15
-cache_read = 0.004
+input = 0.2
+output = 0.75
+cache_read = 0.02
 
 [limit]
 context = 260_000
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
new file mode 100644
index 00000000000..e36bfd9b7fa
--- /dev/null
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
@@ -0,0 +1,27 @@
+name = "Nex-N2.5-Mini (free)"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml b/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml
new file mode 100644
index 00000000000..d658bb181c3
--- /dev/null
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml
@@ -0,0 +1,27 @@
+name = "Nex-N2.5-Pro (free)"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8119fb87874..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From c5ceec9d1c49f6202eb1979d7675863eba3a982a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 18:32:36 +0000
Subject: [PATCH 0872/2012] chore(sync): update Kilo model catalog (#6572)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nex-agi/nex-n2.5-mini:free.toml    | 27 +++++++++++++++++++
 .../models/nex-agi/nex-n2.5-pro:free.toml     | 27 +++++++++++++++++++
 .../kilo/models/~z-ai/glm-flash-latest.toml   |  8 +++---
 3 files changed, 58 insertions(+), 4 deletions(-)
 create mode 100644 providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
 create mode 100644 providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml

diff --git a/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
new file mode 100644
index 00000000000..9709a5a038d
--- /dev/null
+++ b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
@@ -0,0 +1,27 @@
+name = "Nex AGI: Nex-N2.5-Mini (free)"
+description = "Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes. Its core strength is agentic coding within a visual feedback loop: it can explore codebases, implement multi-file..."
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml b/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml
new file mode 100644
index 00000000000..8fa74143452
--- /dev/null
+++ b/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml
@@ -0,0 +1,27 @@
+name = "Nex AGI: Nex-N2.5-Pro (free)"
+description = "Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes. Its core strength is agentic coding within a visual feedback loop: it can explore codebases, implement multi-file..."
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 7b848afc9df..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07125
-output = 0.2375
-cache_read = 0.01425
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From 93be848d080d90e5ecf440018505461d023bfc2e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 19:22:58 +0000
Subject: [PATCH 0873/2012] chore(sync): update OpenRouter model catalog
 (#6574)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml         |  6 +++---
 .../models/deepseek/deepseek-v4-pro.toml           |  6 +++---
 .../{mercury-2.5-preview.toml => mercury-2.5.toml} | 14 +++++++-------
 .../openrouter/models/minimax/minimax-m1.toml      |  2 +-
 .../models/moonshotai/kimi-k2-thinking.toml        |  3 ++-
 .../models/nex-agi/nex-n2.5-mini:free.toml         |  2 +-
 .../models/nex-agi/nex-n2.5-pro:free.toml          |  2 +-
 7 files changed, 18 insertions(+), 17 deletions(-)
 rename providers/openrouter/models/inception/{mercury-2.5-preview.toml => mercury-2.5.toml} (58%)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index fa979a51830..ae6a4dbd88c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.949692
-output = 1.899384
-cache_read = 0.079141
+input = 0.948126
+output = 1.896252
+cache_read = 0.079011
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inception/mercury-2.5-preview.toml b/providers/openrouter/models/inception/mercury-2.5.toml
similarity index 58%
rename from providers/openrouter/models/inception/mercury-2.5-preview.toml
rename to providers/openrouter/models/inception/mercury-2.5.toml
index 41b35c78f10..04931f78a87 100644
--- a/providers/openrouter/models/inception/mercury-2.5-preview.toml
+++ b/providers/openrouter/models/inception/mercury-2.5.toml
@@ -1,8 +1,8 @@
-name = "Mercury 2.5 Preview"
-description = "Preview model for early access evaluation, prototyping, and compatibility testing"
+name = "Mercury 2.5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
 family = "mercury"
-release_date = "2026-08-31"
-last_updated = "2026-08-31"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
 attachment = false
 reasoning = true
 temperature = true
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.2
-output = 0.75
-cache_read = 0.02
+input = 0.04
+output = 0.15
+cache_read = 0.004
 
 [limit]
 context = 260_000
diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 178789497a8..063613f9d24 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -17,7 +17,7 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.55
+input = 0.4
 output = 2.2
 
 [limit]
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 3ccd5a4f3ef..91ad22ed77f 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,6 +8,7 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
+cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
index e36bfd9b7fa..40080615553 100644
--- a/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = true
 
 [[reasoning_options]]
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml b/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml
index d658bb181c3..5d1969c1a4e 100644
--- a/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-pro:free.toml
@@ -7,7 +7,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = true
 
 [[reasoning_options]]

From 86ee3fc3a363244b798641ccf36015d98c1c0adc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 19:23:00 +0000
Subject: [PATCH 0874/2012] chore(sync): update Kilo model catalog (#6575)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../{mercury-2.5-preview.toml => mercury-2.5.toml}          | 6 +++---
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml      | 2 +-
 providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml       | 2 +-
 providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml        | 2 +-
 4 files changed, 6 insertions(+), 6 deletions(-)
 rename providers/kilo/models/inception/{mercury-2.5-preview.toml => mercury-2.5.toml} (86%)

diff --git a/providers/kilo/models/inception/mercury-2.5-preview.toml b/providers/kilo/models/inception/mercury-2.5.toml
similarity index 86%
rename from providers/kilo/models/inception/mercury-2.5-preview.toml
rename to providers/kilo/models/inception/mercury-2.5.toml
index 3ac644d23b9..03b5d5c14c7 100644
--- a/providers/kilo/models/inception/mercury-2.5-preview.toml
+++ b/providers/kilo/models/inception/mercury-2.5.toml
@@ -1,8 +1,8 @@
-name = "Inception: Mercury 2.5 Preview"
+name = "Inception: Mercury 2.5"
 description = "Mercury 2.5 is the fastest reasoning LLM, and the latest diffusion LLM (dLLM) from Inception. Instead of generating tokens sequentially, Mercury 2.5 produces and refines multiple tokens in parallel, achieving..."
 family = "mercury"
-release_date = "2026-08-31"
-last_updated = "2026-08-31"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
 attachment = false
 reasoning = true
 temperature = true
diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 3e83e7378b1..75e607a6e02 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352
diff --git a/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
index 9709a5a038d..59d07170b17 100644
--- a/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
+++ b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = false
 
 [[reasoning_options]]
diff --git a/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml b/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml
index 8fa74143452..6788a3fe86a 100644
--- a/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml
+++ b/providers/kilo/models/nex-agi/nex-n2.5-pro:free.toml
@@ -7,7 +7,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-structured_output = false
+structured_output = true
 open_weights = false
 
 [[reasoning_options]]

From b42db77bf2d4194c26f59fd397c4d8ab7389801c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 19:23:13 +0000
Subject: [PATCH 0875/2012] chore(sync): update NanoGPT model catalog (#6576)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml      | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 50bc9fa38b4..2d0019c9056 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high"]
 
 [cost]
-input = 0.25
-output = 1
-cache_read = 0.125
+input = 0.35
+output = 1.4
+cache_read = 0.175
 
 [limit]
 context = 1_048_576

From 0c7b29e724ff5defa8ad84e18ec73cf6f88b9a62 Mon Sep 17 00:00:00 2001
From: Hugues-Antoine <ha@rabany.eu>
Date: Tue, 8 Sep 2026 22:02:51 +0200
Subject: [PATCH 0876/2012] chore(sync): update cost ovhcloud model qwen*27b
 (#6541)

Co-authored-by: Hugues-Antoine Rabany <ha@aclccp.com>
---
 providers/ovhcloud/models/qwen3.6-27b.toml | 4 ++--
 providers/ovhcloud/models/qwen3.8-27b.toml | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/ovhcloud/models/qwen3.6-27b.toml b/providers/ovhcloud/models/qwen3.6-27b.toml
index 24cf367061c..a6b95592e67 100644
--- a/providers/ovhcloud/models/qwen3.6-27b.toml
+++ b/providers/ovhcloud/models/qwen3.6-27b.toml
@@ -14,8 +14,8 @@ structured_output = true
 open_weights = true
 
 [cost]
-input = 0.47
-output = 3.19
+input = 0.46
+output = 3.13
 
 [limit]
 context = 262_144
diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 45ae7920a40..537063f3d02 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -22,8 +22,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0
-output = 0
+input = 0.46
+output = 3.13
 
 [limit]
 context = 262_144

From e1e63f6ea833b7c86e5d0fc416ab32e81de48122 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:03:22 -0500
Subject: [PATCH 0877/2012] chore(sync): update Requesty model catalog (#6557)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/gpt-6-astra.toml | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/requesty/models/gpt-6-astra.toml

diff --git a/providers/requesty/models/gpt-6-astra.toml b/providers/requesty/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..ebb2823694f
--- /dev/null
+++ b/providers/requesty/models/gpt-6-astra.toml
@@ -0,0 +1,19 @@
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2

From e11c4fc78d8e410f0d46d18546aee880fac8120c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:03:35 -0500
Subject: [PATCH 0878/2012] chore(sync): update DigitalOcean model catalog
 (#6565)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai-gpt-6-astra.toml            | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/digitalocean/models/openai-gpt-6-astra.toml

diff --git a/providers/digitalocean/models/openai-gpt-6-astra.toml b/providers/digitalocean/models/openai-gpt-6-astra.toml
new file mode 100644
index 00000000000..ef287793c30
--- /dev/null
+++ b/providers/digitalocean/models/openai-gpt-6-astra.toml
@@ -0,0 +1,20 @@
+base_model = "openai/gpt-6-astra"
+name = "OpenAI GPT-6 Astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+
+[modalities]
+input = ["text", "image"]

From ef68b9f95380447adf77d4424dfcb0933d23be04 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:09:18 -0500
Subject: [PATCH 0879/2012] chore(sync): update Cortecs model catalog (#6551)

* chore(sync): update Cortecs model catalog

* fix(cortecs): add Gemini 3.8 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
---
 providers/cortecs/models/gemini-3.8-flash.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/cortecs/models/gemini-3.8-flash.toml

diff --git a/providers/cortecs/models/gemini-3.8-flash.toml b/providers/cortecs/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..c3b7bf92671
--- /dev/null
+++ b/providers/cortecs/models/gemini-3.8-flash.toml
@@ -0,0 +1,16 @@
+# Effort: reasoning_effort = low|medium|high
+# Source: https://docs.cortecs.ai/examples/reasoning
+base_model = "google/gemini-3.8-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.825
+output = 4.125
+cache_read = 0.082
+cache_write = 0.084
+
+[limit]
+output = 65_535
+
+[modalities]
+input = ["text", "image", "audio", "pdf"]

From 26131ccd3c6d7cb10cedc75dc6b82002a9e059d4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:09:27 -0500
Subject: [PATCH 0880/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6553)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add reasoning options for new models

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
---
 .../deepseek/deepseek-v4.1-flash-beta.toml    | 25 +++++++++++++++++++
 .../vercel/models/inception/mercury-2.5.toml  | 25 +++++++++++++++++++
 .../xiaomi/mimo-v2.5-pro-ultraspeed.toml      |  9 -------
 3 files changed, 50 insertions(+), 9 deletions(-)
 create mode 100644 providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
 create mode 100644 providers/vercel/models/inception/mercury-2.5.toml
 delete mode 100644 providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml

diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
new file mode 100644
index 00000000000..0a7f78795e6
--- /dev/null
+++ b/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
@@ -0,0 +1,25 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning.effort = high|xhigh (Vercel gateway aliases)
+name = "DeepSeek V4.1 Flash Beta"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+family = "deepseek"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "xhigh"] }]
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/inception/mercury-2.5.toml b/providers/vercel/models/inception/mercury-2.5.toml
new file mode 100644
index 00000000000..24e73a012a7
--- /dev/null
+++ b/providers/vercel/models/inception/mercury-2.5.toml
@@ -0,0 +1,25 @@
+# Effort: reasoning.effort = none|low|medium|high
+# Native levels: https://docs.inceptionlabs.ai/capabilities/instant
+name = "Mercury 2.5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+family = "mercury"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = false
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[cost]
+input = 0.04
+output = 0.15
+cache_read = 0.004
+
+[limit]
+context = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml b/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml
deleted file mode 100644
index 8311903d432..00000000000
--- a/providers/vercel/models/xiaomi/mimo-v2.5-pro-ultraspeed.toml
+++ /dev/null
@@ -1,9 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-base_model = "xiaomi/mimo-v2.5-pro-ultraspeed"
-name = "MiMo V2.5 Pro UltraSpeed"
-reasoning_options = [{ type = "toggle" }]
-
-[cost]
-input = 1.305
-output = 2.61
-cache_read = 0.0108

From 9e2466cdb37542542de8d0ff33331321041ea840 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 20:25:51 +0000
Subject: [PATCH 0881/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#6577)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ovhcloud/models/qwen3.6-27b.toml | 10 ++++++----
 providers/ovhcloud/models/qwen3.8-27b.toml |  4 ++--
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/providers/ovhcloud/models/qwen3.6-27b.toml b/providers/ovhcloud/models/qwen3.6-27b.toml
index a6b95592e67..b0decce5dc7 100644
--- a/providers/ovhcloud/models/qwen3.6-27b.toml
+++ b/providers/ovhcloud/models/qwen3.6-27b.toml
@@ -1,21 +1,23 @@
 # Chat `reasoning_effort` supports none|minimal|low|medium|high; `none` disables it.
 # https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog/qwen-3-6-27b/
-
 name = "Qwen3.6-27B"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 release_date = "2026-06-01"
 last_updated = "2026-06-01"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
 temperature = true
 tool_call = true
 structured_output = true
 open_weights = true
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
+
 [cost]
-input = 0.46
-output = 3.13
+input = 0.47
+output = 3.19
 
 [limit]
 context = 262_144
diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 537063f3d02..45ae7920a40 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -22,8 +22,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.46
-output = 3.13
+input = 0
+output = 0
 
 [limit]
 context = 262_144

From 574239535cd932330873fd0d1856de3e7c3a0c96 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Tue, 8 Sep 2026 15:56:11 -0500
Subject: [PATCH 0882/2012] feat(amazon-bedrock): add GPT-6 Astra models
 (#6578)

---
 .../models/global.openai.gpt-6-astra.toml     | 22 +++++++++++++++
 .../models/openai.gpt-6-astra.toml            | 27 +++++++++++++++++++
 .../models/us.openai.gpt-6-astra.toml         | 22 +++++++++++++++
 3 files changed, 71 insertions(+)
 create mode 100644 providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
 create mode 100644 providers/amazon-bedrock/models/openai.gpt-6-astra.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml

diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
new file mode 100644
index 00000000000..e03babec72a
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
@@ -0,0 +1,22 @@
+# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
+# Global CRIS pricing inferred at 1.0x OpenAI-direct from the GPT-5.6 Sol AWS card precedent.
+# Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+name = "GPT-6 Astra (Global)"
+structured_output = false
+
+[cost]
+input = 10.00
+output = 50.00
+cache_read = 1.00
+cache_write = 12.50
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 20.00
+output = 75.00
+cache_read = 2.00
+cache_write = 25.00
diff --git a/providers/amazon-bedrock/models/openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/openai.gpt-6-astra.toml
new file mode 100644
index 00000000000..b6abd665d27
--- /dev/null
+++ b/providers/amazon-bedrock/models/openai.gpt-6-astra.toml
@@ -0,0 +1,27 @@
+# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
+# Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
+# In-region pricing inferred at 1.1x OpenAI-direct from the GPT-5.6 Sol precedent
+# (AWS card prices in-region and Geo CRIS at 1.1x, Global CRIS at 1.0x); no AWS Astra model card published yet.
+# Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
+# Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+
+[cost]
+input = 11.00
+output = 55.00
+cache_read = 1.10
+cache_write = 13.75
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 22.00
+output = 82.50
+cache_read = 2.20
+cache_write = 27.50
+
+[provider]
+npm = "@ai-sdk/amazon-bedrock/mantle"
+api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+shape = "responses"
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
new file mode 100644
index 00000000000..faa2fdd8679
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
@@ -0,0 +1,22 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
+# Geo CRIS pricing inferred at 1.1x OpenAI-direct (= in-region) from the GPT-5.6 Sol AWS card precedent.
+# Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+name = "GPT-6 Astra (US)"
+structured_output = false
+
+[cost]
+input = 11.00
+output = 55.00
+cache_read = 1.10
+cache_write = 13.75
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 22.00
+output = 82.50
+cache_read = 2.20
+cache_write = 27.50

From e00d4f66324d4bdea6cff6bd5c9ec08fad56d66e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 21:22:24 +0000
Subject: [PATCH 0883/2012] chore(sync): update OpenRouter model catalog
 (#6580)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ae6a4dbd88c..799137f52c2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.948126
-output = 1.896252
-cache_read = 0.079011
+input = 0.946386
+output = 1.892772
+cache_read = 0.078866
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 052fb05abd3..70251750002 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 14
-cache_read = 0.29
+input = 2.55
+output = 12.75
+cache_read = 0.256
 
 [limit]
 context = 1_048_576

From ee25f4c39b0ada930d3ba1e5dfc8e145b605d1cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 21:22:42 +0000
Subject: [PATCH 0884/2012] chore(sync): update Kilo model catalog (#6579)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 6481e4e1993..de31079e2dd 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.5
-output = 14
-cache_read = 0.29
+input = 2.55
+output = 12.75
+cache_read = 0.256
 
 [limit]
 context = 1_048_576

From b3c1ce236bc03b533332f1d76b5fdc29562802cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 22:24:06 +0000
Subject: [PATCH 0885/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6583)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-image-2.5-flare.toml    | 22 +++++++++++++++++++
 .../models/openai/gpt-image-2.5-sunburst.toml | 22 +++++++++++++++++++
 2 files changed, 44 insertions(+)
 create mode 100644 providers/vercel/models/openai/gpt-image-2.5-flare.toml
 create mode 100644 providers/vercel/models/openai/gpt-image-2.5-sunburst.toml

diff --git a/providers/vercel/models/openai/gpt-image-2.5-flare.toml b/providers/vercel/models/openai/gpt-image-2.5-flare.toml
new file mode 100644
index 00000000000..b64fffc4712
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-image-2.5-flare.toml
@@ -0,0 +1,22 @@
+name = "GPT Image 2.5 Flare"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "gpt-image"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 5
+output = 30
+cache_read = 1.25
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["image"]
diff --git a/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml b/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml
new file mode 100644
index 00000000000..948747ae075
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml
@@ -0,0 +1,22 @@
+name = "GPT Image 2.5 Sunburst"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "gpt-image"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 5
+output = 30
+cache_read = 1.25
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["image"]

From ab870c636a556677e827e10a9a19d9aab69bd773 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 22:24:25 +0000
Subject: [PATCH 0886/2012] chore(sync): update OpenRouter model catalog
 (#6582)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 799137f52c2..b7feb0289f9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.946386
-output = 1.892772
-cache_read = 0.078866
+input = 0.942906
+output = 1.885812
+cache_read = 0.078576
 
 [limit]
 context = 1_048_576

From 63d97e2cdd0220c9df18235fa4c23e2012e60fb6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 23:22:46 +0000
Subject: [PATCH 0887/2012] chore(sync): update OpenRouter model catalog
 (#6586)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/minimax/minimax-m1.toml         | 2 +-
 .../models/nvidia/nemotron-3-super-120b-a12b.toml           | 5 +++--
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b7feb0289f9..c1bd24986c4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.942906
-output = 1.885812
-cache_read = 0.078576
+input = 0.9396
+output = 1.8792
+cache_read = 0.0783
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 063613f9d24..178789497a8 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -17,7 +17,7 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.4
+input = 0.55
 output = 2.2
 
 [limit]
diff --git a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
index 0c8ec81b077..9376472a1c3 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -1,5 +1,7 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3-super-120b-a12b"
-structured_output = true
+structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -16,5 +18,4 @@ input = 0.085
 output = 0.4
 
 [limit]
-context = 1_000_000
 output = 16_384

From 14e8513b8f19df6f39476a57243e9139c2fb57ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 23:22:50 +0000
Subject: [PATCH 0888/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6587)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/zai/glm-5.3-promo-50.toml   | 23 -------------------
 providers/vercel/models/zai/glm-5.3.toml      |  6 ++---
 2 files changed, 3 insertions(+), 26 deletions(-)
 delete mode 100644 providers/vercel/models/zai/glm-5.3-promo-50.toml

diff --git a/providers/vercel/models/zai/glm-5.3-promo-50.toml b/providers/vercel/models/zai/glm-5.3-promo-50.toml
deleted file mode 100644
index 3d49e4731b7..00000000000
--- a/providers/vercel/models/zai/glm-5.3-promo-50.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "GLM 5.3 (50% off)"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2026-08-18"
-last_updated = "2026-08-18"
-attachment = false
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/vercel/models/zai/glm-5.3.toml b/providers/vercel/models/zai/glm-5.3.toml
index ca69c65568b..05cdccc7976 100644
--- a/providers/vercel/models/zai/glm-5.3.toml
+++ b/providers/vercel/models/zai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 1.4
+output = 4.4
+cache_read = 0.14
 
 [limit]
 output = 1_000_000

From c7e2692389a38260fded62697f9937ac8d40f2f1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 8 Sep 2026 23:23:03 +0000
Subject: [PATCH 0889/2012] chore(sync): update Kilo model catalog (#6585)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
index 39cb4ddae08..f3a2124746d 100644
--- a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -1,5 +1,5 @@
 base_model = "nvidia/nemotron-3-super-120b-a12b"
-structured_output = true
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"

From 81f3a9546967bc967d5e4d44d2f2b418fd736f8c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 00:59:36 +0000
Subject: [PATCH 0890/2012] chore(sync): update Kilo model catalog (#6588)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml  |  1 +
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml |  2 +-
 .../kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml  |  2 +-
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml      |  1 -
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                | 10 +++++++---
 6 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index f39407dc694..a280435f49f 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,3 +12,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 75e607a6e02..3e83e7378b1 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 8247cd4ca01..8071bb6d3c0 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.78
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index c79bdfce589..39ad2aa079f 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -16,4 +16,3 @@ cache_write = 2.5
 
 [limit]
 context = 1_000_000
-output = 262_144
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fee21b1f2d8..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,6 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+context = 198_000
+output = 16_384

From ff46fe0971dea46848af5416af672bfb50a67da3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 00:59:39 +0000
Subject: [PATCH 0891/2012] chore(sync): update OpenRouter model catalog
 (#6589)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 1 +
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 .../openrouter/models/ibm-granite/granite-4.2-8b.toml    | 6 +++---
 .../openrouter/models/moonshotai/kimi-k2-thinking.toml   | 3 +--
 .../models/qwen/qwen3-next-80b-a3b-instruct.toml         | 5 ++---
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml  | 1 -
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 ++++++---
 9 files changed, 22 insertions(+), 21 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 9e898f11a05..2186c60b6fb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -17,3 +17,4 @@ cache_read = 0.007
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index c1bd24986c4..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.9396
-output = 1.8792
-cache_read = 0.0783
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
index 1b683dd536e..62943cfaadb 100644
--- a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
+++ b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.1
-output = 0.15
-cache_read = 0.05
+input = 0.06
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 131_072
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 91ad22ed77f..3ccd5a4f3ef 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,7 +8,6 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
-cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 31bb83cc8bb..64deac24f44 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -2,10 +2,9 @@ base_model = "alibaba/qwen3-next-80b-a3b-instruct"
 structured_output = true
 
 [cost]
-input = 0.1
+input = 0.09
 output = 1.1
-cache_read = 0.07
 
 [limit]
 context = 262_144
-output = 235_929
+output = 16_384
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index abacdd68c88..74b5d1276f6 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -18,4 +18,3 @@ cache_read = 0.25
 
 [limit]
 context = 1_048_576
-output = 262_144
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 6e636f26b9c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.55
-output = 2.2
-cache_read = 0.11
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384

From 0e550e21eeca2ac4f1f7ca2361f7b9a11198b453 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 01:00:03 +0000
Subject: [PATCH 0892/2012] chore(sync): update Venice model catalog (#6590)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/mercury-2-5.toml | 27 ++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/venice/models/mercury-2-5.toml

diff --git a/providers/venice/models/mercury-2-5.toml b/providers/venice/models/mercury-2-5.toml
new file mode 100644
index 00000000000..a61068676a6
--- /dev/null
+++ b/providers/venice/models/mercury-2-5.toml
@@ -0,0 +1,27 @@
+name = "Mercury 2.5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+family = "mercury"
+release_date = "2026-09-08"
+last_updated = "2026-09-09"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.04999999999999999
+output = 0.18749999999999994
+cache_read = 0.004999999999999999
+
+[limit]
+context = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 469adf3f1223a8c356e133cae33c9e8ee31212ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 01:36:02 +0000
Subject: [PATCH 0893/2012] chore(sync): update NanoGPT model catalog (#6591)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 99c0a7fd3ff95ba045447712136505bab08de1af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 01:36:13 +0000
Subject: [PATCH 0894/2012] chore(sync): update OpenRouter model catalog
 (#6592)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 70251750002..e7ceb0c72c4 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.4
+output = 12
+cache_read = 0.24
 
 [limit]
 context = 1_048_576

From c968d3e5936f7095319895c350e35f96044d8c74 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 02:34:31 +0000
Subject: [PATCH 0895/2012] chore(sync): update Kilo model catalog (#6594)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index de31079e2dd..7b92a6177bd 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.75
-cache_read = 0.256
+input = 2.4
+output = 12
+cache_read = 0.24
 
 [limit]
 context = 1_048_576

From cf1a52df2a9b2d11191bb496a019d0325c57741b Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Tue, 8 Sep 2026 22:21:29 -0500
Subject: [PATCH 0896/2012] feat(amazon-bedrock): add GPT-5.6 US profiles and
 Opus 4.7 AU profile (#6595)

---
 .../models/au.anthropic.claude-opus-4-7.toml  | 12 ++++++++++
 .../models/us.openai.gpt-5.6-luna.toml        | 22 +++++++++++++++++++
 .../models/us.openai.gpt-5.6-sol.toml         | 22 +++++++++++++++++++
 .../models/us.openai.gpt-5.6-terra.toml       | 22 +++++++++++++++++++
 4 files changed, 78 insertions(+)
 create mode 100644 providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml

diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
new file mode 100644
index 00000000000..6f693e67bde
--- /dev/null
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
@@ -0,0 +1,12 @@
+# AU cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
+# AU geo pricing (Price List ap-southeast-2 standard): input/output/cache-read/cache-write USD per 1M tokens.
+# Note: sibling au.* entries are priced at 1.0x; the Price List bills AU-geo calls at the 1.1x standard rate.
+base_model = "anthropic/claude-opus-4-7"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 4.7 (AU)"
+
+[cost]
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
new file mode 100644
index 00000000000..98e3214639b
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
@@ -0,0 +1,22 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
+# Geo CRIS pricing (= in-region) per the GPT-5.6 AWS card pattern; OpenAI-on-Bedrock is absent from the Price List API.
+# Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-luna.toml
+base_model = "openai/gpt-5.6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+name = "GPT-5.6 Luna (US)"
+structured_output = false
+
+[cost]
+input = 0.22
+output = 1.32
+cache_read = 0.022
+cache_write = 0.275
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.44
+output = 1.98
+cache_read = 0.044
+cache_write = 0.55
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
new file mode 100644
index 00000000000..e492f79ec73
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
@@ -0,0 +1,22 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
+# Geo CRIS pricing (= in-region) per the GPT-5.6 Sol AWS card; OpenAI-on-Bedrock is absent from the Price List API.
+# Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-sol.toml
+base_model = "openai/gpt-5.6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+name = "GPT-5.6 Sol (US)"
+structured_output = false
+
+[cost]
+input = 4.40
+output = 22.00
+cache_read = 0.44
+cache_write = 5.50
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 8.80
+output = 33.00
+cache_read = 0.88
+cache_write = 11.00
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
new file mode 100644
index 00000000000..12add6584d1
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
@@ -0,0 +1,22 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
+# Geo CRIS pricing (= in-region) per the GPT-5.6 AWS card pattern; OpenAI-on-Bedrock is absent from the Price List API.
+# Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-terra.toml
+base_model = "openai/gpt-5.6-terra"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
+name = "GPT-5.6 Terra (US)"
+structured_output = false
+
+[cost]
+input = 2.20
+output = 13.20
+cache_read = 0.22
+cache_write = 2.75
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 4.40
+output = 19.80
+cache_read = 0.44
+cache_write = 5.50

From 71ceff635dcadf5b3441cd66c2a1bf3b2733a19f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 03:30:37 +0000
Subject: [PATCH 0897/2012] chore(sync): update OpenRouter model catalog
 (#6600)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 660f0f992c4..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,7 +12,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.57948
 output = 1.73844
-cache_read = 0.057948
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From ee4cde1d4975971ae9ce682991d59dccbfc5b8fa Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Tue, 8 Sep 2026 22:55:09 -0500
Subject: [PATCH 0898/2012] feat(amazon-bedrock): add Nova cross-region
 inference profiles (#6597)

* feat(amazon-bedrock): add Nova cross-region inference profiles

* fix(amazon-bedrock): correct Nova 2 Lite lab limits, dates, modalities
---
 models/amazon/nova-2-lite.toml                | 18 ++++++++++++++++++
 models/amazon/nova-lite.toml                  | 19 +++++++++++++++++++
 models/amazon/nova-micro.toml                 | 19 +++++++++++++++++++
 models/amazon/nova-pro.toml                   | 19 +++++++++++++++++++
 .../models/apac.amazon.nova-lite-v1:0.toml    | 10 ++++++++++
 .../models/apac.amazon.nova-micro-v1:0.toml   | 10 ++++++++++
 .../models/apac.amazon.nova-pro-v1:0.toml     | 10 ++++++++++
 .../models/ca.amazon.nova-lite-v1:0.toml      |  9 +++++++++
 .../models/eu.amazon.nova-2-lite-v1:0.toml    | 11 +++++++++++
 .../models/eu.amazon.nova-lite-v1:0.toml      |  9 +++++++++
 .../models/eu.amazon.nova-micro-v1:0.toml     |  9 +++++++++
 .../models/eu.amazon.nova-pro-v1:0.toml       |  9 +++++++++
 .../global.amazon.nova-2-lite-v1:0.toml       | 11 +++++++++++
 .../models/jp.amazon.nova-2-lite-v1:0.toml    | 11 +++++++++++
 .../models/us.amazon.nova-2-lite-v1:0.toml    | 11 +++++++++++
 .../models/us.amazon.nova-lite-v1:0.toml      |  9 +++++++++
 .../models/us.amazon.nova-micro-v1:0.toml     |  9 +++++++++
 .../models/us.amazon.nova-pro-v1:0.toml       |  9 +++++++++
 18 files changed, 212 insertions(+)
 create mode 100644 models/amazon/nova-2-lite.toml
 create mode 100644 models/amazon/nova-lite.toml
 create mode 100644 models/amazon/nova-micro.toml
 create mode 100644 models/amazon/nova-pro.toml
 create mode 100644 providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml

diff --git a/models/amazon/nova-2-lite.toml b/models/amazon/nova-2-lite.toml
new file mode 100644
index 00000000000..f818123c9c7
--- /dev/null
+++ b/models/amazon/nova-2-lite.toml
@@ -0,0 +1,18 @@
+name = "Nova 2 Lite"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "nova"
+release_date = "2025-12-01"
+last_updated = "2025-12-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]
diff --git a/models/amazon/nova-lite.toml b/models/amazon/nova-lite.toml
new file mode 100644
index 00000000000..5b14adbd695
--- /dev/null
+++ b/models/amazon/nova-lite.toml
@@ -0,0 +1,19 @@
+name = "Nova Lite"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "nova-lite"
+release_date = "2024-12-03"
+last_updated = "2024-12-03"
+attachment = true
+reasoning = false
+temperature = true
+knowledge = "2024-10"
+tool_call = true
+open_weights = false
+
+[limit]
+context = 300_000
+output = 8_192
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/models/amazon/nova-micro.toml b/models/amazon/nova-micro.toml
new file mode 100644
index 00000000000..e786a6f05c4
--- /dev/null
+++ b/models/amazon/nova-micro.toml
@@ -0,0 +1,19 @@
+name = "Nova Micro"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "nova-micro"
+release_date = "2024-12-03"
+last_updated = "2024-12-03"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2024-10"
+tool_call = true
+open_weights = false
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/models/amazon/nova-pro.toml b/models/amazon/nova-pro.toml
new file mode 100644
index 00000000000..40119549868
--- /dev/null
+++ b/models/amazon/nova-pro.toml
@@ -0,0 +1,19 @@
+name = "Nova Pro"
+description = "Flagship model for demanding analysis, coding, and production agent workflows"
+family = "nova-pro"
+release_date = "2024-12-03"
+last_updated = "2024-12-03"
+attachment = true
+reasoning = false
+temperature = true
+knowledge = "2024-10"
+tool_call = true
+open_weights = false
+
+[limit]
+context = 300_000
+output = 8_192
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
new file mode 100644
index 00000000000..e513171b623
--- /dev/null
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
@@ -0,0 +1,10 @@
+# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
+# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
+# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-lite"
+name = "Nova Lite (APAC)"
+
+[cost]
+input = 0.063
+output = 0.252
+cache_read = 0.01575
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
new file mode 100644
index 00000000000..efab7cd712f
--- /dev/null
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
@@ -0,0 +1,10 @@
+# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
+# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
+# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-micro"
+name = "Nova Micro (APAC)"
+
+[cost]
+input = 0.037
+output = 0.148
+cache_read = 0.00925
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
new file mode 100644
index 00000000000..da293e8ec9d
--- /dev/null
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
@@ -0,0 +1,10 @@
+# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
+# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
+# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-pro"
+name = "Nova Pro (APAC)"
+
+[cost]
+input = 0.840
+output = 3.360
+cache_read = 0.210
diff --git a/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
new file mode 100644
index 00000000000..316aa5c3b62
--- /dev/null
+++ b/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
@@ -0,0 +1,9 @@
+# CA cross-Region inference profile, verified via ListInferenceProfiles from ca-central-1.
+# CA geo pricing (Price List ca-central-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-lite"
+name = "Nova Lite (CA)"
+
+[cost]
+input = 0.064
+output = 0.256
+cache_read = 0.016
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
new file mode 100644
index 00000000000..13a858c8b57
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
@@ -0,0 +1,11 @@
+# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
+# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+base_model = "amazon/nova-2-lite"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+name = "Nova 2 Lite (EU)"
+
+[cost]
+input = 0.374
+output = 3.157
+cache_read = 0.0935
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
new file mode 100644
index 00000000000..d1f01cfcb38
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
@@ -0,0 +1,9 @@
+# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
+# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-lite"
+name = "Nova Lite (EU)"
+
+[cost]
+input = 0.069
+output = 0.276
+cache_read = 0.01725
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
new file mode 100644
index 00000000000..08d27c708f0
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
@@ -0,0 +1,9 @@
+# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
+# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-micro"
+name = "Nova Micro (EU)"
+
+[cost]
+input = 0.040
+output = 0.160
+cache_read = 0.010
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
new file mode 100644
index 00000000000..68e21d0d284
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
@@ -0,0 +1,9 @@
+# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
+# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-pro"
+name = "Nova Pro (EU)"
+
+[cost]
+input = 0.920
+output = 3.680
+cache_read = 0.230
diff --git a/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
new file mode 100644
index 00000000000..aa803d4bf43
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
@@ -0,0 +1,11 @@
+# Global cross-Region inference profile, verified via ListInferenceProfiles (visible from all regions).
+# Global pricing (Price List us-east-1 global): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+base_model = "amazon/nova-2-lite"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+name = "Nova 2 Lite (Global)"
+
+[cost]
+input = 0.300
+output = 2.500
+cache_read = 0.075
diff --git a/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
new file mode 100644
index 00000000000..f652010870d
--- /dev/null
+++ b/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
@@ -0,0 +1,11 @@
+# JP cross-Region inference profile, verified via ListInferenceProfiles from ap-northeast-1.
+# JP geo pricing (Price List ap-northeast-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+base_model = "amazon/nova-2-lite"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+name = "Nova 2 Lite (JP)"
+
+[cost]
+input = 0.396
+output = 3.311
+cache_read = 0.099
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
new file mode 100644
index 00000000000..18c7833c41b
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
@@ -0,0 +1,11 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1 and ca-central-1.
+# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+base_model = "amazon/nova-2-lite"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+name = "Nova 2 Lite (US)"
+
+[cost]
+input = 0.33
+output = 2.75
+cache_read = 0.0825
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
new file mode 100644
index 00000000000..38b93c68908
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-lite"
+name = "Nova Lite (US)"
+
+[cost]
+input = 0.06
+output = 0.24
+cache_read = 0.015
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
new file mode 100644
index 00000000000..781b77c917a
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-micro"
+name = "Nova Micro (US)"
+
+[cost]
+input = 0.035
+output = 0.14
+cache_read = 0.00875
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml
new file mode 100644
index 00000000000..259d180c70d
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+base_model = "amazon/nova-pro"
+name = "Nova Pro (US)"
+
+[cost]
+input = 0.80
+output = 3.20
+cache_read = 0.20

From 6d0cc3506e5e70085fe507f7ff31339a7ecf4028 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Tue, 8 Sep 2026 22:55:23 -0500
Subject: [PATCH 0899/2012] feat(amazon-bedrock): add US Llama/Pixtral/Writer
 and EU Pixtral profiles (#6598)

* feat(amazon-bedrock): add US Llama/Pixtral/Writer and EU Pixtral profiles

* fix(amazon-bedrock): correct Pixtral 25.02 lab family and attachment
---
 models/meta/llama-3.1-70b-instruct.toml       | 23 +++++++++++++++++++
 models/mistral/pixtral-large-2502.toml        | 18 +++++++++++++++
 models/writer/palmyra-x4.toml                 | 18 +++++++++++++++
 models/writer/palmyra-x5.toml                 | 18 +++++++++++++++
 .../eu.mistral.pixtral-large-2502-v1:0.toml   |  8 +++++++
 .../us.meta.llama3-1-70b-instruct-v1:0.toml   |  8 +++++++
 .../us.meta.llama3-1-8b-instruct-v1:0.toml    |  8 +++++++
 .../us.meta.llama3-3-70b-instruct-v1:0.toml   |  9 ++++++++
 .../us.mistral.pixtral-large-2502-v1:0.toml   |  8 +++++++
 .../models/us.writer.palmyra-x4-v1:0.toml     |  9 ++++++++
 .../models/us.writer.palmyra-x5-v1:0.toml     |  9 ++++++++
 11 files changed, 136 insertions(+)
 create mode 100644 models/meta/llama-3.1-70b-instruct.toml
 create mode 100644 models/mistral/pixtral-large-2502.toml
 create mode 100644 models/writer/palmyra-x4.toml
 create mode 100644 models/writer/palmyra-x5.toml
 create mode 100644 providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.meta.llama3-1-70b-instruct-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.meta.llama3-1-8b-instruct-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml

diff --git a/models/meta/llama-3.1-70b-instruct.toml b/models/meta/llama-3.1-70b-instruct.toml
new file mode 100644
index 00000000000..cc71a1871d4
--- /dev/null
+++ b/models/meta/llama-3.1-70b-instruct.toml
@@ -0,0 +1,23 @@
+name = "Llama-3.1-70B-Instruct"
+description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
+family = "llama"
+release_date = "2024-07-23"
+last_updated = "2024-07-23"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2023-12"
+open_weights = true
+
+[limit]
+context = 128_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct"
diff --git a/models/mistral/pixtral-large-2502.toml b/models/mistral/pixtral-large-2502.toml
new file mode 100644
index 00000000000..28f902e7f32
--- /dev/null
+++ b/models/mistral/pixtral-large-2502.toml
@@ -0,0 +1,18 @@
+name = "Pixtral Large (25.02)"
+description = "Mistral vision-language model for image understanding and multimodal chat"
+family = "pixtral"
+release_date = "2025-04-08"
+last_updated = "2025-04-08"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/models/writer/palmyra-x4.toml b/models/writer/palmyra-x4.toml
new file mode 100644
index 00000000000..265d3ce5855
--- /dev/null
+++ b/models/writer/palmyra-x4.toml
@@ -0,0 +1,18 @@
+name = "Palmyra X4"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+family = "palmyra"
+release_date = "2025-04-28"
+last_updated = "2025-04-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 122_880
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/models/writer/palmyra-x5.toml b/models/writer/palmyra-x5.toml
new file mode 100644
index 00000000000..328c9dc5011
--- /dev/null
+++ b/models/writer/palmyra-x5.toml
@@ -0,0 +1,18 @@
+name = "Palmyra X5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+family = "palmyra"
+release_date = "2025-04-28"
+last_updated = "2025-04-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_040_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml b/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
new file mode 100644
index 00000000000..50089982857
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
@@ -0,0 +1,8 @@
+# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
+# EU geo pricing (Price List eu-west-1 standard): input/output USD per 1M tokens.
+base_model = "mistral/pixtral-large-2502"
+name = "Pixtral Large (25.02) (EU)"
+
+[cost]
+input = 2.00
+output = 6.00
diff --git a/providers/amazon-bedrock/models/us.meta.llama3-1-70b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama3-1-70b-instruct-v1:0.toml
new file mode 100644
index 00000000000..c8bb679c410
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.meta.llama3-1-70b-instruct-v1:0.toml
@@ -0,0 +1,8 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "meta/llama-3.1-70b-instruct"
+name = "Llama 3.1 70B Instruct (US)"
+
+[cost]
+input = 0.72
+output = 0.72
diff --git a/providers/amazon-bedrock/models/us.meta.llama3-1-8b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama3-1-8b-instruct-v1:0.toml
new file mode 100644
index 00000000000..0489bbbd0f7
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.meta.llama3-1-8b-instruct-v1:0.toml
@@ -0,0 +1,8 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "meta/llama-3.1-8b-instruct"
+name = "Llama 3.1 8B Instruct (US)"
+
+[cost]
+input = 0.22
+output = 0.22
diff --git a/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
new file mode 100644
index 00000000000..c0cdccb6f37
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "meta/llama-3.3-70b-instruct"
+name = "Llama 3.3 70B Instruct (US)"
+attachment = false
+
+[cost]
+input = 0.72
+output = 0.72
diff --git a/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml b/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
new file mode 100644
index 00000000000..d0e81bd25a7
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
@@ -0,0 +1,8 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "mistral/pixtral-large-2502"
+name = "Pixtral Large (25.02) (US)"
+
+[cost]
+input = 2.00
+output = 6.00
diff --git a/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml b/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
new file mode 100644
index 00000000000..f2b5bfb7981
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "writer/palmyra-x4"
+reasoning_options = []
+name = "Palmyra X4 (US)"
+
+[cost]
+input = 2.50
+output = 10.00
diff --git a/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml b/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml
new file mode 100644
index 00000000000..b56d9601c9f
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml
@@ -0,0 +1,9 @@
+# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
+# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+base_model = "writer/palmyra-x5"
+reasoning_options = []
+name = "Palmyra X5 (US)"
+
+[cost]
+input = 0.60
+output = 6.00

From 5b22f954cd1c2b92e49b66007c694cf959dbad1f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:14 +0000
Subject: [PATCH 0900/2012] chore(sync): update Merge Gateway model catalog
 (#6604)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta/llama-3.1-70b-instruct.toml            | 10 ++++++++++
 1 file changed, 10 insertions(+)
 create mode 100644 providers/merge-gateway/models/meta/llama-3.1-70b-instruct.toml

diff --git a/providers/merge-gateway/models/meta/llama-3.1-70b-instruct.toml b/providers/merge-gateway/models/meta/llama-3.1-70b-instruct.toml
new file mode 100644
index 00000000000..1a6508d569e
--- /dev/null
+++ b/providers/merge-gateway/models/meta/llama-3.1-70b-instruct.toml
@@ -0,0 +1,10 @@
+base_model = "meta/llama-3.1-70b-instruct"
+name = "Llama 3.1 70B"
+structured_output = false
+
+[cost]
+input = 0.99
+output = 0.99
+
+[limit]
+output = 2_048

From 93db50db8530a2b6033f2434672ca9d75e8e46a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:20 +0000
Subject: [PATCH 0901/2012] chore(sync): update NanoGPT model catalog (#6608)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 1502b603059c32ba15ffd38efbaff17c087ee124 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:22 +0000
Subject: [PATCH 0902/2012] chore(sync): update Kilo model catalog (#6605)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta-llama/llama-3.1-70b-instruct.toml  | 14 +-------------
 1 file changed, 1 insertion(+), 13 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 9897bdffb94..881a90e0d15 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -1,14 +1,6 @@
-name = "Meta: Llama 3.1 70B Instruct"
+base_model = "meta/llama-3.1-70b-instruct"
 description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.4
@@ -17,7 +9,3 @@ output = 0.4
 [limit]
 context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 4b71bc1acf94b899af31bbde0beb89f8bbc0f68b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:24 +0000
Subject: [PATCH 0903/2012] chore(sync): update Eden AI model catalog (#6607)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../edenai/models/amazon/amazon.nova-lite-v1:0.toml   |  9 +++++++++
 .../models/amazon/amazon.nova-lite-v1:0@us.toml       | 10 ++++++++++
 .../edenai/models/amazon/amazon.nova-micro-v1:0.toml  |  6 ++++++
 .../models/amazon/amazon.nova-micro-v1:0@us.toml      |  7 +++++++
 .../edenai/models/amazon/amazon.nova-pro-v1:0.toml    |  9 +++++++++
 .../edenai/models/amazon/amazon.nova-pro-v1:0@us.toml | 10 ++++++++++
 .../amazon/mistral.pixtral-large-2502-v1:0.toml       | 11 +++++++++++
 .../amazon/mistral.pixtral-large-2502-v1:0@us.toml    | 11 +++++++++++
 8 files changed, 73 insertions(+)
 create mode 100644 providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
 create mode 100644 providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
 create mode 100644 providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
 create mode 100644 providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
 create mode 100644 providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
 create mode 100644 providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
 create mode 100644 providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0.toml
 create mode 100644 providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0@us.toml

diff --git a/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
new file mode 100644
index 00000000000..ea127cc7b46
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
@@ -0,0 +1,9 @@
+base_model = "amazon/nova-lite"
+structured_output = true
+
+[cost]
+input = 0.06
+output = 0.24
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
new file mode 100644
index 00000000000..db196fcd529
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
@@ -0,0 +1,10 @@
+base_model = "amazon/nova-lite"
+name = "Nova Lite (US)"
+structured_output = true
+
+[cost]
+input = 0.06
+output = 0.24
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
new file mode 100644
index 00000000000..f742ca81b6b
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
@@ -0,0 +1,6 @@
+base_model = "amazon/nova-micro"
+structured_output = true
+
+[cost]
+input = 0.035
+output = 0.14
diff --git a/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
new file mode 100644
index 00000000000..c0229f11c40
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
@@ -0,0 +1,7 @@
+base_model = "amazon/nova-micro"
+name = "Nova Micro (US)"
+structured_output = true
+
+[cost]
+input = 0.035
+output = 0.14
diff --git a/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
new file mode 100644
index 00000000000..d8ba4575921
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
@@ -0,0 +1,9 @@
+base_model = "amazon/nova-pro"
+structured_output = true
+
+[cost]
+input = 0.8
+output = 3.2
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
new file mode 100644
index 00000000000..25bbe563d46
--- /dev/null
+++ b/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
@@ -0,0 +1,10 @@
+base_model = "amazon/nova-pro"
+name = "Nova Pro (US)"
+structured_output = true
+
+[cost]
+input = 0.8
+output = 3.2
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0.toml b/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0.toml
new file mode 100644
index 00000000000..b8f937de99a
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/pixtral-large-2502"
+name = "Pixtral Large (25.02) (Amazon Bedrock)"
+attachment = false
+structured_output = false
+
+[cost]
+input = 2
+output = 6
+
+[modalities]
+input = ["text"]
diff --git a/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0@us.toml b/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0@us.toml
new file mode 100644
index 00000000000..22be2866378
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.pixtral-large-2502-v1:0@us.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/pixtral-large-2502"
+name = "Pixtral Large (25.02) (Amazon Bedrock, US)"
+attachment = false
+structured_output = false
+
+[cost]
+input = 2
+output = 6
+
+[modalities]
+input = ["text"]

From cd76187ed43ffc85aaa6ddbf539a03d93ff3b183 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:26 +0000
Subject: [PATCH 0904/2012] chore(sync): update Weights & Biases model catalog
 (#6606)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta-llama/Llama-3.1-70B-Instruct.toml   | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml b/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
index 7f9934ecc96..e4025c8baa5 100644
--- a/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
+++ b/providers/wandb/models/meta-llama/Llama-3.1-70B-Instruct.toml
@@ -1,14 +1,7 @@
+base_model = "meta/llama-3.1-70b-instruct"
 name = "Llama 3.1 70B"
 description = "Efficient conversational model optimized for responsive multilingual chatbot interactions."
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.8
@@ -18,7 +11,3 @@ cache_read = 0.8
 [limit]
 context = 131_072
 output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 2898e30b39bcd67b9d01114fd25254798c81dd90 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 04:30:35 +0000
Subject: [PATCH 0905/2012] chore(sync): update OpenRouter model catalog
 (#6602)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta-llama/llama-3.1-70b-instruct.toml | 15 +--------------
 .../models/qwen/qwen3-235b-a22b-2507.toml         |  4 ++--
 2 files changed, 3 insertions(+), 16 deletions(-)

diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index b2d0588ba33..881a90e0d15 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -1,15 +1,6 @@
-name = "Llama 3.1 70B Instruct"
+base_model = "meta/llama-3.1-70b-instruct"
 description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-knowledge = "2023-12-31"
-open_weights = true
 
 [cost]
 input = 0.4
@@ -18,7 +9,3 @@ output = 0.4
 [limit]
 context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index aa706cc488b..0711c94269c 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,8 +12,8 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.55
+input = 0.22
+output = 0.88
 
 [limit]
 context = 262_144

From b352d4bc9c0ee061c52d63845d313af4a6fc391c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 05:26:55 +0000
Subject: [PATCH 0906/2012] chore(sync): update Eden AI model catalog (#6609)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From bc9d45a9dc11aaef8f9c6abfe2c2fc428eda7e44 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 06:40:51 +0000
Subject: [PATCH 0907/2012] chore(sync): update NanoGPT model catalog (#6610)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 07c890d8967f331b6ada182a50d42e9b5ec1ced5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 08:32:54 +0000
Subject: [PATCH 0908/2012] chore(sync): update OpenRouter model catalog
 (#6613)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 2ce7e2e4c07..c0c2c8b5cd7 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.04815
-output = 0.19305
+input = 0.09
+output = 0.3
 
 [limit]
 context = 262_144
-output = 32_000
+output = 235_929
 
 [modalities]
 input = ["text"]

From 5a097c5f3fb1d26dd8936c4f66bfd5e0703e777a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 08:33:12 +0000
Subject: [PATCH 0909/2012] chore(sync): update Kilo model catalog (#6614)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 48baf441187..1da6f655c94 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 128_000
-output = 32_000
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 47f8b5e9c68eb9e75173e68269fe1361ed8105a4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 08:33:37 +0000
Subject: [PATCH 0910/2012] chore(sync): update NanoGPT model catalog (#6612)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/azure-gpt-4-turbo.toml    | 23 ------------------
 .../nano-gpt/models/azure-gpt-4o-mini.toml    | 24 -------------------
 providers/nano-gpt/models/azure-gpt-4o.toml   | 24 -------------------
 providers/nano-gpt/models/azure-o1.toml       | 24 -------------------
 providers/nano-gpt/models/azure-o3-mini.toml  | 24 -------------------
 5 files changed, 119 deletions(-)
 delete mode 100644 providers/nano-gpt/models/azure-gpt-4-turbo.toml
 delete mode 100644 providers/nano-gpt/models/azure-gpt-4o-mini.toml
 delete mode 100644 providers/nano-gpt/models/azure-gpt-4o.toml
 delete mode 100644 providers/nano-gpt/models/azure-o1.toml
 delete mode 100644 providers/nano-gpt/models/azure-o3-mini.toml

diff --git a/providers/nano-gpt/models/azure-gpt-4-turbo.toml b/providers/nano-gpt/models/azure-gpt-4-turbo.toml
deleted file mode 100644
index a3722d86120..00000000000
--- a/providers/nano-gpt/models/azure-gpt-4-turbo.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Azure gpt-4-turbo"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-01"
-last_updated = "2024-01-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 10
-output = 30
-
-[limit]
-context = 128_000
-input = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/azure-gpt-4o-mini.toml b/providers/nano-gpt/models/azure-gpt-4o-mini.toml
deleted file mode 100644
index 4ff5e0d0ccf..00000000000
--- a/providers/nano-gpt/models/azure-gpt-4o-mini.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Azure gpt-4o-mini"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-01"
-last_updated = "2024-07-18"
-attachment = true
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.075
-
-[limit]
-context = 128_000
-input = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/azure-gpt-4o.toml b/providers/nano-gpt/models/azure-gpt-4o.toml
deleted file mode 100644
index a09a59fff1b..00000000000
--- a/providers/nano-gpt/models/azure-gpt-4o.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Azure gpt-4o"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-01"
-last_updated = "2024-05-13"
-attachment = true
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 2.5
-output = 10
-cache_read = 1.25
-
-[limit]
-context = 128_000
-input = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/azure-o1.toml b/providers/nano-gpt/models/azure-o1.toml
deleted file mode 100644
index e040c38d60c..00000000000
--- a/providers/nano-gpt/models/azure-o1.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Azure o1"
-description = "O-series reasoning model for hard analysis, math, coding, and planning"
-family = "o"
-release_date = "2024-01-01"
-last_updated = "2024-12-17"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 15
-output = 60
-cache_read = 7.5
-
-[limit]
-context = 200_000
-input = 200_000
-output = 100_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/azure-o3-mini.toml b/providers/nano-gpt/models/azure-o3-mini.toml
deleted file mode 100644
index 9daa6421a7c..00000000000
--- a/providers/nano-gpt/models/azure-o3-mini.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Azure o3-mini"
-description = "O-series reasoning model for hard analysis, math, coding, and planning"
-family = "o"
-release_date = "2024-01-01"
-last_updated = "2025-01-31"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 1.1
-output = 4.4
-cache_read = 0.55
-
-[limit]
-context = 200_000
-input = 200_000
-output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 2eba3466cc154c3a8cba3dd43c70a058fac7b6fa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 09:29:33 +0000
Subject: [PATCH 0911/2012] chore(sync): update OpenRouter model catalog
 (#6615)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 +++----
 .../models/qwen/qwen3-30b-a3b-instruct-2507.toml           | 6 +++---
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..5bc4e6af877 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 1.0494
+output = 3.1482
+cache_read = 0.03498
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index c0c2c8b5cd7..2ce7e2e4c07 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.3
+input = 0.04815
+output = 0.19305
 
 [limit]
 context = 262_144
-output = 235_929
+output = 32_000
 
 [modalities]
 input = ["text"]

From bfde04fba8696490c9c0a652cc9d10825897fd09 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 09:29:47 +0000
Subject: [PATCH 0912/2012] chore(sync): update Kilo model catalog (#6616)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml    | 3 +--
 providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 4 ++--
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 1da6f655c94..48baf441187 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 262_144
-output = 235_929
+context = 128_000
+output = 32_000
 
 [modalities]
 input = ["text"]

From 628c117591c586c3a42317b84bf535d238ddeb05 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 10:27:02 +0000
Subject: [PATCH 0913/2012] chore(sync): update NanoGPT model catalog (#6617)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 10a138f5dc0d1d6ffe322a8e39d593d586ca400e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 12:37:38 +0000
Subject: [PATCH 0914/2012] chore(sync): update NanoGPT model catalog (#6619)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gemma-4-12b-it-semancer.toml       | 25 +++++++++++++++++++
 .../models/gemma-4-12b-it-station-keeper.toml | 25 +++++++++++++++++++
 .../models/qwen/qwen3.8-27b-queen.toml        | 25 +++++++++++++++++++
 3 files changed, 75 insertions(+)
 create mode 100644 providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
 create mode 100644 providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml

diff --git a/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
new file mode 100644
index 00000000000..a0e66888780
--- /dev/null
+++ b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
@@ -0,0 +1,25 @@
+name = "Gemma 4 12B Semancer"
+description = "Gemma 4 12B Semancer is an open-weight roleplay finetune with image understanding, tool calling, optional reasoning, and a 131,072-token context window."
+family = "gemma"
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.05
+output = 0.25
+cache_read = 0.025
+
+[limit]
+context = 131_072
+input = 131_072
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
new file mode 100644
index 00000000000..c4ccf5f509a
--- /dev/null
+++ b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
@@ -0,0 +1,25 @@
+name = "Gemma 4 12B StationKeeper"
+description = "Gemma 4 12B StationKeeper is an open-weight roleplay finetune with image understanding, tool calling, optional reasoning, and a 131,072-token context window."
+family = "gemma"
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.05
+output = 0.25
+cache_read = 0.025
+
+[limit]
+context = 131_072
+input = 131_072
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml
new file mode 100644
index 00000000000..19c33f74ea6
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.8 27B Queen"
+description = "Qwen 3.8 27B Queen is an open-weight roleplay finetune with image understanding, tool calling, optional reasoning, and a 262,144-token context window."
+family = "qwen"
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.125
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 8039d61a6f403c25326a2e60befa35444a439686 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 13:28:12 +0000
Subject: [PATCH 0915/2012] chore(sync): update Charm Hyper model catalog
 (#6620)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.1.toml      | 6 +++---
 providers/hyper/models/gpt-oss-120b.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index b38d94327c2..befdc41c51e 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.358
-output = 4.268
-cache_write = 0.679
+input = 1.386
+output = 4.356
+cache_write = 0.693
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index b8abd9a4a66..d7c068dc409 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.178
-output = 0.68
-cache_write = 0.089
+input = 0.18
+output = 0.61
+cache_write = 0.09
 
 [limit]
 context = 128_072

From a406f46db11180ab732c2be0922f3f9f5bf08b3d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 13:28:14 +0000
Subject: [PATCH 0916/2012] chore(sync): update OpenRouter model catalog
 (#6622)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..e57f7b0b838 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.953172
+output = 1.906344
+cache_read = 0.079431
 
 [limit]
 context = 1_048_576

From 6058d3220bf5d2223776201795438fa84795e423 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 13:28:31 +0000
Subject: [PATCH 0917/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6621)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.2-fast.toml   | 6 +++---
 providers/llmgateway/models/glm-5.2.toml        | 6 +++---
 providers/llmgateway/models/glm-5.3-flash.toml  | 4 ++--
 providers/llmgateway/models/kimi-k2.7-code.toml | 6 +++---
 providers/llmgateway/models/kimi-k3.toml        | 6 +++---
 providers/llmgateway/models/qwen3.8-max.toml    | 6 +++---
 6 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.2-fast.toml b/providers/llmgateway/models/glm-5.2-fast.toml
index 8b274c33bcc..b281adf9f5f 100644
--- a/providers/llmgateway/models/glm-5.2-fast.toml
+++ b/providers/llmgateway/models/glm-5.2-fast.toml
@@ -2,6 +2,6 @@ base_model = "zhipuai/glm-5.2"
 reasoning_options = []
 
 [cost]
-input = 1.99
-output = 6.16
-cache_read = 0.4
+input = 2.2
+output = 6.5
+cache_read = 0.45
diff --git a/providers/llmgateway/models/glm-5.2.toml b/providers/llmgateway/models/glm-5.2.toml
index a6d3d351263..f2b344230e3 100644
--- a/providers/llmgateway/models/glm-5.2.toml
+++ b/providers/llmgateway/models/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.55
-output = 1.784
-cache_read = 0.111
+input = 0.8
+output = 2.55
+cache_read = 0.16
 cache_write = 0
 
 [limit]
diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index 035cb3a9470..d46987e8acc 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.1
+input = 0.088
 output = 0.25
-cache_read = 0.02
+cache_read = 0.025
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway/models/kimi-k2.7-code.toml b/providers/llmgateway/models/kimi-k2.7-code.toml
index 082253a616a..bf7139b1230 100644
--- a/providers/llmgateway/models/kimi-k2.7-code.toml
+++ b/providers/llmgateway/models/kimi-k2.7-code.toml
@@ -11,6 +11,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.89
-output = 3.71
-cache_read = 0.18
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/llmgateway/models/kimi-k3.toml b/providers/llmgateway/models/kimi-k3.toml
index 5196be431f3..b4adfd00e15 100644
--- a/providers/llmgateway/models/kimi-k3.toml
+++ b/providers/llmgateway/models/kimi-k3.toml
@@ -16,6 +16,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.83
-output = 14.13
-cache_read = 0.28
+input = 3
+output = 15
+cache_read = 0.3
diff --git a/providers/llmgateway/models/qwen3.8-max.toml b/providers/llmgateway/models/qwen3.8-max.toml
index e6935723bbf..dde7b85e514 100644
--- a/providers/llmgateway/models/qwen3.8-max.toml
+++ b/providers/llmgateway/models/qwen3.8-max.toml
@@ -8,9 +8,9 @@ values = ["low", "medium", "xhigh"]
 type = "budget_tokens"
 
 [cost]
-input = 1.815
-output = 5.4461
-cache_read = 0.21
+input = 2
+output = 6
+cache_read = 0.25
 cache_write = 2.5
 
 [limit]

From 1b20fabc13186486f089c7f68b68787e8a325af0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 13:28:36 +0000
Subject: [PATCH 0918/2012] chore(sync): update LLM Gateway model catalog
 (#6623)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/scx-ai-gp/glm-5.2-fast.toml | 6 +++---
 .../llmgateway-providers/models/scx-ai-gp/glm-5.2.toml      | 6 +++---
 .../models/scx-ai-gp/glm-5.3-flash.toml                     | 6 +++---
 .../llmgateway-providers/models/scx-ai-gp/glm-5.3.toml      | 6 +++---
 .../models/scx-ai-gp/kimi-k2.7-code.toml                    | 6 +++---
 .../llmgateway-providers/models/scx-ai-gp/kimi-k3.toml      | 6 +++---
 .../llmgateway-providers/models/scx-ai-gp/qwen3.8-max.toml  | 6 +++---
 7 files changed, 21 insertions(+), 21 deletions(-)

diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2-fast.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2-fast.toml
index b20b26804d6..246625941ca 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2-fast.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2-fast.toml
@@ -4,6 +4,6 @@ structured_output = false
 reasoning_options = []
 
 [cost]
-input = 1.99
-output = 6.16
-cache_read = 0.4
+input = 2.2
+output = 6.5
+cache_read = 0.45
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
index a5e3a15d492..3a65bed82fd 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
@@ -10,6 +10,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.55
-output = 1.784
-cache_read = 0.111
+input = 0.8
+output = 2.55
+cache_read = 0.16
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
index 3be5d36f155..1c1bd09c1a7 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3-flash.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.13
-output = 0.4
-cache_read = 0.024
+input = 0.088
+output = 0.25
+cache_read = 0.025
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml
index 9e9ce7cdee3..2c130d77f25 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.3
-output = 4
-cache_read = 0.25
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
 output = 128_000
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
index a52f9e11855..08899a30020 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k2.7-code.toml
@@ -10,6 +10,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.89
-output = 3.71
-cache_read = 0.18
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml
index 665f42b2175..00e8ea93776 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/kimi-k3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.83
-output = 14.13
-cache_read = 0.28
+input = 3.5
+output = 18
+cache_read = 0.35
 
 [limit]
 output = 1_048_576
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/qwen3.8-max.toml b/providers/llmgateway-providers/models/scx-ai-gp/qwen3.8-max.toml
index 89d3c99b3d4..92e0feca6ac 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/qwen3.8-max.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/qwen3.8-max.toml
@@ -10,7 +10,7 @@ values = ["low", "medium", "xhigh"]
 type = "budget_tokens"
 
 [cost]
-input = 1.815
-output = 5.4461
-cache_read = 0.21
+input = 2
+output = 6
+cache_read = 0.25
 cache_write = 2.5

From c13d17bca2f513b804d381cc6d88e4ec258fb64a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:27:46 +0000
Subject: [PATCH 0919/2012] chore(sync): update LLM Gateway model catalog
 (#6625)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/glm-5.3-flash.toml                 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
index d53e927044b..4fd6e3fc0b3 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/glm-5.3-flash.toml
@@ -4,7 +4,7 @@ structured_output = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "high", "max"]
 
 [cost]
 input = 0.1

From 016db50817675fd8382e5ae65c0a6dcb1ffa63ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:27:54 +0000
Subject: [PATCH 0920/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#6627)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3.toml           | 20 -------------------
 1 file changed, 20 deletions(-)
 delete mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
deleted file mode 100644
index a4483d01942..00000000000
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-# Cloudflare model input schema verified 2026-08-28 for @cf/zai-org/glm-5.3.
-# https://developers.cloudflare.com/api/resources/ai/subresources/models/subresources/schema/methods/get/
-# Toggle: chat_template_kwargs.enable_thinking = true|false (default true).
-# Effort: reasoning_effort = low|medium|high; Cloudflare's declared controls differ from Z.ai's API.
-base_model = "zhipuai/glm-5.3"
-name = "Glm 5.3"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["low", "medium", "high"] },
-]
-
-[cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
-
-[limit]
-context = 1_310_720
-output = 1_310_720

From f45023255119776bad8021d755424d2809557c5d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:28:09 +0000
Subject: [PATCH 0921/2012] chore(sync): update Kilo model catalog (#6626)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216

From db22938c9ec88b14dd73c31e7a94370a27583cc4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:29:22 +0000
Subject: [PATCH 0922/2012] chore(sync): update OpenRouter model catalog
 (#6628)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 ++++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml              | 2 +-
 providers/openrouter/models/~z-ai/glm-latest.toml          | 2 +-
 4 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5bc4e6af877..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0494
-output = 3.1482
-cache_read = 0.03498
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e57f7b0b838..a010dd38ba2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.953172
-output = 1.906344
-cache_read = 0.079431
+input = 0.87
+output = 1.74
+cache_read = 0.0725
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index d939b2691d5..3f546789070 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -11,5 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_310_720
+context = 1_048_576
 output = 943_718
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 91a5845fc7b..784c165c505 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -20,7 +20,7 @@ output = 3.498
 cache_read = 0.2067
 
 [limit]
-context = 1_310_720
+context = 1_048_576
 output = 128_000
 
 [modalities]

From f9dd6cb6e6a38f3e939c8eb0d764f262817dd88b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 15:27:19 +0000
Subject: [PATCH 0923/2012] chore(sync): update OpenRouter model catalog
 (#6631)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From 264a30f36b1bee021cd0f47a8b8c152b124ab07a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 15:28:16 +0000
Subject: [PATCH 0924/2012] chore(sync): update Eden AI model catalog (#6630)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 7 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index ae74bc38aee..74cfaa35ff4 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -5,5 +5,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75491
-output = 0.75491
+input = 0.75738
+output = 0.75738
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 80db68aa28e..f6d21095100 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17421
-output = 0.75491
+input = 0.17478
+output = 0.75738
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index e8d614f3c78..bb5dfdbfa65 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46456
-output = 0.92912
+input = 0.46608
+output = 0.93216
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 50eaa6eb24f..6090daee388 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17421
-output = 0.69684
+input = 0.17478
+output = 0.69912
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index bb4a9388c3a..f14d2b0ce24 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04526
-output = 1.04526
+input = 1.04868
+output = 1.04868

From 8d079ab2cc7ba67b2932a3efaa79fd948f70afc3 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 10 Sep 2026 00:03:11 +0800
Subject: [PATCH 0925/2012] fix(go): end GLM-5.3-Flash promotion

---
 providers/opencode-go/models/glm-5.3-flash.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/opencode-go/models/glm-5.3-flash.toml b/providers/opencode-go/models/glm-5.3-flash.toml
index 5e54a54c6b3..f0c6c886ed3 100644
--- a/providers/opencode-go/models/glm-5.3-flash.toml
+++ b/providers/opencode-go/models/glm-5.3-flash.toml
@@ -1,11 +1,11 @@
 base_model = "zhipuai/glm-5.3-flash"
-name = "GLM-5.3-Flash (2x usage)"
+name = "GLM-5.3-Flash"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03

From 5b131aa39f720680b6d0173c98836c12bbdeb2aa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 16:29:14 +0000
Subject: [PATCH 0926/2012] chore(sync): update CrossModel model catalog
 (#6634)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/crossmodel/models/z-ai/glm-5.3-flash.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/crossmodel/models/z-ai/glm-5.3-flash.toml b/providers/crossmodel/models/z-ai/glm-5.3-flash.toml
index b5971512328..6a3e9e0d3b9 100644
--- a/providers/crossmodel/models/z-ai/glm-5.3-flash.toml
+++ b/providers/crossmodel/models/z-ai/glm-5.3-flash.toml
@@ -5,10 +5,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
-cache_write = 0.075
+input = 0.15
+output = 0.5
+cache_read = 0.03
+cache_write = 0.15
 
 [limit]
 output = 128_000

From 1126f91fda92e844bd0cc7639770a2e4a731230f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 16:29:18 +0000
Subject: [PATCH 0927/2012] chore(sync): update Kilo model catalog (#6636)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nex-agi/nex-n2.5-mini:free.toml    |  4 +--
 .../models/nousresearch/hermes-4-70b.toml     | 27 -------------------
 providers/kilo/models/tencent/hy3.toml        |  6 ++---
 providers/kilo/models/z-ai/glm-5.2.toml       |  3 ++-
 .../kilo/models/~z-ai/glm-flash-latest.toml   |  2 +-
 5 files changed, 8 insertions(+), 34 deletions(-)
 delete mode 100644 providers/kilo/models/nousresearch/hermes-4-70b.toml

diff --git a/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
index 59d07170b17..2618f50eb27 100644
--- a/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
+++ b/providers/kilo/models/nex-agi/nex-n2.5-mini:free.toml
@@ -3,7 +3,7 @@ description = "Nex-N2.5 is an agentic model built to turn goals into working, ve
 family = "agi"
 release_date = "2026-09-08"
 last_updated = "2026-09-08"
-attachment = false
+attachment = true
 reasoning = true
 temperature = true
 tool_call = true
@@ -23,5 +23,5 @@ context = 262_144
 output = 235_929
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]
diff --git a/providers/kilo/models/nousresearch/hermes-4-70b.toml b/providers/kilo/models/nousresearch/hermes-4-70b.toml
deleted file mode 100644
index b7876a5d850..00000000000
--- a/providers/kilo/models/nousresearch/hermes-4-70b.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Nous: Hermes 4 70B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "nousresearch"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.13
-output = 0.4
-
-[limit]
-context = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bb89edf75cf..dcb3cc80d7f 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,4 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_024_000
+output = 128_000
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..e4ce4e5dfe6 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 37f32d9b8d76a7e407a41816f9d2a3f8af976048 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 16:29:21 +0000
Subject: [PATCH 0928/2012] chore(sync): update OpenRouter model catalog
 (#6635)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../models/nex-agi/nex-n2.5-mini:free.toml    |  4 +--
 .../models/nousresearch/hermes-4-70b.toml     | 27 -------------------
 providers/openrouter/models/tencent/hy3.toml  |  6 ++---
 providers/openrouter/models/z-ai/glm-5.2.toml |  7 ++---
 .../models/~z-ai/glm-flash-latest.toml        |  2 +-
 6 files changed, 13 insertions(+), 39 deletions(-)
 delete mode 100644 providers/openrouter/models/nousresearch/hermes-4-70b.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
index 40080615553..b2ba3f5965a 100644
--- a/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-mini:free.toml
@@ -3,7 +3,7 @@ description = "Efficient model for low-latency assistance, extraction, and routi
 family = "agi"
 release_date = "2026-09-08"
 last_updated = "2026-09-08"
-attachment = false
+attachment = true
 reasoning = true
 temperature = true
 tool_call = true
@@ -23,5 +23,5 @@ context = 262_144
 output = 235_929
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]
diff --git a/providers/openrouter/models/nousresearch/hermes-4-70b.toml b/providers/openrouter/models/nousresearch/hermes-4-70b.toml
deleted file mode 100644
index 74f7f018fd1..00000000000
--- a/providers/openrouter/models/nousresearch/hermes-4-70b.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Hermes 4 70B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "hermes"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = false
-structured_output = false
-knowledge = "2024-08-31"
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.13
-output = 0.4
-
-[limit]
-context = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 64652242a0e..c38d06275a4 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.966
-output = 3.036
-cache_read = 0.1932
+input = 0.28
+output = 0.88
+cache_read = 0.052
 
 [limit]
 context = 1_048_576
+output = 128_000
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..c32fca67d4b 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From 9fc693ffb14f4ac26d1f21594baf938b05a57ad3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 16:29:26 +0000
Subject: [PATCH 0929/2012] chore(sync): update NanoGPT model catalog (#6629)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/agnes-3.0-flash.toml      | 24 ++++++++++++++++
 .../deepseek-v4.1-flash:thinking.toml         | 28 +++++++++++++++++++
 2 files changed, 52 insertions(+)
 create mode 100644 providers/nano-gpt/models/agnes-3.0-flash.toml
 create mode 100644 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml

diff --git a/providers/nano-gpt/models/agnes-3.0-flash.toml b/providers/nano-gpt/models/agnes-3.0-flash.toml
new file mode 100644
index 00000000000..3906744eea4
--- /dev/null
+++ b/providers/nano-gpt/models/agnes-3.0-flash.toml
@@ -0,0 +1,24 @@
+name = "Agnes 3.0 Flash"
+description = "Agnes 3.0 Flash is a low-cost model for coding, tool use, and multi-turn agent tasks. It supports text and image input, optional thinking, and a 512K-token context window."
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.05
+output = 0.15
+cache_read = 0.005
+
+[limit]
+context = 524_288
+input = 524_288
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
new file mode 100644
index 00000000000..8bf595069bc
--- /dev/null
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek V4.1 Flash Thinking"
+description = "DeepSeek V4.1 Flash supports text and image input, reasoning, tool calling, and structured output with a 1M-token context window. This is a rate-limited beta with limited capacity, intended for testing rather than production use. Assume prompts and responses are logged by the provider and may be used for model training or service improvement. Do not send sensitive or confidential data."
+family = "deepseek"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.1562
+output = 0.3124
+cache_read = 0.03124
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 819be3f94c1e047ba694ff0cdd00e88c549af8b5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 18:33:50 +0000
Subject: [PATCH 0930/2012] chore(sync): update OpenRouter model catalog
 (#6640)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..3eee6643f71 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.07
+output = 0.14
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From e840872551c82dce49fe16c805215ec7f0395390 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 18:35:41 +0000
Subject: [PATCH 0931/2012] chore(sync): update Charm Hyper model catalog
 (#6641)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-flash-0731.toml | 2 +-
 providers/hyper/models/qwen3.6-flash.toml          | 5 ++++-
 providers/hyper/models/qwen3.6-max.toml            | 5 ++++-
 providers/hyper/models/qwen3.6-plus.toml           | 5 ++++-
 providers/hyper/models/qwen3.7-flash.toml          | 5 ++++-
 providers/hyper/models/qwen3.7-max.toml            | 5 ++++-
 providers/hyper/models/qwen3.7-plus.toml           | 5 ++++-
 providers/hyper/models/qwen3.8-2.4t-a95b.toml      | 5 ++++-
 providers/hyper/models/qwen3.8-27b.toml            | 5 ++++-
 providers/hyper/models/qwen3.8-flash.toml          | 5 ++++-
 providers/hyper/models/qwen3.8-max.toml            | 5 ++++-
 11 files changed, 41 insertions(+), 11 deletions(-)

diff --git a/providers/hyper/models/deepseek-v4-flash-0731.toml b/providers/hyper/models/deepseek-v4-flash-0731.toml
index f089eb593d4..dda2f29cdd0 100644
--- a/providers/hyper/models/deepseek-v4-flash-0731.toml
+++ b/providers/hyper/models/deepseek-v4-flash-0731.toml
@@ -4,7 +4,7 @@ last_updated = "2026-08-02"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high", "max"]
+values = ["low", "high", "max"]
 
 [cost]
 input = 0.44
diff --git a/providers/hyper/models/qwen3.6-flash.toml b/providers/hyper/models/qwen3.6-flash.toml
index d686f6e3ca8..0554b788c42 100644
--- a/providers/hyper/models/qwen3.6-flash.toml
+++ b/providers/hyper/models/qwen3.6-flash.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.6-flash"
 release_date = "2026-05-20"
 last_updated = "2026-07-22"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 1
diff --git a/providers/hyper/models/qwen3.6-max.toml b/providers/hyper/models/qwen3.6-max.toml
index c333833038d..f954b161903 100644
--- a/providers/hyper/models/qwen3.6-max.toml
+++ b/providers/hyper/models/qwen3.6-max.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.6-max-preview"
 release_date = "2026-05-20"
 last_updated = "2026-07-22"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 2
diff --git a/providers/hyper/models/qwen3.6-plus.toml b/providers/hyper/models/qwen3.6-plus.toml
index 7bf8a3d4c4b..b5db258f7f9 100644
--- a/providers/hyper/models/qwen3.6-plus.toml
+++ b/providers/hyper/models/qwen3.6-plus.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.6-plus"
 release_date = "2026-05-20"
 last_updated = "2026-07-22"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 2
diff --git a/providers/hyper/models/qwen3.7-flash.toml b/providers/hyper/models/qwen3.7-flash.toml
index 4f4691bb8af..e5ed200d633 100644
--- a/providers/hyper/models/qwen3.7-flash.toml
+++ b/providers/hyper/models/qwen3.7-flash.toml
@@ -2,7 +2,10 @@ base_model = "alibaba/qwen3.7-flash"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 release_date = "2026-07-27"
 last_updated = "2026-07-30"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0.2
diff --git a/providers/hyper/models/qwen3.7-max.toml b/providers/hyper/models/qwen3.7-max.toml
index 745d85c8ecb..1e03fe5328f 100644
--- a/providers/hyper/models/qwen3.7-max.toml
+++ b/providers/hyper/models/qwen3.7-max.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.7-max"
 release_date = "2026-05-28"
 last_updated = "2026-07-22"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 2.5
diff --git a/providers/hyper/models/qwen3.7-plus.toml b/providers/hyper/models/qwen3.7-plus.toml
index e808ce99ff1..781ddcaf367 100644
--- a/providers/hyper/models/qwen3.7-plus.toml
+++ b/providers/hyper/models/qwen3.7-plus.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.7-plus"
 release_date = "2026-06-15"
 last_updated = "2026-07-22"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 1.2
diff --git a/providers/hyper/models/qwen3.8-2.4t-a95b.toml b/providers/hyper/models/qwen3.8-2.4t-a95b.toml
index 5de67d80511..00fdf314a69 100644
--- a/providers/hyper/models/qwen3.8-2.4t-a95b.toml
+++ b/providers/hyper/models/qwen3.8-2.4t-a95b.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.8-2.4t-a95b"
 release_date = "2026-08-26"
 last_updated = "2026-08-26"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
 
 [cost]
 input = 2
diff --git a/providers/hyper/models/qwen3.8-27b.toml b/providers/hyper/models/qwen3.8-27b.toml
index ae043f3c6d1..fd47a185674 100644
--- a/providers/hyper/models/qwen3.8-27b.toml
+++ b/providers/hyper/models/qwen3.8-27b.toml
@@ -1,7 +1,10 @@
 base_model = "alibaba/qwen3.8-27b"
 release_date = "2026-08-26"
 last_updated = "2026-08-26"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0.5
diff --git a/providers/hyper/models/qwen3.8-flash.toml b/providers/hyper/models/qwen3.8-flash.toml
index 23e537cd5d4..f95aa060385 100644
--- a/providers/hyper/models/qwen3.8-flash.toml
+++ b/providers/hyper/models/qwen3.8-flash.toml
@@ -1,6 +1,9 @@
 base_model = "alibaba/qwen3.8-flash"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0.15
diff --git a/providers/hyper/models/qwen3.8-max.toml b/providers/hyper/models/qwen3.8-max.toml
index 03488defed5..88d26a05f7e 100644
--- a/providers/hyper/models/qwen3.8-max.toml
+++ b/providers/hyper/models/qwen3.8-max.toml
@@ -1,5 +1,8 @@
 base_model = "alibaba/qwen3.8-max-preview"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 2

From bd911e59db63d6012b42c60ffac4b45efac6aa68 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 18:35:52 +0000
Subject: [PATCH 0932/2012] chore(sync): update NanoGPT model catalog (#6642)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/inclusionai/ling-3.0-flash-vl.toml | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/nano-gpt/models/inclusionai/ling-3.0-flash-vl.toml

diff --git a/providers/nano-gpt/models/inclusionai/ling-3.0-flash-vl.toml b/providers/nano-gpt/models/inclusionai/ling-3.0-flash-vl.toml
new file mode 100644
index 00000000000..d4212f529b9
--- /dev/null
+++ b/providers/nano-gpt/models/inclusionai/ling-3.0-flash-vl.toml
@@ -0,0 +1,28 @@
+name = "Ling 3.0 Flash VL"
+description = "Ling 3.0 Flash VL is inclusionAI's native multimodal Mixture-of-Experts model with 124B total parameters and 5.5B active parameters per token. It combines image and video understanding with reasoning and tool use for document analysis, charts, visual verification, and interface-based agent tasks. Thinking is enabled by default and can be turned off in settings."
+family = "ling"
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.06
+output = 0.18
+cache_read = 0.012
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

From 1d73fd965d6a1f9c60d9baf1b16d5ade8cea3a75 Mon Sep 17 00:00:00 2001
From: "xiaojie.zj" <799732469@qq.com>
Date: Thu, 10 Sep 2026 03:10:21 +0800
Subject: [PATCH 0933/2012] chore(zenmux): mark 1 offline models as deprecated
 (#6638)

Models taken offline in ZenMux (visible=2):
- openai/gpt-5.1-chat
---
 providers/zenmux/models/openai/gpt-5.1-chat.toml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/providers/zenmux/models/openai/gpt-5.1-chat.toml b/providers/zenmux/models/openai/gpt-5.1-chat.toml
index 33972286e34..c35f6c1085e 100644
--- a/providers/zenmux/models/openai/gpt-5.1-chat.toml
+++ b/providers/zenmux/models/openai/gpt-5.1-chat.toml
@@ -9,6 +9,8 @@ tool_call = true
 knowledge = "2025-01-01"
 open_weights = false
 
+status = "deprecated"
+
 [cost]
 input = 1.25
 output = 10.00

From f1f68c1f56a2538ce9936b551734fcffcce1da19 Mon Sep 17 00:00:00 2001
From: Mykola Biron <mykolabiron@gmail.com>
Date: Wed, 9 Sep 2026 21:10:30 +0200
Subject: [PATCH 0934/2012] feat: add logos for databricks, wandb, clarifai and
 lmstudio (#6643)

These four providers ship no logo.svg, so the catalog serves the default
placeholder for them. Marks come from simple-icons (CC0) and are
normalized to the convention in AGENTS.md: one path, square 0 0 24 24
viewBox, fill="currentColor", no fixed width or height.
---
 providers/clarifai/logo.svg   | 3 +++
 providers/databricks/logo.svg | 3 +++
 providers/lmstudio/logo.svg   | 3 +++
 providers/wandb/logo.svg      | 3 +++
 4 files changed, 12 insertions(+)
 create mode 100644 providers/clarifai/logo.svg
 create mode 100644 providers/databricks/logo.svg
 create mode 100644 providers/lmstudio/logo.svg
 create mode 100644 providers/wandb/logo.svg

diff --git a/providers/clarifai/logo.svg b/providers/clarifai/logo.svg
new file mode 100644
index 00000000000..d15444310f2
--- /dev/null
+++ b/providers/clarifai/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M2.78 14.772A2.768 2.768 0 0 1 0 12.008a2.781 2.781 0 0 1 2.78-2.78 2.765 2.765 0 0 1 2.764 2.764c0 1.52-1.23 2.78-2.764 2.78zM17.119 1.83c-2.05 0-2.142 1.732-5.27 1.732-3.16 0-3.419-1.732-5.287-1.732a2.765 2.765 0 0 0-2.765 2.765 2.775 2.775 0 0 0 2.765 2.764c2.066 0 2.157-1.731 5.286-1.731 3.16 0 3.418 1.731 5.286 1.731A2.765 2.765 0 0 0 19.9 4.595a2.778 2.778 0 0 0-2.78-2.765zm0 14.81c-2.066 0-2.157 1.732-5.286 1.732-3.16 0-3.418-1.731-5.286-1.731a2.765 2.765 0 0 0-2.765 2.764 2.775 2.775 0 0 0 2.765 2.765c2.066 0 2.157-1.732 5.286-1.732 3.16 0 3.418 1.732 5.286 1.732a2.765 2.765 0 0 0 0-5.53zm4.101-7.397c-2.08 0-2.005 1.823-4.617 1.823-2.416-.015-2.446-1.823-4.755-1.823-1.443 0-2.628 1.322-2.628 2.765s1.2 2.764 2.765 2.764c2.08 0 2.005-1.823 4.618-1.823 2.415 0 2.445 1.808 4.77 1.808 1.442 0 2.627-1.322 2.627-2.765 0-1.427-1.2-2.749-2.78-2.749Z" />
+</svg>
diff --git a/providers/databricks/logo.svg b/providers/databricks/logo.svg
new file mode 100644
index 00000000000..2dce2836297
--- /dev/null
+++ b/providers/databricks/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M.95 14.184L12 20.403l9.919-5.55v2.21L12 22.662l-10.484-5.96-.565.308v.77L12 24l11.05-6.218v-4.317l-.515-.309L12 19.118l-9.867-5.653v-2.21L12 16.805l11.05-6.218V6.32l-.515-.308L12 11.974 2.647 6.681 12 1.388l7.76 4.368.668-.411v-.566L12 0 .95 6.27v.72L12 13.207l9.919-5.55v2.26L12 15.52 1.516 9.56l-.565.308Z" />
+</svg>
diff --git a/providers/lmstudio/logo.svg b/providers/lmstudio/logo.svg
new file mode 100644
index 00000000000..bef02e80c29
--- /dev/null
+++ b/providers/lmstudio/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M14.025 0c3.492 0 5.237 0 6.571.68a6.24 6.24 0 0 1 2.725 2.724C24 4.738 24 6.484 24 9.975v4.05c0 3.492 0 5.237-.68 6.571a6.24 6.24 0 0 1-2.724 2.725c-1.334.679-3.08.679-6.571.679h-4.05c-3.492 0-5.237 0-6.571-.68A6.24 6.24 0 0 1 .68 20.597C0 19.262 0 17.516 0 14.025v-4.05c0-3.492 0-5.237.68-6.571A6.23 6.23 0 0 1 3.404.68C4.738 0 6.484 0 9.975 0zM7.688 16.313a1.313 1.313 0 0 0 0 2.625h11.625a1.313 1.313 0 0 0 0-2.625zm-3-3.75a1.313 1.313 0 0 0 0 2.624h11.625a1.313 1.313 0 0 0 0-2.624zm3-3.75a1.313 1.313 0 0 0 0 2.624h11.625a1.313 1.313 0 0 0 0-2.624zm-3-3.75a1.313 1.313 0 0 0 0 2.625h11.625a1.313 1.313 0 0 0 0-2.625z" />
+</svg>
diff --git a/providers/wandb/logo.svg b/providers/wandb/logo.svg
new file mode 100644
index 00000000000..6faa9e9ae5b
--- /dev/null
+++ b/providers/wandb/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M2.48 0a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm19.04 0a1.55 1.55 0 1 0 0 3.101 1.55 1.55 0 0 0 0-3.101zM12 2.295a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zM2.48 5.272a2.48 2.48 0 1 0 0 4.96 2.48 2.48 0 0 0 0-4.96zm19.04 0a2.48 2.48 0 1 0 0 4.96 2.48 2.48 0 0 0 0-4.96zM12 8.496a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm-9.52 3.907a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm19.04 0a1.55 1.55 0 1 0 0 3.102 1.55 1.55 0 0 0 0-3.102zM12 13.767a2.48 2.48 0 1 0 0 4.962 2.48 2.48 0 0 0 0-4.962zm-9.52 3.907a2.48 2.48 0 1 0 .001 4.962 2.48 2.48 0 0 0 0-4.962zm19.04.93a1.55 1.55 0 1 0 0 3.102 1.55 1.55 0 0 0 0-3.101zM12 20.9a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1Z" />
+</svg>

From 032d13dc09ca44a5d5890d6413b7fa819bd820f2 Mon Sep 17 00:00:00 2001
From: Michael Chin <mchin188@yahoo.com>
Date: Wed, 9 Sep 2026 12:11:06 -0700
Subject: [PATCH 0935/2012] indicate native structured output support for
 Bedrock GPT-6 Astra (#6593)

---
 providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml | 1 -
 providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml     | 1 -
 2 files changed, 2 deletions(-)

diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
index e03babec72a..8fd07d08872 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
@@ -6,7 +6,6 @@ base_model = "openai/gpt-6-astra"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-6 Astra (Global)"
-structured_output = false
 
 [cost]
 input = 10.00
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
index faa2fdd8679..eabeed2aede 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
@@ -6,7 +6,6 @@ base_model = "openai/gpt-6-astra"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-6 Astra (US)"
-structured_output = false
 
 [cost]
 input = 11.00

From 3bcd2cf81cf8ee644e86ace7e9766e65a747fb85 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:11:18 -0500
Subject: [PATCH 0936/2012] chore(sync): update Cortecs model catalog (#6603)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/nova-2-lite.toml        | 10 +---------
 providers/cortecs/models/pixtral-large-2502.toml |  9 +--------
 2 files changed, 2 insertions(+), 17 deletions(-)

diff --git a/providers/cortecs/models/nova-2-lite.toml b/providers/cortecs/models/nova-2-lite.toml
index dea37eb50ba..85af7e41e26 100644
--- a/providers/cortecs/models/nova-2-lite.toml
+++ b/providers/cortecs/models/nova-2-lite.toml
@@ -1,13 +1,7 @@
-name = "nova-2-lite"
+base_model = "amazon/nova-2-lite"
 description = "Nova 2 Lite is an advanced multimodal reasoning model that combines efficiency and performance, delivering reliable AI for agentic workflows and enterprise applications."
-release_date = "2025-12-04"
-last_updated = "2025-12-04"
-attachment = true
-reasoning = true
 temperature = false
-tool_call = true
 structured_output = true
-open_weights = false
 reasoning_options = []
 
 [cost]
@@ -15,9 +9,7 @@ input = 0.373
 output = 3.144
 
 [limit]
-context = 1_000_000
 output = 65_535
 
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/cortecs/models/pixtral-large-2502.toml b/providers/cortecs/models/pixtral-large-2502.toml
index 732f24f7613..8ee501e9620 100644
--- a/providers/cortecs/models/pixtral-large-2502.toml
+++ b/providers/cortecs/models/pixtral-large-2502.toml
@@ -1,13 +1,8 @@
-name = "pixtral-large-2502"
+base_model = "mistral/pixtral-large-2502"
 description = "Pixtral Large (25.02) is a 124B open-weight multimodal model built on Mistral Large 2, offering advanced image understanding and strong performance across text and code tasks."
-release_date = "2025-05-26"
-last_updated = "2025-05-26"
-attachment = true
 reasoning = true
 temperature = false
-tool_call = true
 structured_output = true
-open_weights = false
 reasoning_options = []
 
 [cost]
@@ -15,9 +10,7 @@ input = 1.993
 output = 5.978
 
 [limit]
-context = 128_000
 output = 128_000
 
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]

From a9bf5cf8202bfb9186ec166380e296ae1ead9981 Mon Sep 17 00:00:00 2001
From: Robert Keus <robert@brothersmedia.nl>
Date: Wed, 9 Sep 2026 21:11:32 +0200
Subject: [PATCH 0937/2012] feat(greenpt): add GLM-5.3 and GLM-5.3 Flash
 (#6618)

---
 providers/greenpt/models/glm-5.3-flash.toml | 17 +++++++++++++++++
 providers/greenpt/models/glm-5.3.toml       | 13 +++++++++++++
 2 files changed, 30 insertions(+)
 create mode 100644 providers/greenpt/models/glm-5.3-flash.toml
 create mode 100644 providers/greenpt/models/glm-5.3.toml

diff --git a/providers/greenpt/models/glm-5.3-flash.toml b/providers/greenpt/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..94a88aa1d38
--- /dev/null
+++ b/providers/greenpt/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+# Cost: GreenPT EUR list prices converted at 1.1614 USD/EUR (ECB, 2026-09-08).
+# Sources: https://docs.greenpt.ai/pricing and https://docs.greenpt.ai/model-cards
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# GreenPT accepts reasoning_effort = none|minimal|low|medium|high.
+# Source: https://docs.greenpt.ai/reasoning
+# Cache writes are not charged: https://docs.greenpt.ai/prompt-caching
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
+attachment = false
+
+[modalities]
+input = ["text"]
+
+[cost]
+input = 0.127754
+cache_read = 0.0255508
+output = 0.511016
diff --git a/providers/greenpt/models/glm-5.3.toml b/providers/greenpt/models/glm-5.3.toml
new file mode 100644
index 00000000000..a111f618605
--- /dev/null
+++ b/providers/greenpt/models/glm-5.3.toml
@@ -0,0 +1,13 @@
+# Cost: GreenPT EUR list prices converted at 1.1614 USD/EUR (ECB, 2026-09-08).
+# Sources: https://docs.greenpt.ai/pricing and https://docs.greenpt.ai/model-cards
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# GreenPT accepts reasoning_effort = none|minimal|low|medium|high.
+# Source: https://docs.greenpt.ai/reasoning
+# Cache writes are not charged: https://docs.greenpt.ai/prompt-caching
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 1.27754
+cache_read = 0.319385
+output = 5.11016

From 9faf0c52829ae15a68137dfee05ae29a69e6ddf5 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 14:14:12 -0500
Subject: [PATCH 0938/2012] fix: [missing-model] cloudflare-workers-ai:
 @cf/zai-org/glm-5.3 (#6633)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3.toml           | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
new file mode 100644
index 00000000000..a5547585252
--- /dev/null
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
@@ -0,0 +1,24 @@
+# Cloudflare Workers AI host for Z.ai GLM-5.3 (paid plan / AI Gateway credits).
+# Context window and pricing from Cloudflare model page.
+# Toggle: chat_template_kwargs.enable_thinking = true|false (default true)
+# Effort: reasoning_effort = low|medium|high
+# https://developers.cloudflare.com/workers-ai/models/glm-5.3/
+# https://developers.cloudflare.com/workers-ai/models/glm-5.3/sync-input.json
+# https://developers.cloudflare.com/ai/models/%40cf/zai-org/glm-5.3/
+base_model = "zhipuai/glm-5.3"
+name = "Glm 5.3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576

From 35a2ed9758828e2bb762a648292533031a2ea439 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 19:22:28 +0000
Subject: [PATCH 0939/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#6645)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3.toml           | 24 -------------------
 1 file changed, 24 deletions(-)
 delete mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
deleted file mode 100644
index a5547585252..00000000000
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-# Cloudflare Workers AI host for Z.ai GLM-5.3 (paid plan / AI Gateway credits).
-# Context window and pricing from Cloudflare model page.
-# Toggle: chat_template_kwargs.enable_thinking = true|false (default true)
-# Effort: reasoning_effort = low|medium|high
-# https://developers.cloudflare.com/workers-ai/models/glm-5.3/
-# https://developers.cloudflare.com/workers-ai/models/glm-5.3/sync-input.json
-# https://developers.cloudflare.com/ai/models/%40cf/zai-org/glm-5.3/
-base_model = "zhipuai/glm-5.3"
-name = "Glm 5.3"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
-
-[limit]
-context = 1_048_576

From 6d4bbee7609f9225850377f68d9424067c1882eb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 19:22:32 +0000
Subject: [PATCH 0940/2012] chore(sync): update NanoGPT model catalog (#6646)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/k2-horizon-7b.toml | 27 ++++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/nano-gpt/models/k2-horizon-7b.toml

diff --git a/providers/nano-gpt/models/k2-horizon-7b.toml b/providers/nano-gpt/models/k2-horizon-7b.toml
new file mode 100644
index 00000000000..ee81d369b10
--- /dev/null
+++ b/providers/nano-gpt/models/k2-horizon-7b.toml
@@ -0,0 +1,27 @@
+name = "K2-Horizon-7B"
+description = "IFM's open-weight 7B model for coding, reasoning, and chat. Heabsy serves an FP8 deployment on a dedicated RTX 5090 with speculative decoding in the United States, with a 131K context window and an 8K output limit. Zero data retention is not verified for this deployment."
+release_date = "2026-09-09"
+last_updated = "2026-09-09"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.05
+output = 0.2
+cache_read = 0.04
+
+[limit]
+context = 131_072
+input = 131_072
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 36bf217be05c8f61a951b319e94eb3871f3f2d6b Mon Sep 17 00:00:00 2001
From: nearworkerbot <lloyd.mak+bot@near.ai>
Date: Wed, 9 Sep 2026 12:58:02 -0700
Subject: [PATCH 0941/2012] sync NEAR AI Cloud catalog with the live model list
 (#6611)

Co-authored-by: nearai-catalog-sync <noreply@near.ai>
---
 .../Qwen/Qwen3-30B-A3B-Instruct-2507.toml     | 23 -------------------
 .../models/Qwen/Qwen3-Embedding-0.6B.toml     |  4 ++--
 .../Qwen/Qwen3-VL-30B-A3B-Instruct.toml       |  4 ++--
 .../nearai/models/Qwen/Qwen3.5-122B-A10B.toml | 13 -----------
 .../models/Qwen/Qwen3.6-35B-A3B-FP8.toml      |  2 +-
 .../models/anthropic/claude-sonnet-4-5.toml   |  2 +-
 .../black-forest-labs/FLUX.2-klein-4B.toml    |  2 +-
 .../nearai/models/google/gemini-3-pro.toml    |  7 ------
 .../nearai/models/google/gemma-4-31B-it.toml  | 10 --------
 providers/nearai/models/openai/gpt-5.2.toml   |  6 ++---
 .../nearai/models/openai/gpt-oss-120b.toml    | 10 --------
 .../models/openai/whisper-large-v3.toml       |  2 +-
 .../nearai/models/zai-org/GLM-5.1-FP8.toml    |  6 ++---
 13 files changed, 14 insertions(+), 77 deletions(-)
 delete mode 100644 providers/nearai/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
 delete mode 100644 providers/nearai/models/Qwen/Qwen3.5-122B-A10B.toml
 delete mode 100644 providers/nearai/models/google/gemini-3-pro.toml
 delete mode 100644 providers/nearai/models/google/gemma-4-31B-it.toml
 delete mode 100644 providers/nearai/models/openai/gpt-oss-120b.toml

diff --git a/providers/nearai/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml b/providers/nearai/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
deleted file mode 100644
index b8572bec67f..00000000000
--- a/providers/nearai/models/Qwen/Qwen3-30B-A3B-Instruct-2507.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen3 30B-A3B Instruct 2507"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-07-29"
-last_updated = "2025-07-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.15
-output = 0.55
-
-[limit]
-context = 262_144
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nearai/models/Qwen/Qwen3-Embedding-0.6B.toml b/providers/nearai/models/Qwen/Qwen3-Embedding-0.6B.toml
index 3d6dfbb0c09..bda13b5cf4e 100644
--- a/providers/nearai/models/Qwen/Qwen3-Embedding-0.6B.toml
+++ b/providers/nearai/models/Qwen/Qwen3-Embedding-0.6B.toml
@@ -11,10 +11,10 @@ open_weights = true
 
 [cost]
 input = 0.01
-output = 0.00
+output = 0.01
 
 [limit]
-context = 40_960
+context = 32768
 output = 1_024
 
 [modalities]
diff --git a/providers/nearai/models/Qwen/Qwen3-VL-30B-A3B-Instruct.toml b/providers/nearai/models/Qwen/Qwen3-VL-30B-A3B-Instruct.toml
index 1fea069c649..8503d559601 100644
--- a/providers/nearai/models/Qwen/Qwen3-VL-30B-A3B-Instruct.toml
+++ b/providers/nearai/models/Qwen/Qwen3-VL-30B-A3B-Instruct.toml
@@ -15,8 +15,8 @@ input = 0.15
 output = 0.55
 
 [limit]
-context = 256_000
-output = 32_768
+context = 16384
+output = 8192
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/nearai/models/Qwen/Qwen3.5-122B-A10B.toml b/providers/nearai/models/Qwen/Qwen3.5-122B-A10B.toml
deleted file mode 100644
index 7fc15d636fe..00000000000
--- a/providers/nearai/models/Qwen/Qwen3.5-122B-A10B.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "alibaba/qwen3.5-122b-a10b"
-reasoning_options = [{ type = "toggle" }]
-
-[cost]
-input = 0.4
-output = 3.2
-
-[limit]
-context = 131_072
-output = 32_768
-
-[modalities]
-input = ["text"]
diff --git a/providers/nearai/models/Qwen/Qwen3.6-35B-A3B-FP8.toml b/providers/nearai/models/Qwen/Qwen3.6-35B-A3B-FP8.toml
index c1da8f58418..6dd21027b99 100644
--- a/providers/nearai/models/Qwen/Qwen3.6-35B-A3B-FP8.toml
+++ b/providers/nearai/models/Qwen/Qwen3.6-35B-A3B-FP8.toml
@@ -8,7 +8,7 @@ output = 1.1
 cache_read = 0.056
 
 [limit]
-output = 32_768
+output = 8192
 
 [modalities]
 input = ["text"]
diff --git a/providers/nearai/models/anthropic/claude-sonnet-4-5.toml b/providers/nearai/models/anthropic/claude-sonnet-4-5.toml
index edc0053f1aa..f6b5beba2ca 100644
--- a/providers/nearai/models/anthropic/claude-sonnet-4-5.toml
+++ b/providers/nearai/models/anthropic/claude-sonnet-4-5.toml
@@ -3,6 +3,6 @@ reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024
 
 [cost]
 input = 3
-output = 15.5
+output = 15.0
 cache_read = 0.3
 cache_write = 3.75
diff --git a/providers/nearai/models/black-forest-labs/FLUX.2-klein-4B.toml b/providers/nearai/models/black-forest-labs/FLUX.2-klein-4B.toml
index 75fe3ebeb32..4745dd25339 100644
--- a/providers/nearai/models/black-forest-labs/FLUX.2-klein-4B.toml
+++ b/providers/nearai/models/black-forest-labs/FLUX.2-klein-4B.toml
@@ -18,5 +18,5 @@ context = 128_000
 output = 128_000
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
 output = ["image"]
diff --git a/providers/nearai/models/google/gemini-3-pro.toml b/providers/nearai/models/google/gemini-3-pro.toml
deleted file mode 100644
index c5911d6fa54..00000000000
--- a/providers/nearai/models/google/gemini-3-pro.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "google/gemini-3-pro-preview"
-reasoning_options = []
-
-[cost]
-input = 1.25
-output = 15
-cache_read = 0
diff --git a/providers/nearai/models/google/gemma-4-31B-it.toml b/providers/nearai/models/google/gemma-4-31B-it.toml
deleted file mode 100644
index a6f46532bff..00000000000
--- a/providers/nearai/models/google/gemma-4-31B-it.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "google/gemma-4-31b-it"
-reasoning_options = [{ type = "toggle" }]
-
-[cost]
-input = 0.13
-output = 0.4
-cache_read = 0.026
-
-[modalities]
-input = ["text"]
diff --git a/providers/nearai/models/openai/gpt-5.2.toml b/providers/nearai/models/openai/gpt-5.2.toml
index 3d3ced618b9..839990b6c17 100644
--- a/providers/nearai/models/openai/gpt-5.2.toml
+++ b/providers/nearai/models/openai/gpt-5.2.toml
@@ -2,6 +2,6 @@ base_model = "openai/gpt-5.2"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 
 [cost]
-input = 1.8
-output = 15.5
-cache_read = 0.18
+input = 1.75
+output = 14.0
+cache_read = 0.175
diff --git a/providers/nearai/models/openai/gpt-oss-120b.toml b/providers/nearai/models/openai/gpt-oss-120b.toml
deleted file mode 100644
index b3cd0e4651e..00000000000
--- a/providers/nearai/models/openai/gpt-oss-120b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "openai/gpt-oss-120b"
-name = "GPT-OSS 120B"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-
-[cost]
-input = 0.15
-output = 0.55
-
-[limit]
-context = 131_000
diff --git a/providers/nearai/models/openai/whisper-large-v3.toml b/providers/nearai/models/openai/whisper-large-v3.toml
index 9e77565bade..4d5dbf409ba 100644
--- a/providers/nearai/models/openai/whisper-large-v3.toml
+++ b/providers/nearai/models/openai/whisper-large-v3.toml
@@ -11,7 +11,7 @@ open_weights = true
 
 [cost]
 input = 0.01
-output = 0.00
+output = 0.01
 
 [limit]
 context = 448
diff --git a/providers/nearai/models/zai-org/GLM-5.1-FP8.toml b/providers/nearai/models/zai-org/GLM-5.1-FP8.toml
index 6ce067b5080..8043e4c0cf5 100644
--- a/providers/nearai/models/zai-org/GLM-5.1-FP8.toml
+++ b/providers/nearai/models/zai-org/GLM-5.1-FP8.toml
@@ -15,12 +15,12 @@ open_weights = true
 field = "reasoning_content"
 
 [cost]
-input = 0.85
-output = 3.30
+input = 1.4
+output = 4.4
 
 [limit]
 context = 202_752
-output = 131_072
+output = 16384
 
 [modalities]
 input = ["text"]

From 32e0f7206e470768720f53eb089bd56497c5f1b8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 20:26:21 +0000
Subject: [PATCH 0942/2012] chore(sync): update OpenRouter model catalog
 (#6648)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3.toml     | 2 +-
 providers/openrouter/models/~z-ai/glm-latest.toml | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 3f546789070..d939b2691d5 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -11,5 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 943_718
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 784c165c505..91a5845fc7b 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -20,7 +20,7 @@ output = 3.498
 cache_read = 0.2067
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 128_000
 
 [modalities]

From d53317f19c7207111eef2e59dcb62049fda75a95 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 21:22:35 +0000
Subject: [PATCH 0943/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6650)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../grok-imagine-video-1.5-preview.toml         | 17 -----------------
 1 file changed, 17 deletions(-)
 delete mode 100644 providers/vercel/models/spacexai/grok-imagine-video-1.5-preview.toml

diff --git a/providers/vercel/models/spacexai/grok-imagine-video-1.5-preview.toml b/providers/vercel/models/spacexai/grok-imagine-video-1.5-preview.toml
deleted file mode 100644
index ab2429fa128..00000000000
--- a/providers/vercel/models/spacexai/grok-imagine-video-1.5-preview.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-name = "Grok Imagine Video 1.5 Preview"
-description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "grok"
-release_date = "2026-05-30"
-last_updated = "2026-05-30"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["text"]
-output = ["video"]

From 39afe73c6d60f65983b90606313704330bf6c5c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 21:22:49 +0000
Subject: [PATCH 0944/2012] chore(sync): update Deep Infra model catalog
 (#6649)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml        | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
index 510ea23b43c..46c971a8379 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
@@ -12,7 +12,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.44
 output = 1.32
-cache_read = 0.14
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 126dcebfba60767dfb72ca9934be1e9d08369249 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 23:23:23 +0000
Subject: [PATCH 0945/2012] chore(sync): update Venice model catalog (#6652)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/openai-gpt-52.toml       | 3 +--
 providers/venice/models/openai-gpt-53-codex.toml | 3 +--
 providers/venice/models/openai-gpt-54-mini.toml  | 3 +--
 providers/venice/models/openai-gpt-54-pro.toml   | 3 +--
 providers/venice/models/openai-gpt-54.toml       | 3 +--
 5 files changed, 5 insertions(+), 10 deletions(-)

diff --git a/providers/venice/models/openai-gpt-52.toml b/providers/venice/models/openai-gpt-52.toml
index b400e076236..83a8cee2188 100644
--- a/providers/venice/models/openai-gpt-52.toml
+++ b/providers/venice/models/openai-gpt-52.toml
@@ -1,6 +1,5 @@
 # Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "openai/gpt-5.2"
 release_date = "2025-12-13"
 last_updated = "2026-06-11"
@@ -8,7 +7,7 @@ attachment = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 2.19
diff --git a/providers/venice/models/openai-gpt-53-codex.toml b/providers/venice/models/openai-gpt-53-codex.toml
index fc27c32d66d..8bbd5a49fe7 100644
--- a/providers/venice/models/openai-gpt-53-codex.toml
+++ b/providers/venice/models/openai-gpt-53-codex.toml
@@ -1,6 +1,5 @@
 # Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "openai/gpt-5.3-codex"
 family = "gpt"
 release_date = "2026-02-24"
@@ -8,7 +7,7 @@ last_updated = "2026-06-11"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 2.19
diff --git a/providers/venice/models/openai-gpt-54-mini.toml b/providers/venice/models/openai-gpt-54-mini.toml
index b0de418a6df..cbc1d598e26 100644
--- a/providers/venice/models/openai-gpt-54-mini.toml
+++ b/providers/venice/models/openai-gpt-54-mini.toml
@@ -1,6 +1,5 @@
 # Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "openai/gpt-5.4-mini"
 name = "GPT-5.4 Mini"
 family = "gpt"
@@ -9,7 +8,7 @@ last_updated = "2026-06-11"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 0.9375
diff --git a/providers/venice/models/openai-gpt-54-pro.toml b/providers/venice/models/openai-gpt-54-pro.toml
index 46b420c7b1f..75b2af050fa 100644
--- a/providers/venice/models/openai-gpt-54-pro.toml
+++ b/providers/venice/models/openai-gpt-54-pro.toml
@@ -1,6 +1,5 @@
 # Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "openai/gpt-5.4-pro"
 family = "gpt"
 last_updated = "2026-06-11"
@@ -8,7 +7,7 @@ structured_output = true
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
+values = ["none", "medium", "high", "xhigh"]
 
 [cost]
 input = 37.5
diff --git a/providers/venice/models/openai-gpt-54.toml b/providers/venice/models/openai-gpt-54.toml
index e261036f262..6d7e63e7e6f 100644
--- a/providers/venice/models/openai-gpt-54.toml
+++ b/providers/venice/models/openai-gpt-54.toml
@@ -1,12 +1,11 @@
 # Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
 base_model = "openai/gpt-5.4"
 last_updated = "2026-06-11"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 3.13

From 411b4c214fadb932bfaa78cf906935564e99c4bb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 23:23:26 +0000
Subject: [PATCH 0946/2012] chore(sync): update OpenRouter model catalog
 (#6653)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml     | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index d5918c075a2..4dac1f2fc80 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index c32fca67d4b..644122b25af 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_310_720

From 5dc2ef24dc14a5820e8fb2423134035e11d07576 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 23:23:31 +0000
Subject: [PATCH 0947/2012] chore(sync): update Kilo model catalog (#6654)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index e4ce4e5dfe6..5d5fb4fce66 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From b71d0251eb95f6d95a20d9ed546b481541a07c26 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 00:58:09 +0000
Subject: [PATCH 0948/2012] chore(sync): update Venice model catalog (#6655)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/mercury-2-5.toml    |  6 +++---
 providers/venice/models/qwen-3-8-flash.toml | 10 ++++++++++
 2 files changed, 13 insertions(+), 3 deletions(-)
 create mode 100644 providers/venice/models/qwen-3-8-flash.toml

diff --git a/providers/venice/models/mercury-2-5.toml b/providers/venice/models/mercury-2-5.toml
index a61068676a6..efa97bb08f8 100644
--- a/providers/venice/models/mercury-2-5.toml
+++ b/providers/venice/models/mercury-2-5.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.04999999999999999
-output = 0.18749999999999994
-cache_read = 0.004999999999999999
+input = 0.25
+output = 0.9375
+cache_read = 0.025
 
 [limit]
 context = 260_000
diff --git a/providers/venice/models/qwen-3-8-flash.toml b/providers/venice/models/qwen-3-8-flash.toml
new file mode 100644
index 00000000000..d5e7b60805f
--- /dev/null
+++ b/providers/venice/models/qwen-3-8-flash.toml
@@ -0,0 +1,10 @@
+base_model = "alibaba/qwen3.8-flash"
+name = "Qwen 3.8 Flash"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+reasoning_options = []
+
+[cost]
+input = 0.14
+output = 0.49
+cache_read = 0.014

From 94dfb18dbddb1b7d977757d0fa5b8d444493c1e7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 00:58:14 +0000
Subject: [PATCH 0949/2012] chore(sync): update Kilo model catalog (#6657)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat.toml         | 8 ++++----
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 3 +--
 providers/kilo/models/minimax/minimax-m2.5.toml           | 4 ----
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml    | 2 +-
 .../kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml     | 4 ++--
 providers/kilo/models/qwen/qwen3.5-122b-a10b.toml         | 3 ---
 providers/kilo/models/tencent/hy3.toml                    | 6 +++---
 providers/kilo/models/thinkingmachines/inkling.toml       | 3 +--
 providers/kilo/models/z-ai/glm-5.2.toml                   | 3 +--
 9 files changed, 13 insertions(+), 23 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat.toml b/providers/kilo/models/deepseek/deepseek-chat.toml
index 42fda83c47b..f67bbe80e1f 100644
--- a/providers/kilo/models/deepseek/deepseek-chat.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat.toml
@@ -4,9 +4,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.32
-output = 0.89
+input = 0.2574
+output = 1.0287
 
 [limit]
-context = 163_840
-output = 16_384
+context = 128_000
+output = 16_000
diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index e89b49a7fbe..3f15c0bc21d 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,7 +10,3 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
-
-[limit]
-context = 200_000
-output = 128_000
diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 3e83e7378b1..75e607a6e02 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352
diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 48baf441187..1da6f655c94 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 128_000
-output = 32_000
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml b/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
index 375dea8cc3c..3a1fc219280 100644
--- a/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-122b-a10b.toml
@@ -9,8 +9,5 @@ values = ["none", "high"]
 input = 0.26
 output = 2.08
 
-[limit]
-output = 81_920
-
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/thinkingmachines/inkling.toml b/providers/kilo/models/thinkingmachines/inkling.toml
index a367cecf27a..79cba59e119 100644
--- a/providers/kilo/models/thinkingmachines/inkling.toml
+++ b/providers/kilo/models/thinkingmachines/inkling.toml
@@ -12,5 +12,4 @@ output = 4.05
 cache_read = 0.16
 
 [limit]
-context = 524_288
-output = 471_859
+output = 32_768
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index dcb3cc80d7f..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,5 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 1_048_576

From f85e7ba42c37722db424a1b39b62eb99102eb16f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 00:58:24 +0000
Subject: [PATCH 0950/2012] chore(sync): update DigitalOcean model catalog
 (#6656)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/kimi-k3.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/digitalocean/models/kimi-k3.toml b/providers/digitalocean/models/kimi-k3.toml
index b71a3e29e2e..7533a7423e2 100644
--- a/providers/digitalocean/models/kimi-k3.toml
+++ b/providers/digitalocean/models/kimi-k3.toml
@@ -11,8 +11,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.85
-output = 14.25
+input = 2.55
+output = 12.95
 cache_read = 0.285
 
 [modalities]

From 887ee5d219d03a8ab50251a54bf097a7aea4d07b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 00:58:53 +0000
Subject: [PATCH 0951/2012] chore(sync): update OpenRouter model catalog
 (#6658)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-chat.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 7 +++----
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 providers/openrouter/models/minimax/minimax-m2.5.toml    | 9 +++------
 .../openrouter/models/moonshotai/kimi-k2-thinking.toml   | 3 ++-
 .../models/qwen/qwen3-30b-a3b-instruct-2507.toml         | 6 +++---
 providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml  | 7 ++-----
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 .../openrouter/models/thinkingmachines/inkling.toml      | 2 +-
 providers/openrouter/models/z-ai/glm-5.2.toml            | 7 +++----
 11 files changed, 29 insertions(+), 36 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat.toml b/providers/openrouter/models/deepseek/deepseek-chat.toml
index 7768aadee98..a01fc1ca488 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat.toml
@@ -3,9 +3,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.32
-output = 0.89
+input = 0.2574
+output = 1.0287
 
 [limit]
 context = 163_840
-output = 16_384
+output = 16_000
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3eee6643f71..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07
-output = 0.14
-cache_read = 0.014
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..5bc4e6af877 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 1.0494
+output = 3.1482
+cache_read = 0.03498
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a010dd38ba2..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.0725
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index ca75f8e4927..1edeb5ecdfc 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,9 +6,6 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.27
-output = 1.08
-cache_read = 0.027
-
-[limit]
-output = 128_000
+input = 0.3
+output = 1.2
+cache_read = 0.03
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 3ccd5a4f3ef..91ad22ed77f 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,6 +8,7 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
+cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352
diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 2ce7e2e4c07..c0c2c8b5cd7 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.04815
-output = 0.19305
+input = 0.09
+output = 0.3
 
 [limit]
 context = 262_144
-output = 32_000
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml b/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
index c07c101174e..cf242b0a0f9 100644
--- a/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml
@@ -6,11 +6,8 @@ base_model = "alibaba/qwen3.5-122b-a10b"
 type = "toggle"
 
 [cost]
-input = 0.29
-output = 2.4
-
-[limit]
-output = 81_920
+input = 0.26
+output = 2.08
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/thinkingmachines/inkling.toml b/providers/openrouter/models/thinkingmachines/inkling.toml
index 1a64ecd26f3..bbed1c7e0bc 100644
--- a/providers/openrouter/models/thinkingmachines/inkling.toml
+++ b/providers/openrouter/models/thinkingmachines/inkling.toml
@@ -12,4 +12,4 @@ output = 4.05
 cache_read = 0.17
 
 [limit]
-output = 471_859
+output = 32_768
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index c38d06275a4..64652242a0e 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.28
-output = 0.88
-cache_read = 0.052
+input = 0.966
+output = 3.036
+cache_read = 0.1932
 
 [limit]
 context = 1_048_576
-output = 128_000

From 1b822856711ca6251a769c32d77e4b76b4ee1310 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 9 Sep 2026 20:24:26 -0500
Subject: [PATCH 0952/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#6647)

* chore(sync): update Cloudflare Workers AI model catalog

* fix(cloudflare-workers-ai): add GLM 5.3 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/@cf/zai-org/glm-5.3.toml                 | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml

diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
new file mode 100644
index 00000000000..a5f20bbaf80
--- /dev/null
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
@@ -0,0 +1,13 @@
+base_model = "zhipuai/glm-5.3"
+name = "Glm 5.3"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_310_720
+output = 1_310_720

From d87b965fd0182aeed7c7b998861e04fbc8080a49 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 01:33:21 +0000
Subject: [PATCH 0953/2012] chore(sync): update NanoGPT model catalog (#6660)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From cc066e352bbf924bf14c4ca9368dd5215bb783fd Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 10 Sep 2026 11:27:04 +0800
Subject: [PATCH 0954/2012] feat(go): add deepseek v4.1 flash

---
 models/deepseek/deepseek-v4.1-flash.toml      | 21 +++++++++++++++++++
 .../opencode-go/models/deepseek-flash.toml    | 14 +++++++++++++
 .../models/deepseek-v4-flash-vision-exp.toml  |  6 +++---
 .../opencode-go/models/deepseek-v4-flash.toml |  6 +++---
 4 files changed, 41 insertions(+), 6 deletions(-)
 create mode 100644 models/deepseek/deepseek-v4.1-flash.toml
 create mode 100644 providers/opencode-go/models/deepseek-flash.toml

diff --git a/models/deepseek/deepseek-v4.1-flash.toml b/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..aac72588623
--- /dev/null
+++ b/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,21 @@
+name = "DeepSeek V4.1 Flash"
+description = "DeepSeek V4.1 Flash model for reasoning and agentic coding"
+family = "deepseek-flash"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-05"
+open_weights = true
+license = "MIT"
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
\ No newline at end of file
diff --git a/providers/opencode-go/models/deepseek-flash.toml b/providers/opencode-go/models/deepseek-flash.toml
new file mode 100644
index 00000000000..21a954201c4
--- /dev/null
+++ b/providers/opencode-go/models/deepseek-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
diff --git a/providers/opencode-go/models/deepseek-v4-flash-vision-exp.toml b/providers/opencode-go/models/deepseek-v4-flash-vision-exp.toml
index 1f68e0a4dd5..de410add742 100644
--- a/providers/opencode-go/models/deepseek-v4-flash-vision-exp.toml
+++ b/providers/opencode-go/models/deepseek-v4-flash-vision-exp.toml
@@ -8,6 +8,6 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "h
 field = "reasoning_content"
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.15
+output = 0.6
+cache_read = 0.003
diff --git a/providers/opencode-go/models/deepseek-v4-flash.toml b/providers/opencode-go/models/deepseek-v4-flash.toml
index 31fdd5dc3d5..09d83c0a6db 100644
--- a/providers/opencode-go/models/deepseek-v4-flash.toml
+++ b/providers/opencode-go/models/deepseek-v4-flash.toml
@@ -11,6 +11,6 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.15
+output = 0.6
+cache_read = 0.003

From b04413e2dfbe29d0b7e7c29ee500136da3bb23ad Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 03:30:33 +0000
Subject: [PATCH 0955/2012] chore(sync): update OpenRouter model catalog
 (#6666)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 91a5845fc7b..7ee4d88f9e0 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.113
-output = 3.498
-cache_read = 0.2067
+input = 1.092
+output = 3.432
+cache_read = 0.2028
 
 [limit]
 context = 1_310_720
-output = 128_000
+output = 131_072
 
 [modalities]
 input = ["text"]

From 611a8f8b66961df0092af682f537a1380c753874 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 03:30:36 +0000
Subject: [PATCH 0956/2012] chore(sync): update Kilo model catalog (#6665)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 81909334ecc..64165466088 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.113
-output = 3.498
-cache_read = 0.2067
+input = 1.092
+output = 3.432
+cache_read = 0.2028
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text"]

From 924a7c3746e722821757e2cc35f30b5ef22d432e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 03:30:54 +0000
Subject: [PATCH 0957/2012] chore(sync): update NanoGPT model catalog (#6664)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml    | 17 +----------------
 .../deepseek/deepseek-v4.1-flash:thinking.toml  | 16 +---------------
 2 files changed, 2 insertions(+), 31 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 40304b5bd8e..52eca003f08 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -1,13 +1,4 @@
-name = "DeepSeek V4.1 Flash"
-description = "DeepSeek V4.1 Flash supports text and image input, reasoning, tool calling, and structured output with a 1M-token context window. This is a rate-limited beta with limited capacity, intended for testing rather than production use. Assume prompts and responses are logged by the provider and may be used for model training or service improvement. Do not send sensitive or confidential data."
-family = "deepseek"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
+base_model = "deepseek/deepseek-v4.1-flash"
 
 [[reasoning_options]]
 type = "effort"
@@ -19,10 +10,4 @@ output = 0.3124
 cache_read = 0.03124
 
 [limit]
-context = 1_000_000
 input = 1_000_000
-output = 384_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 8bf595069bc..bd30c42f5c9 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -1,13 +1,5 @@
+base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4.1 Flash Thinking"
-description = "DeepSeek V4.1 Flash supports text and image input, reasoning, tool calling, and structured output with a 1M-token context window. This is a rate-limited beta with limited capacity, intended for testing rather than production use. Assume prompts and responses are logged by the provider and may be used for model training or service improvement. Do not send sensitive or confidential data."
-family = "deepseek"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,10 +11,4 @@ output = 0.3124
 cache_read = 0.03124
 
 [limit]
-context = 1_000_000
 input = 1_000_000
-output = 384_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From c24a64b589a593f7fef863273c0b7b606f9731ee Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Thu, 10 Sep 2026 12:09:08 +0800
Subject: [PATCH 0958/2012] feat (go): deprecate omen alpha

---
 providers/opencode-go/models/omen-alpha.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/opencode-go/models/omen-alpha.toml b/providers/opencode-go/models/omen-alpha.toml
index dd02fb4d829..140090f751a 100644
--- a/providers/opencode-go/models/omen-alpha.toml
+++ b/providers/opencode-go/models/omen-alpha.toml
@@ -9,6 +9,7 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = false
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"

From ad5311c33202f1a98e0977cc3bcef3d702b1e4d0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 04:30:35 +0000
Subject: [PATCH 0959/2012] chore(sync): update NanoGPT model catalog (#6669)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 4c483b8422cfca8a3f4a7364315617b966c65ca5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 04:30:40 +0000
Subject: [PATCH 0960/2012] chore(sync): update Kilo model catalog (#6668)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.2.toml           | 3 ---
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 6 +++---
 2 files changed, 3 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bb89edf75cf..521ea40a372 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -9,6 +9,3 @@ values = ["none", "high", "xhigh"]
 input = 1.4
 output = 4.4
 cache_read = 0.26
-
-[limit]
-context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 5d5fb4fce66..e4ce4e5dfe6 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 31982b16662b108b46e96a8a45474add7b0695b3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 04:30:42 +0000
Subject: [PATCH 0961/2012] chore(sync): update OpenRouter model catalog
 (#6667)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/qwen/qwen3.8-27b.toml       | 2 +-
 providers/openrouter/models/z-ai/glm-5.3-flash.toml     | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index d1ddcffa2ed..da9f1b34dbf 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -16,5 +16,5 @@ output = 3
 cache_read = 0.085
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 131_072
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index 4dac1f2fc80..d5918c075a2 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 644122b25af..c32fca67d4b 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720

From 4faf5c7ac4a2fa4f0ad12550b899afcd12c2d857 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 05:26:27 +0000
Subject: [PATCH 0962/2012] chore(sync): update Eden AI model catalog (#6671)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nebius/deepseek-ai/DeepSeek-V4-Pro-0813.toml   | 14 ++++++++++++++
 .../edenai/models/qwen/deepseek-v4-flash-0731.toml |  4 ++--
 .../edenai/models/qwen/deepseek-v4-pro-0813.toml   |  4 ++--
 3 files changed, 18 insertions(+), 4 deletions(-)
 create mode 100644 providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Pro-0813.toml

diff --git a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Pro-0813.toml
new file mode 100644
index 00000000000..ff086c1598a
--- /dev/null
+++ b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek V4 Pro 0813 (Nebius)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 1.32
+output = 3.96
+cache_read = 1.32
+
+[limit]
+context = 979_000
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From 16a80fd88d90c1d54904b5cf2a237c74d635f800 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 06:40:18 +0000
Subject: [PATCH 0963/2012] chore(sync): update NanoGPT model catalog (#6677)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 22035ca9dc0daca2ec1d32532e05a3c196131ee8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 06:40:33 +0000
Subject: [PATCH 0964/2012] chore(sync): update OpenRouter model catalog
 (#6676)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml  | 20 +++++++++++++++++++
 .../openrouter/models/qwen/qwen3.8-27b.toml   |  2 +-
 2 files changed, 21 insertions(+), 1 deletion(-)
 create mode 100644 providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..90ff6cfd1ae
--- /dev/null
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,20 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "deepseek/deepseek-v4.1-flash"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+structured_output = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index da9f1b34dbf..d1ddcffa2ed 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -16,5 +16,5 @@ output = 3
 cache_read = 0.085
 
 [limit]
-context = 1_048_576
+context = 1_000_000
 output = 131_072

From 9621763e96d47a50d71d90009986b747d533d840 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 06:40:41 +0000
Subject: [PATCH 0965/2012] chore(sync): update Venice model catalog (#6678)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/mercury-2-5.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/venice/models/mercury-2-5.toml b/providers/venice/models/mercury-2-5.toml
index efa97bb08f8..a61068676a6 100644
--- a/providers/venice/models/mercury-2-5.toml
+++ b/providers/venice/models/mercury-2-5.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 0.9375
-cache_read = 0.025
+input = 0.04999999999999999
+output = 0.18749999999999994
+cache_read = 0.004999999999999999
 
 [limit]
 context = 260_000

From b3149eb01b25776853165dd5c7e60a89f95feade Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 07:28:50 +0000
Subject: [PATCH 0966/2012] chore(sync): update CrossModel model catalog
 (#6681)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-vision-exp.toml     |  8 ++++----
 .../models/deepseek/deepseek-v4-flash.toml         |  8 ++++----
 .../models/deepseek/deepseek-v4.1-flash.toml       | 14 ++++++++++++++
 3 files changed, 22 insertions(+), 8 deletions(-)
 create mode 100644 providers/crossmodel/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/crossmodel/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/crossmodel/models/deepseek/deepseek-v4-flash-vision-exp.toml
index df9e322eaf6..cae79ce1a08 100644
--- a/providers/crossmodel/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/crossmodel/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -8,7 +8,7 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.405
-output = 1.215
-cache_read = 0.0135
-cache_write = 0.405
+input = 0.27
+output = 1.08
+cache_read = 0.0054
+cache_write = 0.27
diff --git a/providers/crossmodel/models/deepseek/deepseek-v4-flash.toml b/providers/crossmodel/models/deepseek/deepseek-v4-flash.toml
index a87de3eec1d..85e579d997c 100644
--- a/providers/crossmodel/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/crossmodel/models/deepseek/deepseek-v4-flash.toml
@@ -8,7 +8,7 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.405
-output = 1.215
-cache_read = 0.0135
-cache_write = 0.405
+input = 0.27
+output = 1.08
+cache_read = 0.0054
+cache_write = 0.27
diff --git a/providers/crossmodel/models/deepseek/deepseek-v4.1-flash.toml b/providers/crossmodel/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..32428719c16
--- /dev/null
+++ b/providers/crossmodel/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.27
+output = 1.08
+cache_read = 0.0054
+cache_write = 0.27

From b58b83aff9e28c5241d8f98a2b7cab6db03129fb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 07:28:55 +0000
Subject: [PATCH 0967/2012] chore(sync): update Kilo model catalog (#6682)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4.1-flash.toml | 15 +++++++++++++++
 providers/kilo/models/~z-ai/glm-latest.toml       | 10 +++++-----
 2 files changed, 20 insertions(+), 5 deletions(-)
 create mode 100644 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..92364b2c945
--- /dev/null
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+description = "DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the cost-efficient tier of the V4.1 family. DeepSeek reports that it exceeds V4 Pro on performance, speed, and task..."
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 64165466088..b71febab31d 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.092
-output = 3.432
-cache_read = 0.2028
+input = 1.085
+output = 3.41
+cache_read = 0.2015
 
 [limit]
-context = 1_048_576
-output = 131_072
+context = 1_024_000
+output = 128_000
 
 [modalities]
 input = ["text"]

From 22312e7502820e974d045f97b64c496e65f14260 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 07:29:03 +0000
Subject: [PATCH 0968/2012] chore(sync): update Ofox model catalog (#6683)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ofox/models/z-ai/glm-5.2.toml       | 6 +++---
 providers/ofox/models/z-ai/glm-5.3-flash.toml | 6 +++---
 providers/ofox/models/z-ai/glm-5.3.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/ofox/models/z-ai/glm-5.2.toml b/providers/ofox/models/z-ai/glm-5.2.toml
index eadde6f74bc..fe20830a575 100644
--- a/providers/ofox/models/z-ai/glm-5.2.toml
+++ b/providers/ofox/models/z-ai/glm-5.2.toml
@@ -5,6 +5,6 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 0.98
-output = 3.08
-cache_read = 0.182
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/ofox/models/z-ai/glm-5.3-flash.toml b/providers/ofox/models/z-ai/glm-5.3-flash.toml
index ead3405dc30..74eb807ba1b 100644
--- a/providers/ofox/models/z-ai/glm-5.3-flash.toml
+++ b/providers/ofox/models/z-ai/glm-5.3-flash.toml
@@ -10,6 +10,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
diff --git a/providers/ofox/models/z-ai/glm-5.3.toml b/providers/ofox/models/z-ai/glm-5.3.toml
index 6cb486df150..c6ad1a40387 100644
--- a/providers/ofox/models/z-ai/glm-5.3.toml
+++ b/providers/ofox/models/z-ai/glm-5.3.toml
@@ -11,6 +11,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.26
-output = 3.96
-cache_read = 0.234
+input = 1.4
+output = 4.4
+cache_read = 0.26

From 69bbc21a6de5d47144d9e0991b2859450940a092 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 07:29:35 +0000
Subject: [PATCH 0969/2012] chore(sync): update OpenRouter model catalog
 (#6684)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 7ee4d88f9e0..2a98306a5b6 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.092
-output = 3.432
-cache_read = 0.2028
+input = 1.085
+output = 3.41
+cache_read = 0.2015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 128_000
 
 [modalities]
 input = ["text"]

From 4f583e43c1e08273879560ae445122b9690d0202 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 08:34:00 +0000
Subject: [PATCH 0970/2012] chore(sync): update LLM Gateway model catalog
 (#6686)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml     | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/deepseek/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..28944b9ff2c
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (DeepSeek)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
+
+[limit]
+context = 1_050_000
+output = 393_216

From f906ef24635e926e50f5ed94eda2b780332eb9c6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 08:34:03 +0000
Subject: [PATCH 0971/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6687)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway/models/deepseek-v4.1-flash.toml      | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/llmgateway/models/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway/models/deepseek-v4.1-flash.toml b/providers/llmgateway/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..9a73dde4fdf
--- /dev/null
+++ b/providers/llmgateway/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,13 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
+
+[limit]
+context = 1_050_000

From c58e2443918234bd1fc24dd4d2acab8ecf0c75ea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 08:34:46 +0000
Subject: [PATCH 0972/2012] chore(sync): update NanoGPT model catalog (#6689)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/ornith-ai/ornith-1.5-9b.toml       | 25 -------------------
 .../ornith-ai/ornith-1.5-9b:thinking.toml     | 25 -------------------
 .../qwen/qwen3.6-35b-a3b-uncensored.toml      | 25 -------------------
 .../qwen3.6-35b-a3b-uncensored:thinking.toml  | 25 -------------------
 4 files changed, 100 deletions(-)
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
 delete mode 100644 providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml

diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
deleted file mode 100644
index 5d0b5ebe016..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 9B"
-description = "Ornith 1.5 9B is an FP8 dense open-weight model built for agentic coding, tool use, visual understanding, and efficient long-context work. This variant disables thinking for faster direct responses."
-family = "ornith"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml b/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
deleted file mode 100644
index ebbf9c7150b..00000000000
--- a/providers/nano-gpt/models/ornith-ai/ornith-1.5-9b:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Ornith 1.5 9B Thinking"
-description = "Ornith 1.5 9B with thinking enabled for more deliberate agentic coding, tool use, visual understanding, and long-context work."
-family = "ornith"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
deleted file mode 100644
index bbb44da495d..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen 3.6 35B A3B Uncensored"
-description = "Qwen 3.6 35B A3B Uncensored is an NVFP4 open-weight mixture-of-experts model LoRA-tuned for fewer refusals across chat, coding, tool use, and multimodal tasks."
-family = "qwen3.6"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.95
-cache_read = 0.075
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
deleted file mode 100644
index 0cbe6601a36..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b-uncensored:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen 3.6 35B A3B Uncensored Thinking"
-description = "Qwen 3.6 35B A3B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and complex chat tasks."
-family = "qwen3.6"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.95
-cache_read = 0.075
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 695ddc952dac726de952592b5673d162b0713c93 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 09:28:49 +0000
Subject: [PATCH 0973/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6692)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-flash-vision-exp.toml        | 13 -------------
 1 file changed, 13 deletions(-)
 delete mode 100644 providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml b/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml
deleted file mode 100644
index e6f23a8f258..00000000000
--- a/providers/llmgateway/models/deepseek-v4-flash-vision-exp.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-vision-exp"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.0028
-
-[limit]
-context = 1_050_000

From 766e4accbb51d48e35ac6d1d64763f5ca456cd29 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 09:28:58 +0000
Subject: [PATCH 0974/2012] chore(sync): update LLM Gateway model catalog
 (#6693)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-v4-flash-vision-exp.toml         | 16 ----------------
 .../models/deepseek/deepseek-v4-flash.toml    | 19 -------------------
 2 files changed, 35 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml
 delete mode 100644 providers/llmgateway-providers/models/deepseek/deepseek-v4-flash.toml

diff --git a/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml
deleted file mode 100644
index 0b1fac9b6a9..00000000000
--- a/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-vision-exp"
-name = "DeepSeek V4 Flash Vision Exp (DeepSeek)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.0028
-
-[limit]
-context = 1_050_000
-output = 393_216
diff --git a/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash.toml
deleted file mode 100644
index 9f461ad4489..00000000000
--- a/providers/llmgateway-providers/models/deepseek/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash"
-name = "DeepSeek V4 Flash (DeepSeek)"
-structured_output = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.0028
-
-[limit]
-context = 1_050_000
-output = 393_216

From 8168b0766a61b78e6cb7c0f94235a72cfc7ee222 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 10:26:50 +0000
Subject: [PATCH 0975/2012] chore(sync): update OpenRouter model catalog
 (#6698)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml    | 7 +++----
 providers/openrouter/models/z-ai/glm-5.3-flash.toml        | 6 +++---
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 90ff6cfd1ae..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -2,7 +2,6 @@
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "deepseek/deepseek-v4.1-flash"
 description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-structured_output = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index d5918c075a2..c54d27b9d18 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_310_720

From 2ac7755b0943862f99b6dcd144f14548952c7dcd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 10:26:57 +0000
Subject: [PATCH 0976/2012] chore(sync): update NanoGPT model catalog (#6697)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/gemma-4-12b-it-semancer.toml      | 4 ++--
 .../nano-gpt/models/gemma-4-12b-it-station-keeper.toml      | 4 ++--
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
index a0e66888780..0a5b660c5d9 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
@@ -16,8 +16,8 @@ output = 0.25
 cache_read = 0.025
 
 [limit]
-context = 131_072
-input = 131_072
+context = 262_144
+input = 262_144
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
index c4ccf5f509a..a68502a86c6 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
@@ -16,8 +16,8 @@ output = 0.25
 cache_read = 0.025
 
 [limit]
-context = 131_072
-input = 131_072
+context = 262_144
+input = 262_144
 output = 32_768
 
 [modalities]

From 0fb6f2452653d7f6bb0fe9c65f7b51e6d435d39e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 10:27:27 +0000
Subject: [PATCH 0977/2012] chore(sync): update Kilo model catalog (#6696)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index 92364b2c945..5431caf43db 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -1,15 +1,14 @@
 base_model = "deepseek/deepseek-v4.1-flash"
 description = "DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the cost-efficient tier of the V4.1 family. DeepSeek reports that it exceeds V4 Pro on performance, speed, and task..."
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576

From 0d4ad1581664b282fecde9b060424a0488a1fb8e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:24:29 +0000
Subject: [PATCH 0978/2012] chore(sync): update NanoGPT model catalog (#6701)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/alibaba/qwen3.8-flash.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml b/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
index 6b08ac34660..c04f98ae19f 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
+++ b/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
@@ -5,8 +5,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.16
-output = 0.47
+input = 0.14
+output = 0.42
 cache_read = 0.016
 cache_write = 0.2
 

From 5fbbd797c30391c92b7f787842aa24ab6fb5b2f5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:24:56 +0000
Subject: [PATCH 0979/2012] chore(sync): update Kilo model catalog (#6702)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index 5431caf43db..2d43303aad0 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576

From 7e02ee8b9c5a5c3d7d866c5b8cb262e6e6aa2306 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 12:37:23 +0000
Subject: [PATCH 0980/2012] chore(sync): update OpenRouter model catalog
 (#6704)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index c0c2c8b5cd7..2ce7e2e4c07 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.3
+input = 0.04815
+output = 0.19305
 
 [limit]
 context = 262_144
-output = 235_929
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 2a98306a5b6..80a0d4a1f36 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.085
+input = 1.0075
 output = 3.41
 cache_read = 0.2015
 
 [limit]
 context = 1_310_720
-output = 128_000
+output = 235_929
 
 [modalities]
 input = ["text"]

From 2ae13086d0132a9c5517020c03243622b9309dff Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 12:37:39 +0000
Subject: [PATCH 0981/2012] chore(sync): update NanoGPT model catalog (#6706)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 52eca003f08..9c2641248c0 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1562
-output = 0.3124
-cache_read = 0.03124
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index bd30c42f5c9..2e7355d5588 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1562
-output = 0.3124
-cache_read = 0.03124
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From eb5642dfb1f361d0f3d9a759cb61509972d5cae8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 12:37:50 +0000
Subject: [PATCH 0982/2012] chore(sync): update Kilo model catalog (#6705)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 4 ++--
 providers/kilo/models/~z-ai/glm-latest.toml                 | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 1da6f655c94..48baf441187 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 262_144
-output = 235_929
+context = 128_000
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index b71febab31d..c87df1c21dd 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.085
+input = 1.0075
 output = 3.41
 cache_read = 0.2015
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 157a809c99d7f34a73b7a545dd187a7c3b5ff920 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 12:37:55 +0000
Subject: [PATCH 0983/2012] chore(sync): update Venice model catalog (#6707)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/deepseek-v4-1-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/venice/models/deepseek-v4-1-flash.toml

diff --git a/providers/venice/models/deepseek-v4-1-flash.toml b/providers/venice/models/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..2e04252851f
--- /dev/null
+++ b/providers/venice/models/deepseek-v4-1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.375
+output = 1.5
+cache_read = 0.0075
+
+[limit]
+output = 131_072

From 445c6a60511b9d85df1ef18414290ac8af1734af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:25:05 +0000
Subject: [PATCH 0984/2012] chore(sync): update Merge Gateway model catalog
 (#6710)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml        | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/merge-gateway/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4.1-flash.toml b/providers/merge-gateway/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..1e8fa5e5686
--- /dev/null
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,13 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003

From 00c80365227fedcc1448a6286d2b7683e82404ed Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:25:11 +0000
Subject: [PATCH 0985/2012] chore(sync): update Kilo model catalog (#6712)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.2.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index 521ea40a372..f8addeabfc8 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -9,3 +9,7 @@ values = ["none", "high", "xhigh"]
 input = 1.4
 output = 4.4
 cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 163_840

From f197e8b266af66163b79cbc39a4584f7f044066d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:25:36 +0000
Subject: [PATCH 0986/2012] chore(sync): update OpenRouter model catalog
 (#6711)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-4.7-flash.toml | 1 -
 providers/openrouter/models/z-ai/glm-5.2.toml       | 7 ++++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-4.7-flash.toml b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
index 1745eeaa253..f2fda12e7c2 100644
--- a/providers/openrouter/models/z-ai/glm-4.7-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-4.7-flash.toml
@@ -14,5 +14,4 @@ input = 0.0605
 output = 0.4
 
 [limit]
-context = 202_752
 output = 117_964
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 64652242a0e..95694d19041 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.966
-output = 3.036
-cache_read = 0.1932
+input = 0.4875
+output = 1.56
+cache_read = 0.091
 
 [limit]
 context = 1_048_576
+output = 163_840

From 07cfbacaebf80b82388036db64d4e48c4d3f8c51 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:26:11 +0000
Subject: [PATCH 0987/2012] chore(sync): update NanoGPT model catalog (#6709)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/deepseek/deepseek-v4-flash-0731.toml    | 6 +++---
 .../models/deepseek/deepseek-v4-flash-0731:thinking.toml    | 6 +++---
 .../nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml  | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731.toml
index 07432312cae..86743682fe7 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.014
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731:thinking.toml
index 0b5ea0e7dc3..ad56f0f83d1 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-0731:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.014
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
index ecb9c3ce5f2..abb7e798efd 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.28
-cache_read = 0.014
+input = 0.05
+output = 0.16
+cache_read = 0.013
 
 [limit]
 context = 1_048_576

From 2a9c4aa261040adac59405296eadb4dffac84cdf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 14:26:47 +0000
Subject: [PATCH 0988/2012] chore(sync): update OpenRouter model catalog
 (#6715)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 8 ++++----
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5bc4e6af877..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0494
-output = 3.1482
-cache_read = 0.03498
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 95694d19041..c38d06275a4 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,10 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.4875
-output = 1.56
-cache_read = 0.091
+input = 0.28
+output = 0.88
+cache_read = 0.052
 
 [limit]
 context = 1_048_576
-output = 163_840
+output = 128_000

From a3e823ffed7cdff132836aa4423d74d08a0b880f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 14:26:57 +0000
Subject: [PATCH 0989/2012] chore(sync): update Eden AI model catalog (#6713)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-ai/DeepSeek-V4.1-Flash.toml         | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..b282fefd5db
--- /dev/null
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Deep Infra)"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+context = 1_048_576

From 7ad19fd54a56e6eefdf76e427adefd1025d8a247 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 14:26:58 +0000
Subject: [PATCH 0990/2012] chore(sync): update NanoGPT model catalog (#6714)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 ...epseek-v4-flash-vision-exp-uncensored.toml | 25 -------------------
 1 file changed, 25 deletions(-)
 delete mode 100644 providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml
deleted file mode 100644
index 4e196584f84..00000000000
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp-uncensored.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "DeepSeek V4 Flash Vision Exp Uncensored"
-description = "An uncensored variant of the experimental vision-enabled DeepSeek V4 Flash model for chat, image understanding, reasoning, coding, and tool use, with a 524K context window."
-family = "deepseek"
-release_date = "2026-09-07"
-last_updated = "2026-09-07"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = false
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.44
-output = 1.6
-cache_read = 0.22
-
-[limit]
-context = 524_288
-input = 524_288
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 6f1a986cf6f7f81b413116b366b5c0e59d647f11 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 14:27:06 +0000
Subject: [PATCH 0991/2012] chore(sync): update Kilo model catalog (#6716)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 ++-
 providers/kilo/models/z-ai/glm-5.2.toml                  | 4 ++--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index f8addeabfc8..dcb3cc80d7f 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,5 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
-output = 163_840
+context = 1_024_000
+output = 128_000

From d826e394ff2aba0a8a12165f052f3608d98904bf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 15:27:30 +0000
Subject: [PATCH 0992/2012] chore(sync): update Kilo model catalog (#6717)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index c87df1c21dd..114538a3af7 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0075
-output = 3.41
-cache_read = 0.2015
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_024_000
+output = 128_000
 
 [modalities]
 input = ["text"]

From b1afc39ea72c98efac9122113c897f1b8e157835 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 15:27:37 +0000
Subject: [PATCH 0993/2012] chore(sync): update Eden AI model catalog (#6718)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml     | 4 +---
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 8 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
index b282fefd5db..3a219aa35fb 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -1,7 +1,5 @@
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4.1 Flash (Deep Infra)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -10,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.3
 output = 1.2
-cache_read = 0.06
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 74cfaa35ff4..646d2146fb0 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -5,5 +5,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75738
-output = 0.75738
+input = 0.75504
+output = 0.75504
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index f6d21095100..32722586cef 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17478
-output = 0.75738
+input = 0.17424
+output = 0.75504
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index bb5dfdbfa65..eb910dd9b4a 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46608
-output = 0.93216
+input = 0.46464
+output = 0.92928
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 6090daee388..55e8357eee7 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17478
-output = 0.69912
+input = 0.17424
+output = 0.69696
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index f14d2b0ce24..1d9a07d1077 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.04868
-output = 1.04868
+input = 1.045439
+output = 1.045439

From cc35b10c2f936644379b80e5797fa5f7f5228c31 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 15:27:49 +0000
Subject: [PATCH 0994/2012] chore(sync): update NanoGPT model catalog (#6720)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gemma-4-12b-it-semancer.toml       |  4 +--
 .../models/gemma-4-12b-it-station-keeper.toml |  4 +--
 providers/nano-gpt/models/k2-horizon-7b.toml  | 27 -------------------
 3 files changed, 4 insertions(+), 31 deletions(-)
 delete mode 100644 providers/nano-gpt/models/k2-horizon-7b.toml

diff --git a/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
index 0a5b660c5d9..a0e66888780 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it-semancer.toml
@@ -16,8 +16,8 @@ output = 0.25
 cache_read = 0.025
 
 [limit]
-context = 262_144
-input = 262_144
+context = 131_072
+input = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
index a68502a86c6..c4ccf5f509a 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it-station-keeper.toml
@@ -16,8 +16,8 @@ output = 0.25
 cache_read = 0.025
 
 [limit]
-context = 262_144
-input = 262_144
+context = 131_072
+input = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/k2-horizon-7b.toml b/providers/nano-gpt/models/k2-horizon-7b.toml
deleted file mode 100644
index ee81d369b10..00000000000
--- a/providers/nano-gpt/models/k2-horizon-7b.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "K2-Horizon-7B"
-description = "IFM's open-weight 7B model for coding, reasoning, and chat. Heabsy serves an FP8 deployment on a dedicated RTX 5090 with speculative decoding in the United States, with a 131K context window and an 8K output limit. Zero data retention is not verified for this deployment."
-release_date = "2026-09-09"
-last_updated = "2026-09-09"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = false
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.05
-output = 0.2
-cache_read = 0.04
-
-[limit]
-context = 131_072
-input = 131_072
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]

From ec598c02ae559ed8c16c6f47360aa5a6b452318e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 15:28:04 +0000
Subject: [PATCH 0995/2012] chore(sync): update OpenRouter model catalog
 (#6719)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 80a0d4a1f36..2b14e7625ce 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0075
-output = 3.41
-cache_read = 0.2015
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 128_000
 
 [modalities]
 input = ["text"]

From f7b0eda070c66ba93d2b0d0e444ce8e29e2c0d48 Mon Sep 17 00:00:00 2001
From: Jiangfan Ran <104256115+ran894694447@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:10:24 +0800
Subject: [PATCH 0996/2012] Add DeepSeek V4.1 Flash for deepseek provider
 (#6721)

* Add DeepSeek V4.1 Flash for deepseek provider

- add providers/deepseek/models/deepseek-flash.toml mapping to deepseek/deepseek-v4.1-flash
- mark deepseek-v4-flash and deepseek-v4-flash-vision-exp deprecated (retired; served by V4.1 Flash)

* Address review: drop redundant name, restore wire-path comments

* Comment out deprecated status in deepseek-v4-flash.toml

Commented out the deprecated status line in the TOML file.

* Comment out deprecated status in TOML configuration

Commented out the deprecated status line for clarity.

---------

Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
---
 providers/deepseek/models/deepseek-flash.toml | 28 +++++++++++++++++++
 .../models/deepseek-v4-flash-vision-exp.toml  | 26 +++++++++--------
 .../deepseek/models/deepseek-v4-flash.toml    | 25 +++++++++--------
 3 files changed, 55 insertions(+), 24 deletions(-)
 create mode 100644 providers/deepseek/models/deepseek-flash.toml

diff --git a/providers/deepseek/models/deepseek-flash.toml b/providers/deepseek/models/deepseek-flash.toml
new file mode 100644
index 00000000000..cb294223344
--- /dev/null
+++ b/providers/deepseek/models/deepseek-flash.toml
@@ -0,0 +1,28 @@
+# DeepSeek-V4.1-Flash, served as `deepseek-flash`.
+# The legacy names `deepseek-v4-flash` and `deepseek-v4-flash-vision-exp` are
+# still accepted, but those models are retired and requests are served by
+# DeepSeek-V4.1-Flash at the Flash price.
+# Reasoning tokens are billed at the output rate (no separate CoT price).
+# Off-peak rates are half of peak; models.dev records the off-peak rate.
+# OpenAI: `thinking.type = enabled|disabled`, `reasoning_effort = low|high|max`.
+# Anthropic: `thinking.type`, `output_config.effort = low|high|max`; budget ignored.
+# Flash maps requested low -> low (unlike Pro). xhigh -> high.
+# https://api-docs.deepseek.com/quick_start/pricing (accessed 2026-09-10)
+# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-09-10)
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.6
+reasoning = 0.6
+cache_read = 0.003
diff --git a/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml b/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
index ec428639dc9..e1227c7a467 100644
--- a/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
+++ b/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
@@ -1,11 +1,13 @@
-# DeepSeek-V4-Flash-Vision-Exp is priced the same as DeepSeek V4 Flash.
-# https://api-docs.deepseek.com/quick_start/pricing/ (accessed 2026-08-22)
-# https://api-docs.deepseek.com/guides/vision (accessed 2026-08-22)
-# OpenAI: `thinking.type = enabled|disabled`, `reasoning_effort = low|high|max`.
-# Anthropic: `thinking.type`, `output_config.effort = low|high|max`; budget ignored.
-base_model = "deepseek/deepseek-v4-flash-vision-exp"
-status = "beta"
-
+# Legacy alias. DeepSeek-V4-Flash-Vision-Exp is retired; requests to
+# `deepseek-v4-flash-vision-exp` are still accepted but served by
+# DeepSeek-V4.1-Flash and billed at the Flash price.
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+# https://api-docs.deepseek.com/quick_start/pricing (accessed 2026-09-10)
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4 Flash Vision Exp"
+## mark as deprecated soon
+# status = "deprecated"
 [[reasoning_options]]
 type = "toggle"
 
@@ -17,7 +19,7 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.14
-output = 0.28
-reasoning = 0.28
-cache_read = 0.0028
+input = 0.15
+output = 0.6
+reasoning = 0.6
+cache_read = 0.003
diff --git a/providers/deepseek/models/deepseek-v4-flash.toml b/providers/deepseek/models/deepseek-v4-flash.toml
index 9dca906c5c6..ae7ff2fc79b 100644
--- a/providers/deepseek/models/deepseek-v4-flash.toml
+++ b/providers/deepseek/models/deepseek-v4-flash.toml
@@ -1,12 +1,13 @@
-# Reasoning tokens are billed at the output rate (no separate CoT price).
-# `completion_tokens_details.reasoning_tokens` is a subset of completion_tokens.
-# https://api-docs.deepseek.com/quick_start/pricing/ (accessed 2026-07-31)
-# OpenAI: `thinking.type = enabled|disabled`, `reasoning_effort = low|high|max`.
-# Anthropic: `thinking.type`, `output_config.effort = low|high|max`; budget ignored.
-# Flash maps requested low→low (unlike Pro, which maps low→high). xhigh→high.
-# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-08-02)
-base_model = "deepseek/deepseek-v4-flash-0731"
+# Legacy alias. The DeepSeek-V4-Flash model is retired; requests to
+# `deepseek-v4-flash` are still accepted but served by DeepSeek-V4.1-Flash and
+# billed at the Flash price.
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+# https://api-docs.deepseek.com/quick_start/pricing (accessed 2026-09-10)
+base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4 Flash"
+## mark as deprecated soon
+# status = "deprecated"
 
 [[reasoning_options]]
 type = "toggle"
@@ -19,7 +20,7 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.14
-output = 0.28
-reasoning = 0.28
-cache_read = 0.0028
+input = 0.15
+output = 0.6
+reasoning = 0.6
+cache_read = 0.003

From 38ddea2e15916c368db1256b1f1cd6064a5c8657 Mon Sep 17 00:00:00 2001
From: David Knaack <david.knaack@sap.com>
Date: Thu, 10 Sep 2026 18:10:36 +0200
Subject: [PATCH 0997/2012] chore(sap-ai-core): add gemini-3.5-flash-lite
 (#6708)

---
 providers/sap-ai-core/models/gemini-3.5-flash-lite.toml | 6 ++++++
 1 file changed, 6 insertions(+)
 create mode 100644 providers/sap-ai-core/models/gemini-3.5-flash-lite.toml

diff --git a/providers/sap-ai-core/models/gemini-3.5-flash-lite.toml b/providers/sap-ai-core/models/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..f2b2b13999b
--- /dev/null
+++ b/providers/sap-ai-core/models/gemini-3.5-flash-lite.toml
@@ -0,0 +1,6 @@
+# Cost information is in CU that are not directly comparable to other providers.
+base_model = "google/gemini-3.5-flash-lite"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
\ No newline at end of file

From ae667d5b8c607efc653a238c3c611966ad140cb8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E8=BD=BB=E5=B0=98?=
 <53001305+SlightDust@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:10:46 +0800
Subject: [PATCH 0998/2012] fix(scnet-token-plan): remove retired DeepSeek-V3.2
 and MiMo-V2.5-Pro (#6703)

---
 providers/scnet-token-plan/models/DeepSeek-V3.2.toml | 10 ----------
 providers/scnet-token-plan/models/MiMo-V2.5-Pro.toml | 10 ----------
 2 files changed, 20 deletions(-)
 delete mode 100644 providers/scnet-token-plan/models/DeepSeek-V3.2.toml
 delete mode 100644 providers/scnet-token-plan/models/MiMo-V2.5-Pro.toml

diff --git a/providers/scnet-token-plan/models/DeepSeek-V3.2.toml b/providers/scnet-token-plan/models/DeepSeek-V3.2.toml
deleted file mode 100644
index 94dbb40951e..00000000000
--- a/providers/scnet-token-plan/models/DeepSeek-V3.2.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "deepseek/deepseek-v3.2"
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
diff --git a/providers/scnet-token-plan/models/MiMo-V2.5-Pro.toml b/providers/scnet-token-plan/models/MiMo-V2.5-Pro.toml
deleted file mode 100644
index 6e465f58869..00000000000
--- a/providers/scnet-token-plan/models/MiMo-V2.5-Pro.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "xiaomi/mimo-v2.5-pro"
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0

From 708b9f1d1bf799eb0585f72ac651907c0447fac0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:10:54 -0500
Subject: [PATCH 0999/2012] chore(sync): update Requesty model catalog (#6700)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../requesty/models/deepseek-v4.1-flash.toml    | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/requesty/models/deepseek-v4.1-flash.toml

diff --git a/providers/requesty/models/deepseek-v4.1-flash.toml b/providers/requesty/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..cf1db66a44a
--- /dev/null
+++ b/providers/requesty/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,17 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216

From 5eb59587e0357c68a8290896ab955d4d3a92cc9e Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:11:38 -0500
Subject: [PATCH 1000/2012] fix: [missing-model] ofox:
 deepseek/deepseek-v4.1-flash (#6694)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4.1-flash.toml   | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/ofox/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml b/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..625c4e0903d
--- /dev/null
+++ b/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,18 @@
+# https://ofox.ai/models/deepseek/deepseek-v4.1-flash
+# https://api-docs.deepseek.com/quick_start/pricing/
+# https://api-docs.deepseek.com/guides/thinking_mode/
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max (Flash; lab maps xhigh→high)
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006

From 47833cc87158e9fbcc88352ef73a3d565f9dab12 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Thu, 10 Sep 2026 23:12:12 +0700
Subject: [PATCH 1001/2012] feat(bothub): add GLM-5.3, GLM-5.3-Flash, DeepSeek
 V4 Flash 0731, V4 Pro 0813 and GPT-5.6 Luna (#6680)

* feat(bothub): add GLM-5.3, GLM-5.3-Flash, DeepSeek V4 Flash 0731, V4 Pro 0813 and GPT-5.6 Luna

* fix(bothub): replace unverified thinking.type toggle with reasoning_effort for DeepSeek V4

---------

Co-authored-by: 4ebuRushka <4ebuRushka@users.noreply.github.com>
---
 .../bothub/models/deepseek-v4-flash-0731.toml   | 17 +++++++++++++++++
 .../bothub/models/deepseek-v4-pro-0813.toml     | 17 +++++++++++++++++
 providers/bothub/models/glm-5.3-flash.toml      | 16 ++++++++++++++++
 providers/bothub/models/glm-5.3.toml            | 16 ++++++++++++++++
 providers/bothub/models/gpt-5.6-luna.toml       | 16 ++++++++++++++++
 5 files changed, 82 insertions(+)
 create mode 100644 providers/bothub/models/deepseek-v4-flash-0731.toml
 create mode 100644 providers/bothub/models/deepseek-v4-pro-0813.toml
 create mode 100644 providers/bothub/models/glm-5.3-flash.toml
 create mode 100644 providers/bothub/models/glm-5.3.toml
 create mode 100644 providers/bothub/models/gpt-5.6-luna.toml

diff --git a/providers/bothub/models/deepseek-v4-flash-0731.toml b/providers/bothub/models/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..24886981b87
--- /dev/null
+++ b/providers/bothub/models/deepseek-v4-flash-0731.toml
@@ -0,0 +1,17 @@
+# Pricing: https://bothub.ru/models (accessed 2026-09-10).
+# RUB converted to USD at the CBR rate 85.4594 RUB/USD (2026-09-10).
+# Controls: bothub exposes a single reasoning_effort parameter on
+# POST /v1/chat/completions (see providers/bothub/provider.toml); no
+# thinking.type toggle is documented for this host. Levels verified
+# against the bothub model page; "none" disables reasoning.
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.1
+output = 0.28
diff --git a/providers/bothub/models/deepseek-v4-pro-0813.toml b/providers/bothub/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..f1dda9cf1a9
--- /dev/null
+++ b/providers/bothub/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,17 @@
+# Pricing: https://bothub.ru/models (accessed 2026-09-10).
+# RUB converted to USD at the CBR rate 85.4594 RUB/USD (2026-09-10).
+# Controls: bothub exposes a single reasoning_effort parameter on
+# POST /v1/chat/completions (see providers/bothub/provider.toml); no
+# thinking.type toggle is documented for this host. Levels verified
+# against the bothub model page; "none" disables reasoning.
+base_model = "deepseek/deepseek-v4-pro-0813"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[cost]
+input = 1.61
+output = 4.84
diff --git a/providers/bothub/models/glm-5.3-flash.toml b/providers/bothub/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..1a2c037bfb1
--- /dev/null
+++ b/providers/bothub/models/glm-5.3-flash.toml
@@ -0,0 +1,16 @@
+# Pricing: https://bothub.ru/models (accessed 2026-09-10) — includes the
+# site's current -20% promo (list 13.15/47.33 RUB per 1M tokens).
+# RUB converted to USD at the CBR rate 85.4594 RUB/USD (2026-09-10).
+# Controls mirrored from the lab entry: GLM-5.3-Flash always reasons;
+# effort low|high|max with default max, no toggle.
+base_model = "zhipuai/glm-5.3-flash"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.12
+output = 0.44
diff --git a/providers/bothub/models/glm-5.3.toml b/providers/bothub/models/glm-5.3.toml
new file mode 100644
index 00000000000..11332821aeb
--- /dev/null
+++ b/providers/bothub/models/glm-5.3.toml
@@ -0,0 +1,16 @@
+# Pricing: https://bothub.ru/models (accessed 2026-09-10) — includes the
+# site's current -20% promo (list 184.04/578.41 RUB per 1M tokens).
+# RUB converted to USD at the CBR rate 85.4594 RUB/USD (2026-09-10).
+# Controls mirrored from the lab entry: GLM-5.3 always reasons; effort
+# low|high|max with default max, no toggle.
+base_model = "zhipuai/glm-5.3"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.72
+output = 5.41
diff --git a/providers/bothub/models/gpt-5.6-luna.toml b/providers/bothub/models/gpt-5.6-luna.toml
new file mode 100644
index 00000000000..4b59060abd2
--- /dev/null
+++ b/providers/bothub/models/gpt-5.6-luna.toml
@@ -0,0 +1,16 @@
+# Pricing: https://bothub.ru/models (accessed 2026-09-10) — includes the
+# site's current -80% promo (list 26.30/157.75 RUB per 1M tokens).
+# RUB converted to USD at the CBR rate 85.4594 RUB/USD (2026-09-10).
+# Controls mirrored from the lab entry: off is effort=none; graded
+# levels — no toggle.
+base_model = "openai/gpt-5.6-luna"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.06
+output = 0.37

From d80318b4383dcf203a9c5afe8ea9d1bc7cffd749 Mon Sep 17 00:00:00 2001
From: realtimegb <support@realtimecomms.co.uk>
Date: Thu, 10 Sep 2026 17:12:19 +0100
Subject: [PATCH 1002/2012] above: remove MiMo V2.5 Pro UltraSpeed (#6679)

Xiaomi ended the MiMo-V2.5-Pro-UltraSpeed limited beta and discontinued the
model API at 00:00 Beijing time on 2026-09-08. above.dev no longer serves
mimo-v2.5-pro-ultraspeed, so the listing is removed to stop agent tools
offering a model that now returns an error.

The standard mimo-v2.5-pro listing is unchanged; its pricing comment drops
the UltraSpeed rate that no longer applies.


Claude-Session: https://claude.ai/code/session_018AGd989iTeCRCrupfdh7ES

Co-authored-by: realtimegb <pkreil3@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
---
 .../models/mimo-v2.5-pro-ultraspeed.toml      | 20 -------------------
 providers/above/models/mimo-v2.5-pro.toml     |  2 +-
 2 files changed, 1 insertion(+), 21 deletions(-)
 delete mode 100644 providers/above/models/mimo-v2.5-pro-ultraspeed.toml

diff --git a/providers/above/models/mimo-v2.5-pro-ultraspeed.toml b/providers/above/models/mimo-v2.5-pro-ultraspeed.toml
deleted file mode 100644
index cebe2ae0c69..00000000000
--- a/providers/above/models/mimo-v2.5-pro-ultraspeed.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-# Passthrough to the official MiMo API (api.xiaomimimo.com); same model as
-# mimo-v2.5-pro served at up to ~1000 tok/s, 3x price for speed.
-# Toggle: `thinking.type = enabled|disabled`, as providers/xiaomi; verified
-# end-to-end on this host 2026-08-29.
-# RMB upstream (pro CNY 3/0.025/6, ultraspeed CNY 9/0.075/18 per MTok)
-# converted at 6.5 CNY/USD (operator rate set 2026-07, spot ~6.79; buffer
-# covers FX drift), then a flat 10% markup; rounded to 4dp. Checked 2026-08-29.
-base_model = "xiaomi/mimo-v2.5-pro-ultraspeed"
-name = "MiMo V2.5 Pro UltraSpeed"
-
-[[reasoning_options]]
-type = "toggle"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.5231
-output = 3.0462
-cache_read = 0.0127
diff --git a/providers/above/models/mimo-v2.5-pro.toml b/providers/above/models/mimo-v2.5-pro.toml
index c1a7dec3005..682e84d3866 100644
--- a/providers/above/models/mimo-v2.5-pro.toml
+++ b/providers/above/models/mimo-v2.5-pro.toml
@@ -1,7 +1,7 @@
 # Passthrough to the official MiMo API (api.xiaomimimo.com).
 # Toggle: `thinking.type = enabled|disabled`, as providers/xiaomi; verified
 # end-to-end on this host 2026-08-29.
-# RMB upstream (pro CNY 3/0.025/6, ultraspeed CNY 9/0.075/18 per MTok)
+# RMB upstream (CNY 3/0.025/6 per MTok)
 # converted at 6.5 CNY/USD (operator rate set 2026-07, spot ~6.79; buffer
 # covers FX drift), then a flat 10% markup; rounded to 4dp. Checked 2026-08-29.
 base_model = "xiaomi/mimo-v2.5-pro"

From 7bfc10da7c84367ef814668105151f6da79b2ea1 Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Fri, 11 Sep 2026 00:13:31 +0800
Subject: [PATCH 1003/2012] fix(vivgrid): openai compitable api (#6659)

* feat(vivgrid): add gpt-6-astra model

* fix(vivgrid): openai compatible api
---
 providers/vivgrid/models/claude-fable-5-1.toml | 3 +++
 providers/vivgrid/models/claude-fable-5.toml   | 3 +++
 providers/vivgrid/models/gemini-3.8-flash.toml | 3 +++
 3 files changed, 9 insertions(+)

diff --git a/providers/vivgrid/models/claude-fable-5-1.toml b/providers/vivgrid/models/claude-fable-5-1.toml
index 93b3075879a..471e3cb571d 100644
--- a/providers/vivgrid/models/claude-fable-5-1.toml
+++ b/providers/vivgrid/models/claude-fable-5-1.toml
@@ -11,3 +11,6 @@ input = 10
 output = 50
 cache_read = 0.5
 cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
\ No newline at end of file
diff --git a/providers/vivgrid/models/claude-fable-5.toml b/providers/vivgrid/models/claude-fable-5.toml
index 239c250c0d4..912e2ff029a 100644
--- a/providers/vivgrid/models/claude-fable-5.toml
+++ b/providers/vivgrid/models/claude-fable-5.toml
@@ -11,3 +11,6 @@ input = 10
 output = 50
 cache_read = 1.25
 cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
\ No newline at end of file
diff --git a/providers/vivgrid/models/gemini-3.8-flash.toml b/providers/vivgrid/models/gemini-3.8-flash.toml
index 32db1ca1276..758290832b2 100644
--- a/providers/vivgrid/models/gemini-3.8-flash.toml
+++ b/providers/vivgrid/models/gemini-3.8-flash.toml
@@ -12,3 +12,6 @@ cache_read = 0.15
 
 [limit]
 output = 128_000
+
+[provider]
+npm = "@ai-sdk/openai-compatible"

From 487fbf1cb2a4c95ddd1e412c4ee41c1fb86def1a Mon Sep 17 00:00:00 2001
From: Karl Sabo <karl.sabo@klaviyo.com>
Date: Thu, 10 Sep 2026 10:13:51 -0600
Subject: [PATCH 1004/2012] feat(fireworks): add Mistral Large 3 FP8 (#6651)

---
 .../accounts/fireworks/models/mistral-large-3-fp8.toml      | 6 ++++++
 1 file changed, 6 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml b/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml
new file mode 100644
index 00000000000..4f5831e7930
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml
@@ -0,0 +1,6 @@
+# Fireworks identifies this on-demand FP8 endpoint as Mistral Large 3 675B Instruct 2512,
+# with 256K context, function calling, and image input. It is not offered serverlessly,
+# so there is no per-token price to record:
+# https://fireworks.ai/models/fireworks/mistral-large-3-fp8 (accessed 2026-09-05)
+base_model = "mistral/mistral-large-2512"
+name = "Mistral Large 3 675B Instruct 2512"

From 87e4c465c76ad0839b8dbe1092d984ab866e6ea3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:16:30 -0500
Subject: [PATCH 1005/2012] chore(sync): update Hugging Face model catalog
 (#6695)

* chore(sync): update Hugging Face model catalog

* fix(huggingface): add DeepSeek V4.1 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/deepseek-ai/DeepSeek-V4.1-Flash.toml      | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/huggingface/models/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/huggingface/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/huggingface/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..4907c3e93df
--- /dev/null
+++ b/providers/huggingface/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,12 @@
+# Effort: reasoning_effort = none|low|high|xhigh|max; none disables thinking.
+# https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4.1-Flash (accessed 2026-09-10)
+base_model = "deepseek/deepseek-v4.1-flash"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.3
+output = 1.2
+
+[limit]
+context = 1_048_576

From 7a93ae74e15ef83f4aaa5483ee0a913bc5df4595 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:16:38 -0500
Subject: [PATCH 1006/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6675)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add DeepSeek V4.1 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/deepseek/deepseek-v3.2.toml |  5 ++--
 .../deepseek/deepseek-v4.1-flash-beta.toml    | 25 -------------------
 .../models/deepseek/deepseek-v4.1-flash.toml  | 13 ++++++++++
 3 files changed, 15 insertions(+), 28 deletions(-)
 delete mode 100644 providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
 create mode 100644 providers/vercel/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/vercel/models/deepseek/deepseek-v3.2.toml b/providers/vercel/models/deepseek/deepseek-v3.2.toml
index af269904ff9..32a4b99d4ca 100644
--- a/providers/vercel/models/deepseek/deepseek-v3.2.toml
+++ b/providers/vercel/models/deepseek/deepseek-v3.2.toml
@@ -4,9 +4,8 @@ reasoning = false
 tool_call = false
 
 [cost]
-input = 0.28
-output = 0.42
-cache_read = 0.028
+input = 0.62
+output = 1.85
 
 [limit]
 output = 8_000
diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
deleted file mode 100644
index 0a7f78795e6..00000000000
--- a/providers/vercel/models/deepseek/deepseek-v4.1-flash-beta.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# Effort: reasoning.effort = high|xhigh (Vercel gateway aliases)
-name = "DeepSeek V4.1 Flash Beta"
-description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-family = "deepseek"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "xhigh"] }]
-
-[cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
-
-[limit]
-context = 1_000_000
-output = 384_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..3381d473c24
--- /dev/null
+++ b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,13 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning.effort = high|xhigh (Vercel gateway aliases)
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "xhigh"] }]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216

From f51c5e61d0df19b476967b983c4327cb3a132900 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:16:47 -0500
Subject: [PATCH 1007/2012] chore(sync): update Deep Infra model catalog
 (#6688)

* chore(sync): update Deep Infra model catalog

* fix(deepinfra): add DeepSeek V4.1 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/deepseek-ai/DeepSeek-V4.1-Flash.toml     | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..719c681bfaf
--- /dev/null
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning_effort = none|low|high|xhigh|max; none disables thinking.
+# https://api.deepinfra.com/models/deepseek-ai/DeepSeek-V4.1-Flash/schema/openai-chat-completions
+# https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4.1-Flash (accessed 2026-09-10)
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576

From 7226cd595334f9d0b213840fb03c4f43250b04b5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 16:27:00 +0000
Subject: [PATCH 1008/2012] chore(sync): update OpenRouter model catalog
 (#6724)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 .../openrouter/models/google/gemma-4-26b-a4b-it.toml     | 9 ++++-----
 providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml    | 7 +++----
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 4 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..a010dd38ba2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.87
+output = 1.74
+cache_read = 0.0725
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index a79fba4f50b..a27c6e610ef 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -1,14 +1,13 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "google/gemma-4-26b-a4b-it"
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-input = 0.07
-output = 0.34
-
-[limit]
-output = 16_384
+input = 0.042
+output = 0.22
 
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index ac95b282785..f33ff42c33e 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,12 +6,11 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.9
-cache_read = 0.05
+input = 0.05
+output = 0.7
 
 [limit]
-output = 235_929
+output = 32_768
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 9a7fd21f10db0f5ae41a6f8c9acc168a0840a4e4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 16:27:29 +0000
Subject: [PATCH 1009/2012] chore(sync): update Kilo model catalog (#6725)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/google/gemma-4-26b-a4b-it.toml | 2 +-
 providers/kilo/models/qwen/qwen3.6-35b-a3b.toml      | 7 +++----
 providers/kilo/models/tencent/hy3.toml               | 6 +++---
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/google/gemma-4-26b-a4b-it.toml b/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
index 512c2f23404..efcf99dab3b 100644
--- a/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
@@ -10,7 +10,7 @@ input = 0.042
 output = 0.22
 
 [limit]
-output = 16_384
+context = 131_072
 
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index 1385363812a..19df61b6e3c 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,12 +6,11 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.1
-output = 0.9
-cache_read = 0.05
+input = 0.05
+output = 0.7
 
 [limit]
-output = 235_929
+output = 32_768
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 068d3f4ab3255599ab3bd5c63ed82632d793a36a Mon Sep 17 00:00:00 2001
From: Chirag Patil <chirag.patil@onixnet.com>
Date: Thu, 10 Sep 2026 21:57:59 +0530
Subject: [PATCH 1010/2012] feat(google-vertex): add xAI Grok models (4.6, 4.3,
 4.20, 4.1 Fast) (#6690)

* feat(google-vertex): add Grok 4.6

* feat(google-vertex): add Grok 4.1 Fast, 4.20, and 4.3 models

* fix(google-vertex): correct Grok model metadata

---------

Co-authored-by: Chirag Patil <244892032+chiragpatil-lp@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/xai/grok-4.1-fast-reasoning.toml       | 22 +++++++++++++++
 .../xai/grok-4.1-fast-non-reasoning.toml      | 18 ++++++++++++
 .../models/xai/grok-4.1-fast-reasoning.toml   | 20 +++++++++++++
 .../models/xai/grok-4.20-non-reasoning.toml   | 25 +++++++++++++++++
 .../models/xai/grok-4.20-reasoning.toml       | 27 ++++++++++++++++++
 .../google-vertex/models/xai/grok-4.3.toml    | 28 +++++++++++++++++++
 .../google-vertex/models/xai/grok-4.6.toml    | 25 +++++++++++++++++
 7 files changed, 165 insertions(+)
 create mode 100644 models/xai/grok-4.1-fast-reasoning.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.1-fast-non-reasoning.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.1-fast-reasoning.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.20-non-reasoning.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.20-reasoning.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.3.toml
 create mode 100644 providers/google-vertex/models/xai/grok-4.6.toml

diff --git a/models/xai/grok-4.1-fast-reasoning.toml b/models/xai/grok-4.1-fast-reasoning.toml
new file mode 100644
index 00000000000..497dc490856
--- /dev/null
+++ b/models/xai/grok-4.1-fast-reasoning.toml
@@ -0,0 +1,22 @@
+# Sources:
+# - https://x.ai/news/grok-4-1-fast
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-1-fast
+name = "Grok 4.1 Fast (Reasoning)"
+description = "xAI's fast agentic tool-calling model with a 2M context window and built-in reasoning"
+family = "grok"
+release_date = "2025-11-19"
+last_updated = "2025-11-19"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 2_000_000
+output = 30_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/google-vertex/models/xai/grok-4.1-fast-non-reasoning.toml b/providers/google-vertex/models/xai/grok-4.1-fast-non-reasoning.toml
new file mode 100644
index 00000000000..d46be5858d7
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.1-fast-non-reasoning.toml
@@ -0,0 +1,18 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-1-fast
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.1-fast"
+description = "Fast Grok model for responsive chat, tool-assisted work, and low-latency responses"
+status = "deprecated"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 128_000
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/xai/grok-4.1-fast-reasoning.toml b/providers/google-vertex/models/xai/grok-4.1-fast-reasoning.toml
new file mode 100644
index 00000000000..842a77c5559
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.1-fast-reasoning.toml
@@ -0,0 +1,20 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-1-fast
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/capabilities/reasoning
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.1-fast-reasoning"
+description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
+reasoning_options = []
+status = "deprecated"
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 128_000
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/xai/grok-4.20-non-reasoning.toml b/providers/google-vertex/models/xai/grok-4.20-non-reasoning.toml
new file mode 100644
index 00000000000..6d7b518406b
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.20-non-reasoning.toml
@@ -0,0 +1,25 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-20
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.20-0309-non-reasoning"
+
+[cost]
+input = 1.25
+output = 2.5
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
+[limit]
+context = 2_000_000
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/xai/grok-4.20-reasoning.toml b/providers/google-vertex/models/xai/grok-4.20-reasoning.toml
new file mode 100644
index 00000000000..5ddd9c53548
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.20-reasoning.toml
@@ -0,0 +1,27 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-20
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/capabilities/reasoning
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.20-0309-reasoning"
+reasoning_options = []
+
+[cost]
+input = 1.25
+output = 2.5
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
+[limit]
+context = 2_000_000
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/xai/grok-4.3.toml b/providers/google-vertex/models/xai/grok-4.3.toml
new file mode 100644
index 00000000000..926ef300cf9
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.3.toml
@@ -0,0 +1,28 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-3
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/capabilities/reasoning
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.3"
+reasoning_options = []
+status = "beta"
+
+[cost]
+input = 1.25
+output = 2.5
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
+[limit]
+context = 200_000
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/xai/grok-4.6.toml b/providers/google-vertex/models/xai/grok-4.6.toml
new file mode 100644
index 00000000000..b9a3ebb2596
--- /dev/null
+++ b/providers/google-vertex/models/xai/grok-4.6.toml
@@ -0,0 +1,25 @@
+# Sources:
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/grok-4-6
+# - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/grok/capabilities/reasoning
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "xai/grok-4.6"
+reasoning_options = []
+status = "beta"
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+
+[limit]
+context = 524_288
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"

From 59d8546ddb649cd42608c9d6d4afd134df86c576 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 11:34:03 -0500
Subject: [PATCH 1011/2012] fix: Remove stale qwen3.8-max-preview rows from
 Alibaba Token Plan (#6663)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/qwen3.8-max-preview.toml           | 30 +++++++------------
 .../models/qwen3.8-max-preview.toml           | 30 +++++++------------
 2 files changed, 22 insertions(+), 38 deletions(-)

diff --git a/providers/alibaba-token-plan-cn/models/qwen3.8-max-preview.toml b/providers/alibaba-token-plan-cn/models/qwen3.8-max-preview.toml
index 2d969d268dc..5f48a231c3e 100644
--- a/providers/alibaba-token-plan-cn/models/qwen3.8-max-preview.toml
+++ b/providers/alibaba-token-plan-cn/models/qwen3.8-max-preview.toml
@@ -1,22 +1,14 @@
-# Sources (accessed 2026-07-20):
+# Sources (accessed 2026-09-10):
 # https://platform.qianwenai.com/docs/token-plan/personal/token-plan-personal-overview
-# https://platform.qianwenai.com/docs/token-plan/personal/token-plan-personal-quickstart
-# https://platform.qianwenai.com/docs/developer-guides/clients-and-developer-tools/opencode
-# https://platform.qianwenai.com/docs/developer-guides/clients-and-developer-tools/kilo-cli
-# https://platform.qianwenai.com/docs/developer-guides/text-generation/thinking
-# https://platform.qianwenai.com/docs/api-reference/chat/openai-chat
-# Thinking is always enabled. In thinking mode, temperature defaults to 0.6
-# and values below 0.6 are raised to 0.6. This provider uses the OpenAI-compatible
-# Chat API, whose native qwen3.8 effort values are low/medium/xhigh (default
-# xhigh); the standard high value is accepted as an alias and maps to xhigh.
-# The OpenCode guide labels its tiers low/high/xhigh and configures a 262144-token
-# thinking budget for the China endpoint. `thinking_budget` (0..262144) cannot
-# be combined with `reasoning_effort`. A zero budget maps to low effort; it does
-# not disable thinking. API: {"reasoning_effort":"medium"} or
-# {"thinking_budget":16384}. `preserve_thinking` defaults to true, so clients must
-# replay every historical `reasoning_content` value unchanged.
-# structured_output: the gateway accepts response_format json_schema (probed
-# 2026-07-24 on the Singapore endpoint; China shares the catalogue).
+# https://help.aliyun.com/zh/model-studio/token-plan-personal-overview
+# Model update: qwen3.8-max-preview ended preview and is officially retired from
+# the Token Plan supported-models allowlist. Docs say legacy calls may still
+# route to production qwen3.8-max (Credits billed as qwen3.8-max); clients should
+# use model ID qwen3.8-max. Marked deprecated so it is not offered as a current
+# Token Plan selection. Prefer providers/.../qwen3.8-max.toml.
+# Prior sources (accessed 2026-07-20): thinking always-on; OpenAI-compatible Chat
+# effort low/medium/xhigh (high aliases to xhigh); thinking_budget 0..262144
+# mutually exclusive with reasoning_effort; preserve_thinking defaults true.
 
 base_model = "alibaba/qwen3.8-max-preview"
 structured_output = true
@@ -24,7 +16,7 @@ reasoning_options = [
   { type = "effort", values = ["low", "medium", "xhigh"] },
   { type = "budget_tokens", min = 0, max = 262_144 },
 ]
-status = "beta"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/alibaba-token-plan/models/qwen3.8-max-preview.toml b/providers/alibaba-token-plan/models/qwen3.8-max-preview.toml
index f887d05a68a..deb4f08b1ea 100644
--- a/providers/alibaba-token-plan/models/qwen3.8-max-preview.toml
+++ b/providers/alibaba-token-plan/models/qwen3.8-max-preview.toml
@@ -1,22 +1,14 @@
-# Sources (accessed 2026-07-20):
+# Sources (accessed 2026-09-10):
 # https://docs.qwencloud.com/token-plan/personal/token-plan-personal-overview
-# https://docs.qwencloud.com/token-plan/personal/token-plan-personal-quickstart
-# https://docs.qwencloud.com/developer-guides/clients-and-developer-tools/opencode
-# https://docs.qwencloud.com/developer-guides/clients-and-developer-tools/kilo-cli
-# https://docs.qwencloud.com/developer-guides/text-generation/thinking
-# https://docs.qwencloud.com/api-reference/chat/openai-chat
-# Thinking is always enabled. In thinking mode, temperature defaults to 0.6
-# and values below 0.6 are raised to 0.6. This provider uses the OpenAI-compatible
-# Chat API, whose native qwen3.8 effort values are low/medium/xhigh (default
-# xhigh); the standard high value is accepted as an alias and maps to xhigh.
-# The OpenCode guide labels its tiers low/high/xhigh and configures a 99072-token
-# thinking budget for the international endpoint. `thinking_budget` (0..262144)
-# cannot be combined with `reasoning_effort`. A zero budget maps to low effort;
-# it does not disable thinking. API: {"reasoning_effort":"medium"} or
-# {"thinking_budget":16384}. `preserve_thinking` defaults to true, so clients must
-# replay every historical `reasoning_content` value unchanged.
-# structured_output: the gateway accepts response_format json_schema (probed
-# 2026-07-24: strict schema request returned {"n": 7}).
+# https://www.alibabacloud.com/help/en/model-studio/token-plan-overview
+# Model update: qwen3.8-max-preview ended preview and is officially retired from
+# the Token Plan supported-models allowlist. Docs say legacy calls may still
+# route to production qwen3.8-max (Credits billed as qwen3.8-max); clients should
+# use model ID qwen3.8-max. Marked deprecated so it is not offered as a current
+# Token Plan selection. Prefer providers/.../qwen3.8-max.toml.
+# Prior sources (accessed 2026-07-20): thinking always-on; OpenAI-compatible Chat
+# effort low/medium/xhigh (high aliases to xhigh); thinking_budget 0..262144
+# mutually exclusive with reasoning_effort; preserve_thinking defaults true.
 
 base_model = "alibaba/qwen3.8-max-preview"
 structured_output = true
@@ -24,7 +16,7 @@ reasoning_options = [
   { type = "effort", values = ["low", "medium", "xhigh"] },
   { type = "budget_tokens", min = 0, max = 262_144 },
 ]
-status = "beta"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"

From 28a2974f5568a612398e6d9668f30bbec6061928 Mon Sep 17 00:00:00 2001
From: KisenaMiyuki <69066601+KisenaMiyuki@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:34:13 +0800
Subject: [PATCH 1012/2012] feat(inception): add mercury 2.5 model definition
 (#6674)

Co-authored-by: sugar@work <ssagisetsu@gmail.com>
---
 providers/inception/models/mercury-2.5.toml | 29 +++++++++++++++++++++
 1 file changed, 29 insertions(+)
 create mode 100644 providers/inception/models/mercury-2.5.toml

diff --git a/providers/inception/models/mercury-2.5.toml b/providers/inception/models/mercury-2.5.toml
new file mode 100644
index 00000000000..306f7ae383e
--- /dev/null
+++ b/providers/inception/models/mercury-2.5.toml
@@ -0,0 +1,29 @@
+name = "Mercury 2.5"
+description = "Mercury 2.5 is the fastest reasoning LLM, and the latest diffusion LLM (dLLM) from Inception"
+# Model exception (accessed 2026-06-25): `reasoning_effort = "instant"` is also
+# accepted for near-realtime responses, but it is not a models.dev effort enum.
+# https://docs.inceptionlabs.ai/capabilities/instant
+family = "mercury"
+release_date = "2026-09-08"
+last_updated = "2026-09-10"
+attachment = false
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+temperature = true
+knowledge = "2025-11-01"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.04
+output = 0.15
+cache_read = 0.004
+
+[limit]
+context = 260_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 1bbe545ba23247d9b4ccc460cb66bbc33b262ebd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 17:23:36 +0000
Subject: [PATCH 1013/2012] chore(sync): update LLM Gateway model catalog
 (#6729)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aws-bedrock/claude-fable-5-1.toml  | 13 ++++++++++++
 .../consensusprotocol/deepseek-v4-flash.toml  |  1 -
 .../consensusprotocol/gemma-4-31b-it.toml     | 16 +++++++++++++++
 .../models/vichar-ai/glm-5.3-flash.toml       | 20 +++++++++++++++++++
 4 files changed, 49 insertions(+), 1 deletion(-)
 create mode 100644 providers/llmgateway-providers/models/aws-bedrock/claude-fable-5-1.toml
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/gemma-4-31b-it.toml
 create mode 100644 providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml

diff --git a/providers/llmgateway-providers/models/aws-bedrock/claude-fable-5-1.toml b/providers/llmgateway-providers/models/aws-bedrock/claude-fable-5-1.toml
new file mode 100644
index 00000000000..2d570728f40
--- /dev/null
+++ b/providers/llmgateway-providers/models/aws-bedrock/claude-fable-5-1.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-fable-5-1"
+name = "Claude Fable 5.1 (AWS Bedrock)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
index 38a7c0ad2ce..5e6b181be08 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -14,5 +14,4 @@ output = 0.1
 cache_read = 0.01
 
 [limit]
-context = 524_288
 output = 393_216
diff --git a/providers/llmgateway-providers/models/consensusprotocol/gemma-4-31b-it.toml b/providers/llmgateway-providers/models/consensusprotocol/gemma-4-31b-it.toml
new file mode 100644
index 00000000000..25e68298d12
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/gemma-4-31b-it.toml
@@ -0,0 +1,16 @@
+base_model = "google/gemma-4-31b-it"
+name = "Gemma 4 31B IT (Consensus Protocol)"
+attachment = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.25
+cache_read = 0.01
+
+[modalities]
+input = ["text"]
diff --git a/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml b/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..a8d0eb4cba1
--- /dev/null
+++ b/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml
@@ -0,0 +1,20 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (vichar-ai)"
+attachment = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_000
+output = 128_000
+
+[modalities]
+input = ["text"]

From 6bf69f2f7082f7eda3005cbde3bce230bc3591d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 17:23:42 +0000
Subject: [PATCH 1014/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6730)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gemma-4-31b-it.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/gemma-4-31b-it.toml b/providers/llmgateway/models/gemma-4-31b-it.toml
index ade0ba6aa27..b1c4fc2e42e 100644
--- a/providers/llmgateway/models/gemma-4-31b-it.toml
+++ b/providers/llmgateway/models/gemma-4-31b-it.toml
@@ -5,6 +5,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.102
-output = 0.297
-cache_read = 0.012
+input = 0.1
+output = 0.25
+cache_read = 0.01

From 80dc48161641bd20d0abf14b812c34154669cc47 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 17:23:44 +0000
Subject: [PATCH 1015/2012] chore(sync): update OpenRouter model catalog
 (#6731)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inclusionai/ling-3.0-flash-vl:free.toml   | 28 +++++++++++++++++++
 .../models/minimax/minimax-m2.5.toml          |  9 ++++--
 .../models/qwen/qwen3.6-35b-a3b.toml          |  7 +++--
 3 files changed, 38 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml

diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml
new file mode 100644
index 00000000000..b0955907412
--- /dev/null
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml
@@ -0,0 +1,28 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ling 3.0 Flash VL (free)"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "ling"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index 1edeb5ecdfc..ca75f8e4927 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,6 +6,9 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.03
+input = 0.27
+output = 1.08
+cache_read = 0.027
+
+[limit]
+output = 128_000
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index f33ff42c33e..ac95b282785 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,11 +6,12 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.05
-output = 0.7
+input = 0.1
+output = 0.9
+cache_read = 0.05
 
 [limit]
-output = 32_768
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From 4ba084c9c4496e894816f24cee7a52ad077ef64c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 17:23:58 +0000
Subject: [PATCH 1016/2012] chore(sync): update Merge Gateway model catalog
 (#6732)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 3 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index b86a0e589cd..5ef98e7a3a0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index d7d8d42decc..f0885b3822b 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +10,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 84592d1b8dd..0264c6bc733 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 5446186e3fe56d04fbab1e72012673c547674716 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 17:24:51 +0000
Subject: [PATCH 1017/2012] chore(sync): update Kilo model catalog (#6734)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inclusionai/ling-3.0-flash-vl:free.toml   | 27 +++++++++++++++++++
 .../kilo/models/minimax/minimax-m2.5.toml     |  4 +++
 .../kilo/models/qwen/qwen3.6-35b-a3b.toml     |  7 ++---
 3 files changed, 35 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml
new file mode 100644
index 00000000000..25f5cdff910
--- /dev/null
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml
@@ -0,0 +1,27 @@
+name = "inclusionAI: Ling 3.0 Flash VL (free)"
+description = "Ling 3.0 Flash VL builds on Ling 3.0 Flash (124B total / 5.5B active MoE from InclusionAI), further strengthening its language capabilities while adding native visual perception and advanced visual..."
+family = "ling"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index 3f15c0bc21d..e89b49a7fbe 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,3 +10,7 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
+
+[limit]
+context = 200_000
+output = 128_000
diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index 19df61b6e3c..1385363812a 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,11 +6,12 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.05
-output = 0.7
+input = 0.1
+output = 0.9
+cache_read = 0.05
 
 [limit]
-output = 32_768
+output = 235_929
 
 [modalities]
 input = ["text", "image", "video"]

From 2858118c7109c87cdaaeeb6db35d8e6a10f3ea46 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 11 Sep 2026 01:40:19 +0800
Subject: [PATCH 1018/2012] fix(opencode-go): rename DeepSeek V4.1 Flash model
 ID and deprecate old ID

---
 providers/opencode-go/models/deepseek-flash.toml   |  1 +
 .../opencode-go/models/deepseek-v4.1-flash.toml    | 14 ++++++++++++++
 2 files changed, 15 insertions(+)
 create mode 100644 providers/opencode-go/models/deepseek-v4.1-flash.toml

diff --git a/providers/opencode-go/models/deepseek-flash.toml b/providers/opencode-go/models/deepseek-flash.toml
index 21a954201c4..74e51dbe421 100644
--- a/providers/opencode-go/models/deepseek-flash.toml
+++ b/providers/opencode-go/models/deepseek-flash.toml
@@ -1,5 +1,6 @@
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4.1 Flash"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/opencode-go/models/deepseek-v4.1-flash.toml b/providers/opencode-go/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..21a954201c4
--- /dev/null
+++ b/providers/opencode-go/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003

From 3e5c1b60ba755356965060adba5a9bc2a98e93b1 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 11 Sep 2026 01:58:42 +0800
Subject: [PATCH 1019/2012] chore(opencode-go): remove deprecated
 deepseek-flash model entry

---
 providers/opencode-go/models/deepseek-flash.toml | 15 ---------------
 1 file changed, 15 deletions(-)
 delete mode 100644 providers/opencode-go/models/deepseek-flash.toml

diff --git a/providers/opencode-go/models/deepseek-flash.toml b/providers/opencode-go/models/deepseek-flash.toml
deleted file mode 100644
index 74e51dbe421..00000000000
--- a/providers/opencode-go/models/deepseek-flash.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "deepseek/deepseek-v4.1-flash"
-name = "DeepSeek V4.1 Flash"
-status = "deprecated"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003

From 5308729a3d4c3730fcd0351832f55c5b4d7ba240 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 18:28:55 +0000
Subject: [PATCH 1020/2012] chore(sync): update NanoGPT model catalog (#6740)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/TEE/gpt-oss-20b.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/TEE/gpt-oss-20b.toml b/providers/nano-gpt/models/TEE/gpt-oss-20b.toml
index 5dbb5729808..667c4defe72 100644
--- a/providers/nano-gpt/models/TEE/gpt-oss-20b.toml
+++ b/providers/nano-gpt/models/TEE/gpt-oss-20b.toml
@@ -5,9 +5,9 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.2
-output = 0.8
-cache_read = 0.1
+input = 0.04
+output = 0.15
+cache_read = 0.02
 
 [limit]
 input = 131_072

From 62d58b5fcbbb7a77eab4b394bfda1a6670cbabbc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 18:29:19 +0000
Subject: [PATCH 1021/2012] chore(sync): update LLM Gateway model catalog
 (#6739)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/vichar-ai/glm-5.3.toml                | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml

diff --git a/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml b/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml
new file mode 100644
index 00000000000..561369dc9e2
--- /dev/null
+++ b/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml
@@ -0,0 +1,16 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (vichar-ai)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_000
+output = 128_000

From a2aa746f2b706d0e04882d6d5b7edac3e349d9f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:36:27 -0500
Subject: [PATCH 1022/2012] fix: [missing-model] ofox: qwen/qwen3.8-max-0902
 (#6728)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/qwen/qwen3.8-max-0902.toml    | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.8-max-0902.toml

diff --git a/providers/ofox/models/qwen/qwen3.8-max-0902.toml b/providers/ofox/models/qwen/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..50d7d8338e3
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.8-max-0902.toml
@@ -0,0 +1,21 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.8-max-0902
+# https://api.ofox.ai/v1/models/qwen/qwen3.8-max-0902
+# https://www.qwencloud.com/models/qwen3.8-max-0902
+# https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Toggle: enable_thinking true|false (hybrid; Qwen/DashScope surface through Ofox)
+# Effort: reasoning_effort = low|medium|xhigh (default xhigh); high accepted as alias → xhigh
+# Budget: thinking_budget (0..262144) cannot be combined with reasoning_effort
+base_model = "alibaba/qwen3.8-max-0902"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", min = 0, max = 262_144 },
+]
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.25
+cache_write = 2.5

From 13e4fd2bd0b9d4346f5154a8d48139d17c540f87 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:36:43 -0500
Subject: [PATCH 1023/2012] fix: [missing-model] ofox: qwen/qwen3.8-max (#6727)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.8-max.toml | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.8-max.toml

diff --git a/providers/ofox/models/qwen/qwen3.8-max.toml b/providers/ofox/models/qwen/qwen3.8-max.toml
new file mode 100644
index 00000000000..733f1266841
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.8-max.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.8-max
+# https://api.ofox.ai/v1/models/qwen/qwen3.8-max
+# https://www.qwencloud.com/models/qwen3.8-max
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Toggle: enable_thinking true|false (hybrid; Qwen/DashScope surface through Ofox)
+# Effort: reasoning_effort = low|medium|xhigh (default xhigh); high accepted as alias → xhigh
+# Budget: thinking_budget (0..262144) cannot be combined with reasoning_effort
+base_model = "alibaba/qwen3.8-max"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "xhigh"] },
+  { type = "budget_tokens", min = 0, max = 262_144 },
+]
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.25
+cache_write = 2.5

From 882725164dc883bbe188cd682d000c1c6d70e539 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:36:52 -0500
Subject: [PATCH 1024/2012] fix: [missing-model] ofox: qwen/qwen3.8-flash
 (#6726)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.8-flash.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.8-flash.toml

diff --git a/providers/ofox/models/qwen/qwen3.8-flash.toml b/providers/ofox/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..9e1592e3e53
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,15 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.8-flash
+# https://api.ofox.ai/v1/models/qwen/qwen3.8-flash
+# https://www.qwencloud.com/models/qwen3.8-flash
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Toggle: enable_thinking true|false (hybrid; Qwen/DashScope surface through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From ac1cc10309e8dc769e0b17b539d8b53b7b082b52 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:36:57 -0500
Subject: [PATCH 1025/2012] fix: [missing-model] ofox: qwen/qwen3.8-27b (#6736)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.8-27b.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.8-27b.toml

diff --git a/providers/ofox/models/qwen/qwen3.8-27b.toml b/providers/ofox/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..ea631083ca0
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.8-27b
+# https://api.ofox.ai/v1/models/qwen/qwen3.8-27b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.8-27b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.45
+output = 3.2
+cache_read = 0.05
+cache_write = 0.5625
+
+[limit]
+context = 1_131_072
+output = 131_072

From ac86bfbf1fc719a7c1391e61eb481a35f91e14fe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:37:06 -0500
Subject: [PATCH 1026/2012] fix: [missing-model] ofox: qwen/qwen3.7-max (#6737)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.7-max.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.7-max.toml

diff --git a/providers/ofox/models/qwen/qwen3.7-max.toml b/providers/ofox/models/qwen/qwen3.7-max.toml
new file mode 100644
index 00000000000..c071652b2c2
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.7-max.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.7-max
+# https://api.ofox.ai/v1/models/qwen/qwen3.7-max
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.7-max"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 2.5
+output = 7.5
+cache_read = 0.5
+cache_write = 3.125
+
+[limit]
+context = 1_064_000
+output = 64_000

From 780c1d7c8f35ca328a4ea8792978b21bae945b16 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 13:37:22 -0500
Subject: [PATCH 1027/2012] fix: [missing-model] ofox: qwen/qwen3.7-plus
 (#6738)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.7-plus.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.7-plus.toml

diff --git a/providers/ofox/models/qwen/qwen3.7-plus.toml b/providers/ofox/models/qwen/qwen3.7-plus.toml
new file mode 100644
index 00000000000..88cbf529408
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.7-plus.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.7-plus
+# https://api.ofox.ai/v1/models/qwen/qwen3.7-plus
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.7-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.4
+output = 1.6
+cache_read = 0.08
+cache_write = 0.5
+
+[limit]
+context = 1_064_000

From 0d423d3ce011ef73b7f391d25560dd5733742098 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 14:19:43 -0500
Subject: [PATCH 1028/2012] fix: [missing-model] ofox: qwen/qwen3.6-plus
 (#6747)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.6-plus.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.6-plus.toml

diff --git a/providers/ofox/models/qwen/qwen3.6-plus.toml b/providers/ofox/models/qwen/qwen3.6-plus.toml
new file mode 100644
index 00000000000..7c0f1c3277e
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.6-plus.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.6-plus
+# https://api.ofox.ai/v1/models/qwen/qwen3.6-plus
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.6-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.5
+output = 3
+cache_read = 0.05
+cache_write = 0.625
+
+[limit]
+output = 64_000

From fb2a44361b9b9a4a731c8036dcc24265516fcb63 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 19:21:15 +0000
Subject: [PATCH 1029/2012] chore(sync): update OpenRouter model catalog
 (#6748)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576

From 8c27c3036623be59cac300300f179a9416e2044d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 19:21:46 +0000
Subject: [PATCH 1030/2012] fix: [missing-model] ofox: qwen/qwen3.6-max-preview
 (#6745)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/qwen/qwen3.6-max-preview.toml   | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.6-max-preview.toml

diff --git a/providers/ofox/models/qwen/qwen3.6-max-preview.toml b/providers/ofox/models/qwen/qwen3.6-max-preview.toml
new file mode 100644
index 00000000000..30c5aff16f1
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.6-max-preview.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.6-max-preview
+# https://api.ofox.ai/v1/models/qwen/qwen3.6-max-preview
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.6-max-preview"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 2.15
+output = 12.86
+cache_read = 0.2
+cache_write = 1.17
+
+[limit]
+context = 256_000
+output = 64_000

From 7523facd85e3a1a890c860237ce429d0bd9fb11e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 19:22:00 +0000
Subject: [PATCH 1031/2012] fix: [missing-model] ofox: qwen/qwen3.6-flash
 (#6746)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.6-flash.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.6-flash.toml

diff --git a/providers/ofox/models/qwen/qwen3.6-flash.toml b/providers/ofox/models/qwen/qwen3.6-flash.toml
new file mode 100644
index 00000000000..ac1359ad408
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.6-flash.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.6-flash
+# https://api.ofox.ai/v1/models/qwen/qwen3.6-flash
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.6-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.31
+
+[limit]
+output = 64_000

From 07c0a57168026f97e621d39201943265b53e4c7b Mon Sep 17 00:00:00 2001
From: Ahmad Shahzad <107808273+shzdehmd@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:31:50 +0500
Subject: [PATCH 1032/2012] feat(fireworks-ai): add DeepSeek V4.1 Flash (#6741)

Add serverless provider entry for accounts/fireworks/models/deepseek-v4p1-flash,
released 2026-09-10. Override-only on deepseek/deepseek-v4.1-flash lab model.

- Reasoning: toggle (thinking.type / reasoning_effort=none) + effort
  (low, high, max). medium promotes to high; low is distinct (not promoted
  despite docs). xhigh and max are distinct tiers (not promoted despite docs).
  Verified across 4 prompts at temperature 0.
- Vision: confirmed working with image input (base64 + URL).
- Pricing: $0.22/$0.007/$0.66 per 1M (input/cached/output).

Co-authored-by: shzdehmd <shzdehmd@users.noreply.github.com>
---
 .../fireworks/models/deepseek-v4p1-flash.toml | 25 +++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml
new file mode 100644
index 00000000000..d6e8a5fa391
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml
@@ -0,0 +1,25 @@
+# Toggle: thinking.type = enabled|disabled (reasoning_effort = "none" or false also disables)
+# Effort: reasoning_effort = low|medium|high|xhigh|max
+#   medium and high produce identical output (medium promoted to high).
+#   low is a distinct tier, not promoted to high despite docs.
+#   xhigh and max are distinct tiers, not promoted despite docs.
+#   Verified 2026-09-10 across 4 prompts at temperature 0.
+# Reasoning: https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-09-10)
+# Pricing: https://fireworks.ai/models/deepseek-ai/deepseek-v4p1-flash (accessed 2026-09-10)
+# Released on Fireworks Serverless API 2026-09-10.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007

From 668cddb01214a83b3786b7926a99f81d52d82573 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 20:24:13 +0000
Subject: [PATCH 1033/2012] chore(sync): update Charm Hyper model catalog
 (#6755)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml              | 6 +++---
 providers/hyper/models/glm-5.1.toml                         | 6 +++---
 providers/hyper/models/glm-5.toml                           | 2 +-
 providers/hyper/models/gpt-oss-120b.toml                    | 6 +++---
 providers/hyper/models/kimi-k2-thinking.toml                | 2 +-
 providers/hyper/models/kimi-k2.5.toml                       | 2 +-
 providers/hyper/models/llama-3.3-70b-instruct.toml          | 2 +-
 .../models/llama-4-maverick-17b-128e-instruct-fp8.toml      | 2 +-
 providers/hyper/models/minimax-m2.7.toml                    | 6 +++---
 .../qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml       | 2 +-
 providers/hyper/models/qwen3-next-80b-a3b-instruct.toml     | 2 +-
 11 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index c98d649e236..fb39f3786c1 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.106
-output = 0.368
-cache_write = 0.053
+input = 0.122
+output = 0.42
+cache_read = 0.061
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index befdc41c51e..24bf7b877f4 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.386
-output = 4.356
-cache_write = 0.693
+input = 1.326
+output = 4.22
+cache_read = 0.663
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 2be85ce1855..108e0f80af8 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -6,7 +6,7 @@ reasoning_options = []
 [cost]
 input = 0.86
 output = 2.784
-cache_write = 0.43
+cache_read = 0.43
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index d7c068dc409..22c22acac3a 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.18
-output = 0.61
-cache_write = 0.09
+input = 0.178
+output = 0.68
+cache_read = 0.089
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/kimi-k2-thinking.toml b/providers/hyper/models/kimi-k2-thinking.toml
index cb77c77068c..03488a06cac 100644
--- a/providers/hyper/models/kimi-k2-thinking.toml
+++ b/providers/hyper/models/kimi-k2-thinking.toml
@@ -6,7 +6,7 @@ reasoning_options = []
 [cost]
 input = 0.6
 output = 2.5
-cache_write = 0.3
+cache_read = 0.3
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index b8552721ba3..59545971994 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -7,7 +7,7 @@ reasoning_options = []
 [cost]
 input = 0.5584
 output = 2.935
-cache_write = 0.2792
+cache_read = 0.2792
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/llama-3.3-70b-instruct.toml b/providers/hyper/models/llama-3.3-70b-instruct.toml
index 7ca7f9333ed..bb98c102871 100644
--- a/providers/hyper/models/llama-3.3-70b-instruct.toml
+++ b/providers/hyper/models/llama-3.3-70b-instruct.toml
@@ -6,7 +6,7 @@ attachment = false
 [cost]
 input = 0.6066
 output = 1.0386
-cache_write = 0.3033
+cache_read = 0.3033
 
 [limit]
 output = 12_800
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
index 9b588de13d6..10568630e53 100644
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -6,7 +6,7 @@ attachment = false
 [cost]
 input = 0.274
 output = 0.8992
-cache_write = 0.137
+cache_read = 0.137
 
 [limit]
 context = 430_000
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 6ec098e5a6b..4a47482e34e 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.458
-output = 1.712
-cache_write = 0.229
+input = 0.404
+output = 1.496
+cache_read = 0.202
 
 [limit]
 context = 262_100
diff --git a/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml b/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
index 3f34d120a9b..12c7d4b7413 100644
--- a/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
+++ b/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
@@ -5,7 +5,7 @@ last_updated = "2026-07-22"
 [cost]
 input = 0.445
 output = 2.145
-cache_write = 0.2225
+cache_read = 0.2225
 
 [limit]
 context = 106_000
diff --git a/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml b/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
index fd9d9ca3d08..5cfa8ce4d56 100644
--- a/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
@@ -5,7 +5,7 @@ last_updated = "2026-07-22"
 [cost]
 input = 0.1175
 output = 1.136
-cache_write = 0.05875
+cache_read = 0.05875
 
 [limit]
 context = 262_144

From 277ca4449712078468e0d46274c2afe20578bac0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 20:24:24 +0000
Subject: [PATCH 1034/2012] chore(sync): update OpenRouter model catalog
 (#6756)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576

From 598b6c7ffd8a2afb022a3cb31ef5506db470bf38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 20:24:47 +0000
Subject: [PATCH 1035/2012] chore(sync): update Merge Gateway model catalog
 (#6758)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 5ef98e7a3a0..b86a0e589cd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +11,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 74989687cca80cd436c1e105f0dc79eb32516024 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 20:25:05 +0000
Subject: [PATCH 1036/2012] chore(sync): update Venice model catalog (#6757)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/deepseek-v4-pro.toml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/providers/venice/models/deepseek-v4-pro.toml b/providers/venice/models/deepseek-v4-pro.toml
index edc4ced1b96..8f02eb73ba9 100644
--- a/providers/venice/models/deepseek-v4-pro.toml
+++ b/providers/venice/models/deepseek-v4-pro.toml
@@ -1,11 +1,14 @@
 base_model = "deepseek/deepseek-v4-pro"
 family = "deepseek"
 last_updated = "2026-06-11"
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
 [cost]
 input = 1.65
 output = 3.301

From 238640ec00fa85d4f534b7bda1521b0b23102d1b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:18:17 +0000
Subject: [PATCH 1037/2012] fix: [missing-model] ofox: qwen/qwen3.6-27b (#6754)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.6-27b.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.6-27b.toml

diff --git a/providers/ofox/models/qwen/qwen3.6-27b.toml b/providers/ofox/models/qwen/qwen3.6-27b.toml
new file mode 100644
index 00000000000..b2ff03427c1
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.6-27b.toml
@@ -0,0 +1,15 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.6-27b
+# https://api.ofox.ai/v1/models/qwen/qwen3.6-27b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.6-27b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.6
+output = 3.6
+
+[limit]
+context = 256_000
+output = 64_000

From 4a85d7fe07bc62425d38ff06a98b7b2ee50d81c4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:18:21 +0000
Subject: [PATCH 1038/2012] fix: [missing-model] ofox: qwen/qwen3.5-plus
 (#6753)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.5-plus.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-plus.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-plus.toml b/providers/ofox/models/qwen/qwen3.5-plus.toml
new file mode 100644
index 00000000000..7badf80bbb0
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-plus.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-plus
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-plus
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-plus"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.4
+output = 2.4
+cache_read = 0.04
+cache_write = 0.4
+
+[limit]
+output = 64_000

From 4e701aedac0f3c885b917aaf83352b0219af0cca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:18:26 +0000
Subject: [PATCH 1039/2012] fix: [missing-model] ofox: qwen/qwen3.5-flash
 (#6752)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.5-flash.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-flash.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-flash.toml b/providers/ofox/models/qwen/qwen3.5-flash.toml
new file mode 100644
index 00000000000..dd4205c51ac
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-flash.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-flash
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-flash
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.1
+output = 0.4
+cache_read = 0.01
+cache_write = 0.125
+
+[limit]
+output = 64_000

From 3bae856e95adf23d99c1a0859133ec0be2aaf986 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:21:49 +0000
Subject: [PATCH 1040/2012] chore(sync): update Baseten model catalog (#6763)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/baseten/models/zai-org/GLM-5.3.toml | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/providers/baseten/models/zai-org/GLM-5.3.toml b/providers/baseten/models/zai-org/GLM-5.3.toml
index 3e11707249d..3a5a5ce1e12 100644
--- a/providers/baseten/models/zai-org/GLM-5.3.toml
+++ b/providers/baseten/models/zai-org/GLM-5.3.toml
@@ -4,14 +4,15 @@
 base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+attachment = true
+
+[interleaved]
+field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 1.4
 output = 4.4
@@ -20,3 +21,6 @@ cache_read = 0.14
 [limit]
 context = 1_048_576
 output = 262_144
+
+[modalities]
+input = ["text", "image"]

From e6a01a24a0e0257a9c21fd72084af5edb29933e4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:18:18 +0000
Subject: [PATCH 1041/2012] fix: [missing-model] ofox: qwen/qwen3.5-397b-a17b
 (#6762)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/qwen/qwen3.5-397b-a17b.toml   | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-397b-a17b.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-397b-a17b.toml b/providers/ofox/models/qwen/qwen3.5-397b-a17b.toml
new file mode 100644
index 00000000000..f08f155e01c
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-397b-a17b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-397b-a17b
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-397b-a17b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-397b-a17b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.55
+output = 3.5
+cache_read = 0.55
+
+[limit]
+context = 256_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From faa7ae16e6f9bba4ca2515516e13dc207f7e1373 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:18:22 +0000
Subject: [PATCH 1042/2012] fix: [missing-model] ofox: qwen/qwen3.5-35b-a3b
 (#6761)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/qwen/qwen3.5-35b-a3b.toml     | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-35b-a3b.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-35b-a3b.toml b/providers/ofox/models/qwen/qwen3.5-35b-a3b.toml
new file mode 100644
index 00000000000..fe7310792ff
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-35b-a3b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-35b-a3b
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-35b-a3b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-35b-a3b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.29
+output = 1.83
+cache_read = 0.29
+
+[limit]
+context = 256_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From 3c71e80d8be033bb678a71a2e53f57db14653140 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:18:26 +0000
Subject: [PATCH 1043/2012] fix: [missing-model] ofox: qwen/qwen3.5-27b (#6760)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3.5-27b.toml | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-27b.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-27b.toml b/providers/ofox/models/qwen/qwen3.5-27b.toml
new file mode 100644
index 00000000000..ddf63942c12
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-27b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-27b
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-27b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-27b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.29
+output = 2.05
+cache_read = 0.29
+
+[limit]
+context = 256_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From 7c3a8716d9b8c037a98921f48aa934a445a0ae9a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:23:59 +0000
Subject: [PATCH 1044/2012] chore(sync): update OpenRouter model catalog
 (#6767)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576

From 8b066fe45361d3152a71d2a0717bd779643d162d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:24:31 +0000
Subject: [PATCH 1045/2012] chore(sync): update Deep Infra model catalog
 (#6768)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/MiniMaxAI/MiniMax-M2.7.toml | 1 +
 providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml  | 1 +
 providers/deepinfra/models/zai-org/GLM-5.toml          | 1 +
 3 files changed, 3 insertions(+)

diff --git a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.7.toml b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.7.toml
index 6bd61f181c5..7e2b1f39898 100644
--- a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.7.toml
+++ b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.7.toml
@@ -1,4 +1,5 @@
 base_model = "minimax/MiniMax-M2.7"
+status = "deprecated"
 reasoning_options = []
 
 [cost]
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
index 42f815d17d3..fcd0b9ff7de 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
@@ -10,6 +10,7 @@ tool_call = true
 structured_output = true
 knowledge = "2025-04"
 open_weights = true
+status = "deprecated"
 reasoning_options = []
 
 [interleaved]
diff --git a/providers/deepinfra/models/zai-org/GLM-5.toml b/providers/deepinfra/models/zai-org/GLM-5.toml
index e55f3cd2a3d..3d97a980667 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.toml
@@ -10,6 +10,7 @@ tool_call = true
 structured_output = true
 knowledge = "2025-12"
 open_weights = true
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"

From 903c339354bdd7293f58ed806803fd1e05441217 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 22:24:33 +0000
Subject: [PATCH 1046/2012] chore(sync): update Charm Hyper model catalog
 (#6770)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.3.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/hyper/models/glm-5.3.toml b/providers/hyper/models/glm-5.3.toml
index bc974016adc..98852073c09 100644
--- a/providers/hyper/models/glm-5.3.toml
+++ b/providers/hyper/models/glm-5.3.toml
@@ -1,6 +1,7 @@
 base_model = "zhipuai/glm-5.3"
 release_date = "2026-08-28"
 last_updated = "2026-08-31"
+attachment = true
 
 [[reasoning_options]]
 type = "effort"
@@ -14,3 +15,6 @@ cache_read = 0.283088
 [limit]
 context = 1_048_576
 output = 262_144
+
+[modalities]
+input = ["text", "image"]

From 27abf43c3440f41f418e38a4e80bd7bb472d3ff2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:18:18 +0000
Subject: [PATCH 1047/2012] fix: [missing-model] ofox: qwen/qwen3.5-122b-a10b
 (#6766)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/qwen/qwen3.5-122b-a10b.toml   | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3.5-122b-a10b.toml

diff --git a/providers/ofox/models/qwen/qwen3.5-122b-a10b.toml b/providers/ofox/models/qwen/qwen3.5-122b-a10b.toml
new file mode 100644
index 00000000000..6609c1c51b1
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3.5-122b-a10b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3.5-122b-a10b
+# https://api.ofox.ai/v1/models/qwen/qwen3.5-122b-a10b
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.5-122b-a10b"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.29
+output = 2.29
+cache_read = 0.29
+
+[limit]
+context = 256_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]

From ee066124db268f51e77c1303bee37b2d5dcda838 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:18:22 +0000
Subject: [PATCH 1048/2012] fix: [missing-model] ofox: qwen/qwen3-max (#6765)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3-max.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3-max.toml

diff --git a/providers/ofox/models/qwen/qwen3-max.toml b/providers/ofox/models/qwen/qwen3-max.toml
new file mode 100644
index 00000000000..36aacd179c5
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3-max.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3-max
+# https://api.ofox.ai/v1/models/qwen/qwen3-max
+# https://www.alibabacloud.com/help/en/model-studio/deep-thinking
+# Toggle: enable_thinking true|false (hybrid; default off)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3-max"
+reasoning = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.36
+output = 1.43
+cache_read = 0.072
+
+[limit]
+context = 256_000
+output = 64_000

From 0a3569c593c5c91ccf51a5b37d837f308531cb3a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:18:26 +0000
Subject: [PATCH 1049/2012] fix: [missing-model] ofox: qwen/qwen3-coder-plus
 (#6764)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3-coder-plus.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3-coder-plus.toml

diff --git a/providers/ofox/models/qwen/qwen3-coder-plus.toml b/providers/ofox/models/qwen/qwen3-coder-plus.toml
new file mode 100644
index 00000000000..a787d74d0e5
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3-coder-plus.toml
@@ -0,0 +1,14 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3-coder-plus
+# https://api.ofox.ai/v1/models/qwen/qwen3-coder-plus
+base_model = "alibaba/qwen3-coder-plus"
+
+[cost]
+input = 1.8
+output = 9
+cache_read = 0.2
+cache_write = 1
+
+[limit]
+context = 1_000_000
+output = 64_000

From 91972d81a48631c6ede9932faaae253745c124ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:22:02 +0000
Subject: [PATCH 1050/2012] chore(sync): update Charm Hyper model catalog
 (#6774)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4.1-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/hyper/models/deepseek-v4.1-flash.toml

diff --git a/providers/hyper/models/deepseek-v4.1-flash.toml b/providers/hyper/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..f98b45365b0
--- /dev/null
+++ b/providers/hyper/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "xhigh"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 26_214

From e743d20190045aefe29249841a7c3f3c8983d902 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:22:13 +0000
Subject: [PATCH 1051/2012] chore(sync): update OpenRouter model catalog
 (#6775)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576

From 61fe67c6efeeaad10630b1181442705c59c2f5ec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:18:18 +0000
Subject: [PATCH 1052/2012] fix: [missing-model] ofox: qwen/qwen3-coder-next
 (#6772)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3-coder-next.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3-coder-next.toml

diff --git a/providers/ofox/models/qwen/qwen3-coder-next.toml b/providers/ofox/models/qwen/qwen3-coder-next.toml
new file mode 100644
index 00000000000..826c97c4f25
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3-coder-next.toml
@@ -0,0 +1,12 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3-coder-next
+# https://api.ofox.ai/v1/models/qwen/qwen3-coder-next
+base_model = "alibaba/qwen3-coder-next"
+
+[cost]
+input = 0.2
+output = 1.5
+
+[limit]
+context = 256_000
+output = 64_000

From 916bce4cd6b6d07b77410c9fa63f6233d3de8d84 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:18:23 +0000
Subject: [PATCH 1053/2012] fix: [missing-model] ofox: qwen/qwen3-coder-flash
 (#6771)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen3-coder-flash.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen3-coder-flash.toml

diff --git a/providers/ofox/models/qwen/qwen3-coder-flash.toml b/providers/ofox/models/qwen/qwen3-coder-flash.toml
new file mode 100644
index 00000000000..75b3885adaa
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen3-coder-flash.toml
@@ -0,0 +1,13 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen3-coder-flash
+# https://api.ofox.ai/v1/models/qwen/qwen3-coder-flash
+base_model = "alibaba/qwen3-coder-flash"
+
+[cost]
+input = 0.5
+output = 2.5
+cache_read = 0.06
+cache_write = 0.27
+
+[limit]
+output = 64_000

From 2eafd4c71fa2db24ec3d895ece1942dc5bb3102d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:18:28 +0000
Subject: [PATCH 1054/2012] fix: [missing-model] ofox: qwen/qwen-vl-max (#6773)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen-vl-max.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen-vl-max.toml

diff --git a/providers/ofox/models/qwen/qwen-vl-max.toml b/providers/ofox/models/qwen/qwen-vl-max.toml
new file mode 100644
index 00000000000..0692c38b4b0
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen-vl-max.toml
@@ -0,0 +1,13 @@
+# Sources (accessed 2026-09-10):
+# https://ofox.ai/models/qwen/qwen-vl-max
+# https://api.ofox.ai/v1/models/qwen/qwen-vl-max
+base_model = "alibaba/qwen-vl-max"
+
+[cost]
+input = 0.23
+output = 0.58
+cache_read = 0.046
+
+[limit]
+context = 128_000
+output = 8_000

From b1134a6d51c5375991efb25cb5725b20424c2187 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:57:47 +0000
Subject: [PATCH 1055/2012] chore(sync): update OpenRouter model catalog
 (#6782)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 7 +++----
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml        | 6 +++---
 providers/openrouter/models/minimax/minimax-m2.5.toml    | 9 +++------
 providers/openrouter/models/moonshotai/kimi-k3.toml      | 6 +++---
 .../models/qwen/qwen3-30b-a3b-instruct-2507.toml         | 6 +++---
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/upstage/solar-pro4.toml      | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml            | 7 +++----
 providers/openrouter/models/~z-ai/glm-latest.toml        | 8 ++++----
 11 files changed, 34 insertions(+), 39 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..5bc4e6af877 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 1.0494
+output = 3.1482
+cache_read = 0.03498
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a010dd38ba2..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.0725
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..1352c1ee599 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.4
-output = 0.4
+input = 0.72
+output = 0.72
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index ca75f8e4927..1edeb5ecdfc 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,9 +6,6 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.27
-output = 1.08
-cache_read = 0.027
-
-[limit]
-output = 128_000
+input = 0.3
+output = 1.2
+cache_read = 0.03
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index c11eb6c3dae..bc1e985b845 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.6
+output = 13
+cache_read = 0.29
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 2ce7e2e4c07..c0c2c8b5cd7 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.04815
-output = 0.19305
+input = 0.09
+output = 0.3
 
 [limit]
 context = 262_144
-output = 32_000
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/upstage/solar-pro4.toml b/providers/openrouter/models/upstage/solar-pro4.toml
index b88e5ba9776..adc9761fb07 100644
--- a/providers/openrouter/models/upstage/solar-pro4.toml
+++ b/providers/openrouter/models/upstage/solar-pro4.toml
@@ -16,9 +16,9 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.03
-output = 0.12
-cache_read = 0.006
+input = 0.09
+output = 0.36
+cache_read = 0.018
 
 [limit]
 context = 524_288
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index c38d06275a4..64652242a0e 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.28
-output = 0.88
-cache_read = 0.052
+input = 0.966
+output = 3.036
+cache_read = 0.1932
 
 [limit]
 context = 1_048_576
-output = 128_000
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 2b14e7625ce..0818573f975 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 1
+output = 3.41
+cache_read = 0.2
 
 [limit]
 context = 1_310_720
-output = 128_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 2e58761572d44f27a413831433193a2183f7fd41 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:57:50 +0000
Subject: [PATCH 1056/2012] chore(sync): update Deep Infra model catalog
 (#6780)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml     | 6 +++---
 .../models/deepseek-ai/DeepSeek-V4.1-Flash.toml          | 9 ++++++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
index 0d7f8576c32..83d3a4c6ca2 100644
--- a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
+++ b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
@@ -8,9 +8,9 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 0.4
-output = 2
-cache_read = 0.08
+input = 0.14
+output = 0.28
+cache_read = 0.0028
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
index 719c681bfaf..3c4cd385308 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -2,11 +2,14 @@
 # https://api.deepinfra.com/models/deepseek-ai/DeepSeek-V4.1-Flash/schema/openai-chat-completions
 # https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4.1-Flash (accessed 2026-09-10)
 base_model = "deepseek/deepseek-v4.1-flash"
-reasoning_options = [{ type = "effort", values = ["none", "low", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "xhigh", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
+input = 0.2
+output = 0.6
 cache_read = 0.006
 
 [limit]

From 5a23a5a72b09f39fccbe54ef2db5b38158bcbd68 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:57:52 +0000
Subject: [PATCH 1057/2012] chore(sync): update Merge Gateway model catalog
 (#6781)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../merge-gateway/models/deepseek/deepseek-v4-flash.toml   | 7 +++++--
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
index 59726ca8398..5644b59b802 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash.toml
@@ -9,6 +9,9 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 0.22
-output = 0.66
+input = 0.035
+output = 0.07
 cache_read = 0.007
+
+[limit]
+context = 1_048_576
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index b86a0e589cd..5ef98e7a3a0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 901146256a8828a2789910659456c849d4064f2e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 00:58:20 +0000
Subject: [PATCH 1058/2012] chore(sync): update Kilo model catalog (#6783)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 .../kilo/models/meta-llama/llama-3.1-70b-instruct.toml |  2 +-
 providers/kilo/models/minimax/minimax-m2.5.toml        |  4 ----
 .../kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml  |  4 ++--
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                |  4 ----
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 7 files changed, 12 insertions(+), 21 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..6fc67901197 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index e89b49a7fbe..3f15c0bc21d 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,7 +10,3 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
-
-[limit]
-context = 200_000
-output = 128_000
diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 48baf441187..1da6f655c94 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 128_000
-output = 32_000
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index dcb3cc80d7f..521ea40a372 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -9,7 +9,3 @@ values = ["none", "high", "xhigh"]
 input = 1.4
 output = 4.4
 cache_read = 0.26
-
-[limit]
-context = 1_024_000
-output = 128_000
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 114538a3af7..6e543d6cf2b 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 1
+output = 3.41
+cache_read = 0.2
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 0dbcdec03a87ad0cb476df1d5662bc136a1b496b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 01:18:19 +0000
Subject: [PATCH 1059/2012] fix: [missing-model] ofox: qwen/qwen-turbo (#6779)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen-turbo.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen-turbo.toml

diff --git a/providers/ofox/models/qwen/qwen-turbo.toml b/providers/ofox/models/qwen/qwen-turbo.toml
new file mode 100644
index 00000000000..af971283622
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen-turbo.toml
@@ -0,0 +1,14 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/qwen/qwen-turbo
+# https://api.ofox.ai/v1/models/qwen/qwen-turbo
+base_model = "alibaba/qwen-turbo"
+reasoning = false
+
+[cost]
+input = 0.05
+output = 0.09
+cache_read = 0.0086
+
+[limit]
+context = 128_000
+output = 16_000

From f068058eca5b2a8f92d46bd7b05c6bebdb36560c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 01:18:23 +0000
Subject: [PATCH 1060/2012] fix: [missing-model] ofox: qwen/qwen-plus (#6778)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen-plus.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen-plus.toml

diff --git a/providers/ofox/models/qwen/qwen-plus.toml b/providers/ofox/models/qwen/qwen-plus.toml
new file mode 100644
index 00000000000..37c1a0cbf95
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen-plus.toml
@@ -0,0 +1,13 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/qwen/qwen-plus
+# https://api.ofox.ai/v1/models/qwen/qwen-plus
+base_model = "alibaba/qwen-plus"
+reasoning = false
+
+[cost]
+input = 0.12
+output = 0.29
+cache_read = 0.023
+
+[limit]
+output = 32_000

From 1ae85fc0f427f0a40d58cac0108a8a8c5594a8a4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 01:18:27 +0000
Subject: [PATCH 1061/2012] fix: [missing-model] ofox: qwen/qwen-max (#6777)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen-max.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen-max.toml

diff --git a/providers/ofox/models/qwen/qwen-max.toml b/providers/ofox/models/qwen/qwen-max.toml
new file mode 100644
index 00000000000..44ad790b404
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen-max.toml
@@ -0,0 +1,13 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/qwen/qwen-max
+# https://api.ofox.ai/v1/models/qwen/qwen-max
+base_model = "alibaba/qwen-max"
+
+[cost]
+input = 0.35
+output = 1.38
+cache_read = 0.069
+
+[limit]
+context = 32_000
+output = 8_000

From 5462d2f3a58ab1799521c2545214ca95c6bedd24 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 01:33:50 +0000
Subject: [PATCH 1062/2012] chore(sync): update OpenRouter model catalog
 (#6788)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576

From db8755b4f0f54b6df409c08fb67545d837a577df Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 01:34:04 +0000
Subject: [PATCH 1063/2012] chore(sync): update NanoGPT model catalog (#6787)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 9c2641248c0..fb4e3a2dd26 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 2e7355d5588..f23aac53ab8 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From ee77699da79a10fbab0caa9ca62cbdab7c82ba8a Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:16:26 -0500
Subject: [PATCH 1064/2012] fix(amazon-bedrock): correct Nova catalog (#6793)

---
 models/amazon/nova-2-lite.toml                | 10 +++++--
 models/amazon/nova-lite.toml                  |  9 ++++--
 models/amazon/nova-micro.toml                 |  6 +++-
 models/amazon/nova-premier.toml               | 23 ++++++++++++++
 models/amazon/nova-pro.toml                   |  9 ++++--
 .../models/amazon.nova-2-lite-v1:0.toml       | 30 +++++++++----------
 .../models/amazon.nova-lite-v1:0.toml         | 25 ++++------------
 .../models/amazon.nova-micro-v1:0.toml        | 25 ++++------------
 .../models/amazon.nova-pro-v1:0.toml          | 25 ++++------------
 .../models/apac.amazon.nova-lite-v1:0.toml    |  9 ++++--
 .../models/apac.amazon.nova-micro-v1:0.toml   |  9 ++++--
 .../models/apac.amazon.nova-pro-v1:0.toml     |  9 ++++--
 .../models/ca.amazon.nova-lite-v1:0.toml      |  8 +++--
 .../models/eu.amazon.nova-2-lite-v1:0.toml    | 17 +++++++++--
 .../models/eu.amazon.nova-lite-v1:0.toml      |  8 +++--
 .../models/eu.amazon.nova-micro-v1:0.toml     |  8 +++--
 .../models/eu.amazon.nova-pro-v1:0.toml       |  8 +++--
 .../global.amazon.nova-2-lite-v1:0.toml       | 17 +++++++++--
 .../models/jp.amazon.nova-2-lite-v1:0.toml    | 17 +++++++++--
 .../models/us.amazon.nova-2-lite-v1:0.toml    | 17 +++++++++--
 .../models/us.amazon.nova-lite-v1:0.toml      |  8 +++--
 .../models/us.amazon.nova-micro-v1:0.toml     |  8 +++--
 .../models/us.amazon.nova-premier-v1:0.toml   | 16 ++++++++++
 .../models/us.amazon.nova-pro-v1:0.toml       |  8 +++--
 24 files changed, 213 insertions(+), 116 deletions(-)
 create mode 100644 models/amazon/nova-premier.toml
 create mode 100644 providers/amazon-bedrock/models/us.amazon.nova-premier-v1:0.toml

diff --git a/models/amazon/nova-2-lite.toml b/models/amazon/nova-2-lite.toml
index f818123c9c7..5d42c94dbb4 100644
--- a/models/amazon/nova-2-lite.toml
+++ b/models/amazon/nova-2-lite.toml
@@ -1,17 +1,21 @@
+# Sources: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-2-lite.html
+# Launch: https://aws.amazon.com/about-aws/whats-new/2025/12/nova-2-foundation-models-amazon-bedrock/
 name = "Nova 2 Lite"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "nova"
-release_date = "2025-12-01"
-last_updated = "2025-12-01"
+release_date = "2025-12-02"
+last_updated = "2025-12-02"
 attachment = true
 reasoning = true
 temperature = true
+knowledge = "2025-10"
 tool_call = true
 open_weights = false
 
 [limit]
 context = 1_000_000
-output = 64_000
+output = 65_536
 
 [modalities]
 input = ["text", "image", "video", "pdf"]
diff --git a/models/amazon/nova-lite.toml b/models/amazon/nova-lite.toml
index 5b14adbd695..db526061329 100644
--- a/models/amazon/nova-lite.toml
+++ b/models/amazon/nova-lite.toml
@@ -1,3 +1,8 @@
+# Sources: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-lite.html
+# Launch: https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/
+# PDF document input is supported through Bedrock Converse.
+# Output: the Nova V1 guide's 10K is confirmed by US Bedrock Converse boundary checks (2026-09-09).
 name = "Nova Lite"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 family = "nova-lite"
@@ -12,8 +17,8 @@ open_weights = false
 
 [limit]
 context = 300_000
-output = 8_192
+output = 10_000
 
 [modalities]
-input = ["text", "image", "video"]
+input = ["text", "image", "video", "pdf"]
 output = ["text"]
diff --git a/models/amazon/nova-micro.toml b/models/amazon/nova-micro.toml
index e786a6f05c4..bc353d076ae 100644
--- a/models/amazon/nova-micro.toml
+++ b/models/amazon/nova-micro.toml
@@ -1,3 +1,7 @@
+# Sources: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-micro.html
+# Launch: https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/
+# Output: the Nova V1 guide's 10K is confirmed by US Bedrock Converse boundary checks (2026-09-09).
 name = "Nova Micro"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
 family = "nova-micro"
@@ -12,7 +16,7 @@ open_weights = false
 
 [limit]
 context = 128_000
-output = 8_192
+output = 10_000
 
 [modalities]
 input = ["text"]
diff --git a/models/amazon/nova-premier.toml b/models/amazon/nova-premier.toml
new file mode 100644
index 00000000000..15dea69332a
--- /dev/null
+++ b/models/amazon/nova-premier.toml
@@ -0,0 +1,23 @@
+# Launch: https://aws.amazon.com/blogs/aws/amazon-nova-premier-our-most-capable-model-for-complex-tasks-and-teacher-for-model-distillation/
+# Specs (including 10K output): https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Prompted CoT, not a native thinking mode: https://docs.aws.amazon.com/nova/latest/userguide/prompting-chain-of-thought.html
+# Bedrock's generic Premier card conflicts on launch date, output limit, and reasoning; use the Nova V1 guide and launch announcement.
+name = "Nova Premier"
+description = "Multimodal model for complex analysis, long-context understanding, tool use, and model distillation"
+family = "nova"
+release_date = "2025-04-30"
+last_updated = "2025-04-30"
+attachment = true
+reasoning = false
+temperature = true
+knowledge = "2024-10"
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 10_000
+
+[modalities]
+input = ["text", "image", "video", "pdf"]
+output = ["text"]
diff --git a/models/amazon/nova-pro.toml b/models/amazon/nova-pro.toml
index 40119549868..5b04d7dfab4 100644
--- a/models/amazon/nova-pro.toml
+++ b/models/amazon/nova-pro.toml
@@ -1,3 +1,8 @@
+# Sources: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-pro.html
+# Launch: https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/
+# PDF document input is supported through Bedrock Converse.
+# Output: the Nova V1 guide's 10K is confirmed by US Bedrock Converse boundary checks (2026-09-09).
 name = "Nova Pro"
 description = "Flagship model for demanding analysis, coding, and production agent workflows"
 family = "nova-pro"
@@ -12,8 +17,8 @@ open_weights = false
 
 [limit]
 context = 300_000
-output = 8_192
+output = 10_000
 
 [modalities]
-input = ["text", "image", "video"]
+input = ["text", "image", "video", "pdf"]
 output = ["text"]
diff --git a/providers/amazon-bedrock/models/amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/amazon.nova-2-lite-v1:0.toml
index c2ddd59cf87..280c5a8927d 100644
--- a/providers/amazon-bedrock/models/amazon.nova-2-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/amazon.nova-2-lite-v1:0.toml
@@ -1,23 +1,21 @@
-name = "Nova 2 Lite"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "nova"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = true
+# Model: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-2-lite.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Reasoning: https://docs.aws.amazon.com/nova/latest/nova2-userguide/extended-thinking.html
+# Toggle: additionalModelRequestFields.reasoningConfig.type = enabled|disabled (Converse).
+# Effort: additionalModelRequestFields.reasoningConfig.maxReasoningEffort = low|medium|high.
+# High effort requires omitting temperature, topP, topK, and maxTokens.
+# Bedrock output cap: US Converse boundary check (2026-09-09) accepts 65,535 and rejects 65,536.
+# Lab specs: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
+base_model = "amazon/nova-2-lite"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-open_weights = false
 
 [cost]
 input = 0.33
 output = 2.75
+cache_read = 0.0825
+cache_write = 0.33
 
 [limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+output = 65_535
diff --git a/providers/amazon-bedrock/models/amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/amazon.nova-lite-v1:0.toml
index 671e8a15b26..3c0867bc54d 100644
--- a/providers/amazon-bedrock/models/amazon.nova-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/amazon.nova-lite-v1:0.toml
@@ -1,24 +1,11 @@
-name = "Nova Lite"
-description = "Efficient model for low-latency assistance, extraction, and routine automation"
-family = "nova-lite"
-release_date = "2024-12-03"
-last_updated = "2024-12-03"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-10"
-tool_call = true
-open_weights = false
+# Model: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-lite.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+base_model = "amazon/nova-lite"
 
 [cost]
 input = 0.06
 output = 0.24
 cache_read = 0.015
-
-[limit]
-context = 300_000
-output = 8_192
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+cache_write = 0.06
diff --git a/providers/amazon-bedrock/models/amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/amazon.nova-micro-v1:0.toml
index b936adf1844..0693d5a5b59 100644
--- a/providers/amazon-bedrock/models/amazon.nova-micro-v1:0.toml
+++ b/providers/amazon-bedrock/models/amazon.nova-micro-v1:0.toml
@@ -1,24 +1,11 @@
-name = "Nova Micro"
-description = "Efficient model for low-latency assistance, extraction, and routine automation"
-family = "nova-micro"
-release_date = "2024-12-03"
-last_updated = "2024-12-03"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2024-10"
-tool_call = true
-open_weights = false
+# Model: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-micro.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+base_model = "amazon/nova-micro"
 
 [cost]
 input = 0.035
 output = 0.14
 cache_read = 0.00875
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
+cache_write = 0.035
diff --git a/providers/amazon-bedrock/models/amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/amazon.nova-pro-v1:0.toml
index 0c37b88ffe5..39cce486786 100644
--- a/providers/amazon-bedrock/models/amazon.nova-pro-v1:0.toml
+++ b/providers/amazon-bedrock/models/amazon.nova-pro-v1:0.toml
@@ -1,24 +1,11 @@
-name = "Nova Pro"
-description = "Flagship model for demanding analysis, coding, and production agent workflows"
-family = "nova-pro"
-release_date = "2024-12-03"
-last_updated = "2024-12-03"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-10"
-tool_call = true
-open_weights = false
+# Model: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-pro.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+base_model = "amazon/nova-pro"
 
 [cost]
 input = 0.80
 output = 3.20
 cache_read = 0.20
-
-[limit]
-context = 300_000
-output = 8_192
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+cache_write = 0.80
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
index e513171b623..a10f4385413 100644
--- a/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-lite-v1:0.toml
@@ -1,6 +1,8 @@
-# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
-# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
-# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# APAC cross-Region profile verified by ListInferenceProfiles (ap-southeast-2, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/ap-southeast-2/index.json
+# Standard ap-southeast-2 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-lite"
 name = "Nova Lite (APAC)"
 
@@ -8,3 +10,4 @@ name = "Nova Lite (APAC)"
 input = 0.063
 output = 0.252
 cache_read = 0.01575
+cache_write = 0.063
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
index efab7cd712f..15c02027943 100644
--- a/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-micro-v1:0.toml
@@ -1,6 +1,8 @@
-# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
-# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
-# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# APAC cross-Region profile verified by ListInferenceProfiles (ap-southeast-2, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/ap-southeast-2/index.json
+# Standard ap-southeast-2 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-micro"
 name = "Nova Micro (APAC)"
 
@@ -8,3 +10,4 @@ name = "Nova Micro (APAC)"
 input = 0.037
 output = 0.148
 cache_read = 0.00925
+cache_write = 0.037
diff --git a/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
index da293e8ec9d..358a342d708 100644
--- a/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
+++ b/providers/amazon-bedrock/models/apac.amazon.nova-pro-v1:0.toml
@@ -1,6 +1,8 @@
-# APAC cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
-# Priced at ap-southeast-2 standard; also invocable from ap-northeast-1/ap-southeast-1 where regional rates differ slightly.
-# Prices (Price List): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# APAC cross-Region profile verified by ListInferenceProfiles (ap-southeast-2, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/ap-southeast-2/index.json
+# Standard ap-southeast-2 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-pro"
 name = "Nova Pro (APAC)"
 
@@ -8,3 +10,4 @@ name = "Nova Pro (APAC)"
 input = 0.840
 output = 3.360
 cache_read = 0.210
+cache_write = 0.840
diff --git a/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
index 316aa5c3b62..c6f55a4c614 100644
--- a/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/ca.amazon.nova-lite-v1:0.toml
@@ -1,5 +1,8 @@
-# CA cross-Region inference profile, verified via ListInferenceProfiles from ca-central-1.
-# CA geo pricing (Price List ca-central-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# CA cross-Region profile verified by ListInferenceProfiles (ca-central-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/ca-central-1/index.json
+# Standard ca-central-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-lite"
 name = "Nova Lite (CA)"
 
@@ -7,3 +10,4 @@ name = "Nova Lite (CA)"
 input = 0.064
 output = 0.256
 cache_read = 0.016
+cache_write = 0.064
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
index 13a858c8b57..cc33e084317 100644
--- a/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-2-lite-v1:0.toml
@@ -1,6 +1,13 @@
-# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
-# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
-# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+# EU cross-Region profile verified by ListInferenceProfiles (eu-west-1, 2026-09-08).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/eu-west-1/index.json
+# Standard eu-west-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Reasoning: https://docs.aws.amazon.com/nova/latest/nova2-userguide/extended-thinking.html
+# Toggle: additionalModelRequestFields.reasoningConfig.type = enabled|disabled (Converse).
+# Effort: additionalModelRequestFields.reasoningConfig.maxReasoningEffort = low|medium|high.
+# High effort requires omitting temperature, topP, topK, and maxTokens.
+# Bedrock output cap: US Converse boundary check (2026-09-09) accepts 65,535 and rejects 65,536.
+# Lab specs: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
 base_model = "amazon/nova-2-lite"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 name = "Nova 2 Lite (EU)"
@@ -9,3 +16,7 @@ name = "Nova 2 Lite (EU)"
 input = 0.374
 output = 3.157
 cache_read = 0.0935
+cache_write = 0.374
+
+[limit]
+output = 65_535
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
index d1f01cfcb38..bcf2a1a3e5b 100644
--- a/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-lite-v1:0.toml
@@ -1,5 +1,8 @@
-# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
-# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# EU cross-Region profile verified by ListInferenceProfiles (eu-west-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/eu-west-1/index.json
+# Standard eu-west-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-lite"
 name = "Nova Lite (EU)"
 
@@ -7,3 +10,4 @@ name = "Nova Lite (EU)"
 input = 0.069
 output = 0.276
 cache_read = 0.01725
+cache_write = 0.069
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
index 08d27c708f0..d4ac6403928 100644
--- a/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-micro-v1:0.toml
@@ -1,5 +1,8 @@
-# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
-# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# EU cross-Region profile verified by ListInferenceProfiles (eu-west-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/eu-west-1/index.json
+# Standard eu-west-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-micro"
 name = "Nova Micro (EU)"
 
@@ -7,3 +10,4 @@ name = "Nova Micro (EU)"
 input = 0.040
 output = 0.160
 cache_read = 0.010
+cache_write = 0.040
diff --git a/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
index 68e21d0d284..e7b8f776669 100644
--- a/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.amazon.nova-pro-v1:0.toml
@@ -1,5 +1,8 @@
-# EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
-# EU geo pricing (Price List eu-west-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# EU cross-Region profile verified by ListInferenceProfiles (eu-west-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/eu-west-1/index.json
+# Standard eu-west-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-pro"
 name = "Nova Pro (EU)"
 
@@ -7,3 +10,4 @@ name = "Nova Pro (EU)"
 input = 0.920
 output = 3.680
 cache_read = 0.230
+cache_write = 0.920
diff --git a/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
index aa803d4bf43..22f70547f8c 100644
--- a/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/global.amazon.nova-2-lite-v1:0.toml
@@ -1,6 +1,13 @@
-# Global cross-Region inference profile, verified via ListInferenceProfiles (visible from all regions).
-# Global pricing (Price List us-east-1 global): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
-# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+# Global cross-Region profile verified by ListInferenceProfiles in six source Regions (2026-09-08).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Global us-east-1 reference rates (cross-region-global SKUs), USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Reasoning: https://docs.aws.amazon.com/nova/latest/nova2-userguide/extended-thinking.html
+# Toggle: additionalModelRequestFields.reasoningConfig.type = enabled|disabled (Converse).
+# Effort: additionalModelRequestFields.reasoningConfig.maxReasoningEffort = low|medium|high.
+# High effort requires omitting temperature, topP, topK, and maxTokens.
+# Bedrock output cap: US Converse boundary check (2026-09-09) accepts 65,535 and rejects 65,536.
+# Lab specs: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
 base_model = "amazon/nova-2-lite"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 name = "Nova 2 Lite (Global)"
@@ -9,3 +16,7 @@ name = "Nova 2 Lite (Global)"
 input = 0.300
 output = 2.500
 cache_read = 0.075
+cache_write = 0.300
+
+[limit]
+output = 65_535
diff --git a/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
index f652010870d..b689ff48311 100644
--- a/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/jp.amazon.nova-2-lite-v1:0.toml
@@ -1,6 +1,13 @@
-# JP cross-Region inference profile, verified via ListInferenceProfiles from ap-northeast-1.
-# JP geo pricing (Price List ap-northeast-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
-# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+# JP cross-Region profile verified by ListInferenceProfiles (ap-northeast-1, 2026-09-08).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/ap-northeast-1/index.json
+# Standard ap-northeast-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Reasoning: https://docs.aws.amazon.com/nova/latest/nova2-userguide/extended-thinking.html
+# Toggle: additionalModelRequestFields.reasoningConfig.type = enabled|disabled (Converse).
+# Effort: additionalModelRequestFields.reasoningConfig.maxReasoningEffort = low|medium|high.
+# High effort requires omitting temperature, topP, topK, and maxTokens.
+# Bedrock output cap: US Converse boundary check (2026-09-09) accepts 65,535 and rejects 65,536.
+# Lab specs: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
 base_model = "amazon/nova-2-lite"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 name = "Nova 2 Lite (JP)"
@@ -9,3 +16,7 @@ name = "Nova 2 Lite (JP)"
 input = 0.396
 output = 3.311
 cache_read = 0.099
+cache_write = 0.396
+
+[limit]
+output = 65_535
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
index 18c7833c41b..cdd58992c96 100644
--- a/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.amazon.nova-2-lite-v1:0.toml
@@ -1,6 +1,13 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1 and ca-central-1.
-# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
-# Toggle: reasoningConfig.type = enabled|disabled; Effort: reasoningConfig.maxReasoningEffort = low|medium|high
+# US cross-Region profile verified by ListInferenceProfiles (us-east-1 and ca-central-1, 2026-09-08).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; Canadian source rates differ. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Reasoning: https://docs.aws.amazon.com/nova/latest/nova2-userguide/extended-thinking.html
+# Toggle: additionalModelRequestFields.reasoningConfig.type = enabled|disabled (Converse).
+# Effort: additionalModelRequestFields.reasoningConfig.maxReasoningEffort = low|medium|high.
+# High effort requires omitting temperature, topP, topK, and maxTokens.
+# Bedrock output cap: US Converse boundary check (2026-09-09) accepts 65,535 and rejects 65,536.
+# Lab specs: https://docs.aws.amazon.com/nova/latest/nova2-userguide/what-is-nova-2.html
 base_model = "amazon/nova-2-lite"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 name = "Nova 2 Lite (US)"
@@ -9,3 +16,7 @@ name = "Nova 2 Lite (US)"
 input = 0.33
 output = 2.75
 cache_read = 0.0825
+cache_write = 0.33
+
+[limit]
+output = 65_535
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
index 38b93c68908..8ccf82880f4 100644
--- a/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.amazon.nova-lite-v1:0.toml
@@ -1,5 +1,8 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
-# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# US cross-Region profile verified by ListInferenceProfiles (us-east-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-lite"
 name = "Nova Lite (US)"
 
@@ -7,3 +10,4 @@ name = "Nova Lite (US)"
 input = 0.06
 output = 0.24
 cache_read = 0.015
+cache_write = 0.06
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
index 781b77c917a..e949c022662 100644
--- a/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.amazon.nova-micro-v1:0.toml
@@ -1,5 +1,8 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
-# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# US cross-Region profile verified by ListInferenceProfiles (us-east-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-micro"
 name = "Nova Micro (US)"
 
@@ -7,3 +10,4 @@ name = "Nova Micro (US)"
 input = 0.035
 output = 0.14
 cache_read = 0.00875
+cache_write = 0.035
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-premier-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-premier-v1:0.toml
new file mode 100644
index 00000000000..feb36290000
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.amazon.nova-premier-v1:0.toml
@@ -0,0 +1,16 @@
+# US cross-Region profile verified by ListInferenceProfiles (us-east-1, 2026-09-08).
+# Profile: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-premier.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
+# Live Converse (2026-09-09): Legacy access gate denies accounts without use in the last 30 days; profile retained for existing users.
+base_model = "amazon/nova-premier"
+name = "Nova Premier (US)"
+status = "deprecated"
+structured_output = false
+
+[cost]
+input = 2.50
+output = 12.50
+cache_read = 0.625
+cache_write = 2.50
diff --git a/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml b/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml
index 259d180c70d..6e5e88e0e37 100644
--- a/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.amazon.nova-pro-v1:0.toml
@@ -1,5 +1,8 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
-# US geo pricing (Price List us-east-1 standard): input/output/cache-read USD per 1M tokens; Nova has no cache-write charge.
+# US cross-Region profile verified by ListInferenceProfiles (us-east-1, 2026-09-08).
+# Model: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Standard us-east-1 reference rates, USD/MTok; rates vary by source Region. Cache writes billed at input rate; no surcharge.
+# Cache billing: https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html
 base_model = "amazon/nova-pro"
 name = "Nova Pro (US)"
 
@@ -7,3 +10,4 @@ name = "Nova Pro (US)"
 input = 0.80
 output = 3.20
 cache_read = 0.20
+cache_write = 0.80

From 0fdb49b07d0d141d4d8995446bdb21ac2c2aeb90 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:18:15 +0000
Subject: [PATCH 1065/2012] fix: [missing-model] ofox: qwen/qwen-flash (#6784)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/ofox/models/qwen/qwen-flash.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/ofox/models/qwen/qwen-flash.toml

diff --git a/providers/ofox/models/qwen/qwen-flash.toml b/providers/ofox/models/qwen/qwen-flash.toml
new file mode 100644
index 00000000000..eca6704d742
--- /dev/null
+++ b/providers/ofox/models/qwen/qwen-flash.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/qwen/qwen-flash
+# https://api.ofox.ai/v1/models/qwen/qwen-flash
+# https://help.aliyun.com/en/model-studio/deep-thinking
+# Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
+# Budget: thinking_budget
+base_model = "alibaba/qwen-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.022
+output = 0.22
+cache_read = 0.0043
+cache_write = 0.027
+
+[limit]
+output = 32_000

From b03f91b7ba9bd8a27373343d3d2ed0cbed2d6f97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:18:19 +0000
Subject: [PATCH 1066/2012] fix: [missing-model] ofox: google/gemini-3.8-flash
 (#6786)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../ofox/models/google/gemini-3.8-flash.toml  | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100644 providers/ofox/models/google/gemini-3.8-flash.toml

diff --git a/providers/ofox/models/google/gemini-3.8-flash.toml b/providers/ofox/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..62f8fcecf5b
--- /dev/null
+++ b/providers/ofox/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,26 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/google/gemini-3.8-flash
+# https://api.ofox.ai/v1/models/google/gemini-3.8-flash
+# https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+cache_write = 0.0415
+input_audio = 0.75
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image", "audio", "pdf"]
+
+[provider]
+npm = "@ai-sdk/google"
+api = "https://api.ofox.ai/gemini/v1beta"

From 0d80985de86c008afd5f75c57fea96999691b333 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:18:22 +0000
Subject: [PATCH 1067/2012] fix: [missing-model] ofox:
 deepseek/deepseek-v4-flash-0423 (#6785)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/deepseek/deepseek-v4-flash-0423.toml   | 28 +++++++++++++++++++
 .../deepseek/deepseek-v4-flash-0423.toml      | 18 ++++++++++++
 2 files changed, 46 insertions(+)
 create mode 100644 models/deepseek/deepseek-v4-flash-0423.toml
 create mode 100644 providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml

diff --git a/models/deepseek/deepseek-v4-flash-0423.toml b/models/deepseek/deepseek-v4-flash-0423.toml
new file mode 100644
index 00000000000..0de9cc46fd0
--- /dev/null
+++ b/models/deepseek/deepseek-v4-flash-0423.toml
@@ -0,0 +1,28 @@
+# Sources:
+# https://www.deepseek.com/en/news/v4-preview/
+# https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
+name = "DeepSeek V4 Flash 0423"
+description = "Initial DeepSeek V4 Flash snapshot for economical reasoning, coding, and million-token agent workloads"
+family = "deepseek-flash"
+release_date = "2026-04-23"
+last_updated = "2026-04-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-05"
+open_weights = true
+license = "MIT"
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash"
diff --git a/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml b/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml
new file mode 100644
index 00000000000..d5d592e6729
--- /dev/null
+++ b/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-11):
+# https://ofox.ai/models/deepseek/deepseek-v4-flash-0423
+# https://api.ofox.ai/v1/models/deepseek/deepseek-v4-flash-0423
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+base_model = "deepseek/deepseek-v4-flash-0423"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.44
+output = 1.32
+cache_read = 0.014

From cdfb15b7c765dd4272be12c7aeab04d7a21841bd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 21:25:51 -0500
Subject: [PATCH 1068/2012] fix: [missing-model] pioneer: zai-org/GLM-5.2-Fast
 (#6811)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../pioneer/models/zai-org/GLM-5.2-Fast.toml   | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/pioneer/models/zai-org/GLM-5.2-Fast.toml

diff --git a/providers/pioneer/models/zai-org/GLM-5.2-Fast.toml b/providers/pioneer/models/zai-org/GLM-5.2-Fast.toml
new file mode 100644
index 00000000000..9afa02f9bc0
--- /dev/null
+++ b/providers/pioneer/models/zai-org/GLM-5.2-Fast.toml
@@ -0,0 +1,18 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "zhipuai/glm-5.2"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 2.1
+output = 6.6
+cache_read = 0.21
+cache_write = 2.1
+
+[limit]
+output = 128_000

From fb36b8a8616de53f83b39396c77f9563edaf060d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:34:44 +0000
Subject: [PATCH 1069/2012] chore(sync): update Merge Gateway model catalog
 (#6818)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0423.toml       | 18 ++++++++++++++++++
 .../merge-gateway/models/qwen/qwen3.5-27b.toml |  7 ++++---
 .../models/qwen/qwen3.5-397b-a17b.toml         |  7 ++++---
 3 files changed, 26 insertions(+), 6 deletions(-)
 create mode 100644 providers/merge-gateway/models/deepseek/deepseek-v4-flash-0423.toml

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0423.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0423.toml
new file mode 100644
index 00000000000..95eeecf3e4c
--- /dev/null
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0423.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4-flash-0423"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.139
+output = 0.278
+
+[limit]
+context = 131_072
+output = 32_768
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index f0885b3822b..d7d8d42decc 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -10,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 0264c6bc733..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 8ae9265b8b2b4d8fb23946c46d84bc506b6c5b16 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:34:50 +0000
Subject: [PATCH 1070/2012] chore(sync): update OpenRouter model catalog
 (#6819)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k3.toml      | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index bc1e985b845..c831cb49220 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.6
-output = 13
-cache_read = 0.29
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index e7ceb0c72c4..f78c60f8145 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.4
-output = 12
-cache_read = 0.24
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 context = 1_048_576

From b2dc9cc0b2e5308690dc184e03bbee0749ee767a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 02:35:00 +0000
Subject: [PATCH 1071/2012] chore(sync): update Kilo model catalog (#6820)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 7b92a6177bd..4ff6f302821 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.4
-output = 12
-cache_read = 0.24
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 context = 1_048_576

From a4a156f1891bd42f0e55f253194ce85113dcee18 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 03:30:18 +0000
Subject: [PATCH 1072/2012] chore(sync): update Kilo model catalog (#6829)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 2 +-
 providers/kilo/models/~moonshotai/kimi-latest.toml          | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 80dc12b4ce8..bd0a4217a38 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,7 +15,7 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05
+input = 0.04
 output = 0.16
 cache_read = 0.013
 
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 4ff6f302821..ab70d74a6d7 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 context = 1_048_576

From bc0281f9dad9fc6c0a90b39254b970813273bbc8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 03:30:22 +0000
Subject: [PATCH 1073/2012] chore(sync): update OpenRouter model catalog
 (#6828)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/google/gemma-4-26b-a4b-it.toml  | 3 +++
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index a27c6e610ef..f8a013291bc 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -9,5 +9,8 @@ type = "toggle"
 input = 0.042
 output = 0.22
 
+[limit]
+context = 1_000_000
+
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index c831cb49220..af500f205e8 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 27d5c88cb9e..5ab4a3ffd9b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,7 +20,7 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
+input = 0.04
 output = 0.16
 cache_read = 0.013
 
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index f78c60f8145..66be7b84546 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 context = 1_048_576

From bc1ade872930cf943b7534b4f2a1527e8d2a575a Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:19:37 -0500
Subject: [PATCH 1074/2012] fix(amazon): preserve Nova 2 Lite update date
 (#6832)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 models/amazon/nova-2-lite.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/models/amazon/nova-2-lite.toml b/models/amazon/nova-2-lite.toml
index 5d42c94dbb4..4995aa70ddf 100644
--- a/models/amazon/nova-2-lite.toml
+++ b/models/amazon/nova-2-lite.toml
@@ -5,7 +5,7 @@ name = "Nova 2 Lite"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "nova"
 release_date = "2025-12-02"
-last_updated = "2025-12-02"
+last_updated = "2025-12-01"
 attachment = true
 reasoning = true
 temperature = true

From c2e998464eb17b8ad223728698bb5ed31f243dea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:19:52 -0500
Subject: [PATCH 1075/2012] fix: [missing-model] pioneer: mistral-medium
 (#6827)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/pioneer/models/mistral-medium.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/pioneer/models/mistral-medium.toml

diff --git a/providers/pioneer/models/mistral-medium.toml b/providers/pioneer/models/mistral-medium.toml
new file mode 100644
index 00000000000..549214fa0ce
--- /dev/null
+++ b/providers/pioneer/models/mistral-medium.toml
@@ -0,0 +1,12 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "mistral/mistral-medium-2505"
+
+[cost]
+input = 0.4
+output = 2
+cache_read = 0.4
+cache_write = 0.4
+
+[limit]
+context = 128_000
+output = 64_000

From 6a79e17c16ee5c813764efe4590c0417144b87fe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 04:30:30 +0000
Subject: [PATCH 1076/2012] chore(sync): update NanoGPT model catalog (#6836)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index fb4e3a2dd26..9c2641248c0 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index f23aac53ab8..2e7355d5588 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From d3825080af04753835228dc1009df254c8e8315d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 04:30:33 +0000
Subject: [PATCH 1077/2012] chore(sync): update Kilo model catalog (#6837)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index ab70d74a6d7..74831ee82da 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 context = 1_048_576

From 9094603c0d42edeebcca61d44d1be86db4fd5496 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 04:30:40 +0000
Subject: [PATCH 1078/2012] chore(sync): update OpenRouter model catalog
 (#6838)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index af500f205e8..975660fffb4 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 66be7b84546..6eb528f593d 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 context = 1_048_576

From e7daff54c2c56d5930f5b7db4dca6bcc659b4982 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:32:35 -0500
Subject: [PATCH 1079/2012] fix(amazon-bedrock): correct Gemma catalog (#6824)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 models/google/gemma-3-12b-it.toml             | 25 ++++++++++++++++++
 models/google/gemma-3-27b-it.toml             | 25 ++++++++++++++++++
 models/google/gemma-3-4b-it.toml              | 25 ++++++++++++++++++
 .../models/google.gemma-3-12b-it.toml         | 26 ++++++-------------
 .../models/google.gemma-3-27b-it.toml         | 26 +++++++------------
 .../models/google.gemma-3-4b-it.toml          | 21 +++++----------
 .../models/google.gemma-4-26b-a4b.toml        | 19 ++++++++++++++
 .../models/google.gemma-4-31b.toml            | 19 ++++++++++++++
 .../models/google.gemma-4-e2b.toml            | 19 ++++++++++++++
 9 files changed, 155 insertions(+), 50 deletions(-)
 create mode 100644 models/google/gemma-3-12b-it.toml
 create mode 100644 models/google/gemma-3-27b-it.toml
 create mode 100644 models/google/gemma-3-4b-it.toml
 create mode 100644 providers/amazon-bedrock/models/google.gemma-4-26b-a4b.toml
 create mode 100644 providers/amazon-bedrock/models/google.gemma-4-31b.toml
 create mode 100644 providers/amazon-bedrock/models/google.gemma-4-e2b.toml

diff --git a/models/google/gemma-3-12b-it.toml b/models/google/gemma-3-12b-it.toml
new file mode 100644
index 00000000000..5ca0388a3db
--- /dev/null
+++ b/models/google/gemma-3-12b-it.toml
@@ -0,0 +1,25 @@
+# https://ai.google.dev/gemma/docs/core/model_card_3
+# https://blog.google/technology/developers/gemma-3/
+name = "Gemma 3 12B IT"
+description = "Open multimodal Gemma instruction model for multilingual text generation and image understanding"
+family = "gemma"
+release_date = "2025-03-12"
+last_updated = "2025-03-12"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2024-08"
+open_weights = true
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-3-12b-it"
diff --git a/models/google/gemma-3-27b-it.toml b/models/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..2ef351ba4c1
--- /dev/null
+++ b/models/google/gemma-3-27b-it.toml
@@ -0,0 +1,25 @@
+# https://ai.google.dev/gemma/docs/core/model_card_3
+# https://blog.google/technology/developers/gemma-3/
+name = "Gemma 3 27B IT"
+description = "Largest open Gemma 3 instruction model for multilingual text generation and visual understanding"
+family = "gemma"
+release_date = "2025-03-12"
+last_updated = "2025-03-12"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2024-08"
+open_weights = true
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-3-27b-it"
diff --git a/models/google/gemma-3-4b-it.toml b/models/google/gemma-3-4b-it.toml
new file mode 100644
index 00000000000..bef57eb4cac
--- /dev/null
+++ b/models/google/gemma-3-4b-it.toml
@@ -0,0 +1,25 @@
+# https://ai.google.dev/gemma/docs/core/model_card_3
+# https://blog.google/technology/developers/gemma-3/
+name = "Gemma 3 4B IT"
+description = "Open multimodal Gemma instruction model for efficient text generation and image understanding"
+family = "gemma"
+release_date = "2025-03-12"
+last_updated = "2025-03-12"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2024-08"
+open_weights = true
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-3-4b-it"
diff --git a/providers/amazon-bedrock/models/google.gemma-3-12b-it.toml b/providers/amazon-bedrock/models/google.gemma-3-12b-it.toml
index 68b871bf94d..756e88ef4c5 100644
--- a/providers/amazon-bedrock/models/google.gemma-3-12b-it.toml
+++ b/providers/amazon-bedrock/models/google.gemma-3-12b-it.toml
@@ -1,24 +1,14 @@
-name = "Google Gemma 3 12B"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-temperature = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-3-12b-it.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live Converse (us-east-1, 2026-09-09): forced toolUse works, but native toolResult continuation fails with a role-alternation error.
+# Tool calling remains disabled because a complete native tool-use round trip is unavailable.
+base_model = "google/gemma-3-12b-it"
 tool_call = false
 structured_output = true
-knowledge = "2024-12"
-open_weights = false
 
 [cost]
-input = 0.049999999999999996
-output = 0.09999999999999999
+input = 0.09
+output = 0.29
 
 [limit]
-context = 131072
-output = 8192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/google.gemma-3-27b-it.toml b/providers/amazon-bedrock/models/google.gemma-3-27b-it.toml
index c186c98763c..052bc51a507 100644
--- a/providers/amazon-bedrock/models/google.gemma-3-27b-it.toml
+++ b/providers/amazon-bedrock/models/google.gemma-3-27b-it.toml
@@ -1,24 +1,16 @@
-name = "Google Gemma 3 27B Instruct"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-attachment = true
-reasoning = false
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-3-27b-pt.html
+# AWS's card title says PT, but its programmatic ID and discovery name specify IT.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live Converse (us-east-1, 2026-09-09): forced toolUse works, but native toolResult continuation fails with a role-alternation error.
+# Tool calling remains disabled because a complete native tool-use round trip is unavailable.
+base_model = "google/gemma-3-27b-it"
+tool_call = false
 structured_output = true
-temperature = true
-knowledge = "2025-07"
-release_date = "2025-07-27"
-last_updated = "2025-07-27"
-open_weights = true
 
 [cost]
-input = 0.12
-output = 0.2
+input = 0.23
+output = 0.38
 
 [limit]
 context = 202_752
 output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/google.gemma-3-4b-it.toml b/providers/amazon-bedrock/models/google.gemma-3-4b-it.toml
index bcf4d664264..57c569b281b 100644
--- a/providers/amazon-bedrock/models/google.gemma-3-4b-it.toml
+++ b/providers/amazon-bedrock/models/google.gemma-3-4b-it.toml
@@ -1,22 +1,13 @@
-name = "Gemma 3 4B IT"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-3-4b-it.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live Converse (us-east-1, 2026-09-09): forced toolUse works, but native toolResult continuation fails with a role-alternation error.
+# Tool calling remains disabled because a complete native tool-use round trip is unavailable.
+base_model = "google/gemma-3-4b-it"
+tool_call = false
 
 [cost]
 input = 0.04
 output = 0.08
 
 [limit]
-context = 128_000
 output = 4_096
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/google.gemma-4-26b-a4b.toml b/providers/amazon-bedrock/models/google.gemma-4-26b-a4b.toml
new file mode 100644
index 00000000000..89a35733be0
--- /dev/null
+++ b/providers/amazon-bedrock/models/google.gemma-4-26b-a4b.toml
@@ -0,0 +1,19 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-4-26b-a4b.html
+# https://github.com/aws-samples/bedrock-claude-chatbot/blob/eeb2fa35339081d34ea9caa4e4d39cf1c189770a/agent/chat_agent.py
+# Effort: reasoning.effort = none|high on Mantle Responses, matching the Gemma 4 31B surface.
+# Live 2026-09-09: high returns reasoning; strict JSON schema and max_output_tokens=32768 accepted.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+base_model = "google/gemma-4-26b-a4b-it"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.13
+output = 0.40
+
+[modalities]
+input = ["text", "image", "video"]
+
+[provider]
+npm = "@ai-sdk/amazon-bedrock/mantle"
+api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+shape = "responses"
diff --git a/providers/amazon-bedrock/models/google.gemma-4-31b.toml b/providers/amazon-bedrock/models/google.gemma-4-31b.toml
new file mode 100644
index 00000000000..0e01ba73f24
--- /dev/null
+++ b/providers/amazon-bedrock/models/google.gemma-4-31b.toml
@@ -0,0 +1,19 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-4-31b.html
+# https://github.com/aws-samples/bedrock-claude-chatbot/blob/eeb2fa35339081d34ea9caa4e4d39cf1c189770a/agent/chat_agent.py
+# Effort: reasoning.effort = none|high on Mantle Responses.
+# Live 2026-09-09: high returns reasoning, none omits it; strict JSON schema and max_output_tokens=32768 accepted.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+base_model = "google/gemma-4-31b-it"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.14
+output = 0.40
+
+[modalities]
+input = ["text", "image", "video"]
+
+[provider]
+npm = "@ai-sdk/amazon-bedrock/mantle"
+api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+shape = "responses"
diff --git a/providers/amazon-bedrock/models/google.gemma-4-e2b.toml b/providers/amazon-bedrock/models/google.gemma-4-e2b.toml
new file mode 100644
index 00000000000..94501897c8b
--- /dev/null
+++ b/providers/amazon-bedrock/models/google.gemma-4-e2b.toml
@@ -0,0 +1,19 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-google-gemma-4-e2b.html
+# Effort: reasoning.effort = none|high on Mantle Responses, matching the Gemma 4 31B surface.
+# AWS recommends high to keep thinking in the reasoning channel.
+# Live 2026-09-09: high returns reasoning; strict JSON schema and max_output_tokens=8192 accepted.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+base_model = "google/gemma-4-E2B-it"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.04
+output = 0.08
+
+[modalities]
+input = ["text", "image", "audio", "video"]
+
+[provider]
+npm = "@ai-sdk/amazon-bedrock/mantle"
+api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+shape = "responses"

From e2f6eeb18d270a9c81c62219728b295250a6315c Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 10 Sep 2026 23:33:18 -0500
Subject: [PATCH 1080/2012] fix(amazon-bedrock): correct OpenAI catalog (#6792)

* fix(amazon-bedrock): correct OpenAI catalog

* fix(amazon-bedrock): use documented GPT-OSS output limits
---
 models/openai/gpt-oss-safeguard-20b.toml      | 32 +++++++++++++++++++
 .../models/global.openai.gpt-5.6-luna.toml    |  3 +-
 .../models/global.openai.gpt-5.6-sol.toml     |  2 ++
 .../models/global.openai.gpt-5.6-terra.toml   |  2 ++
 .../models/global.openai.gpt-6-astra.toml     |  6 ++--
 .../models/in.openai.gpt-5.6-luna.toml        | 23 +++++++++++++
 .../models/in.openai.gpt-5.6-terra.toml       | 24 ++++++++++++++
 .../amazon-bedrock/models/openai.gpt-5.4.toml |  4 ++-
 .../amazon-bedrock/models/openai.gpt-5.5.toml |  4 ++-
 .../models/openai.gpt-5.6-luna.toml           |  5 +--
 .../models/openai.gpt-5.6-sol.toml            |  5 +--
 .../models/openai.gpt-5.6-terra.toml          |  5 +--
 .../models/openai.gpt-6-astra.toml            |  5 ++-
 .../models/openai.gpt-oss-120b-1:0.toml       | 20 ++++--------
 .../models/openai.gpt-oss-120b.toml           | 19 +++--------
 .../models/openai.gpt-oss-20b-1:0.toml        | 20 ++++--------
 .../models/openai.gpt-oss-20b.toml            | 19 +++--------
 .../models/openai.gpt-oss-safeguard-120b.toml | 23 +++++--------
 .../models/openai.gpt-oss-safeguard-20b.toml  | 23 +++++--------
 .../us-gov.openai.gpt-oss-120b-1:0.toml       | 15 +++++++++
 .../models/us-gov.openai.gpt-oss-20b-1:0.toml | 15 +++++++++
 .../models/us.openai.gpt-5.6-luna.toml        |  6 ++--
 .../models/us.openai.gpt-5.6-sol.toml         |  4 ++-
 .../models/us.openai.gpt-5.6-terra.toml       |  4 ++-
 .../models/us.openai.gpt-6-astra.toml         |  6 ++--
 25 files changed, 188 insertions(+), 106 deletions(-)
 create mode 100644 models/openai/gpt-oss-safeguard-20b.toml
 create mode 100644 providers/amazon-bedrock/models/in.openai.gpt-5.6-luna.toml
 create mode 100644 providers/amazon-bedrock/models/in.openai.gpt-5.6-terra.toml
 create mode 100644 providers/amazon-bedrock/models/us-gov.openai.gpt-oss-120b-1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us-gov.openai.gpt-oss-20b-1:0.toml

diff --git a/models/openai/gpt-oss-safeguard-20b.toml b/models/openai/gpt-oss-safeguard-20b.toml
new file mode 100644
index 00000000000..c4b69735eee
--- /dev/null
+++ b/models/openai/gpt-oss-safeguard-20b.toml
@@ -0,0 +1,32 @@
+# Model and reasoning controls: https://huggingface.co/openai/gpt-oss-safeguard-20b
+# Text-only, reasoning and structured outputs: https://openai.com/index/gpt-oss-safeguard-technical-report/
+# Tool-call format: https://huggingface.co/openai/gpt-oss-safeguard-20b/blob/main/chat_template.jinja
+# Context: https://huggingface.co/openai/gpt-oss-safeguard-20b/blob/8a11e17b25c973a24099d4016bf2e17dd7ec1574/config.json
+# Output is the configured decoder-context ceiling, not a published hosted-API output maximum.
+# Prompt, reasoning and final output share the 131,072-token budget; usable output is the remaining context.
+# No separate fixed output cap: https://huggingface.co/openai/gpt-oss-safeguard-20b/blob/8a11e17b25c973a24099d4016bf2e17dd7ec1574/generation_config.json
+# Reference generation has a caller-supplied token cap (0 = uncapped): https://github.com/openai/gpt-oss/blob/main/gpt_oss/torch/model.py
+# Release: https://openai.com/index/introducing-gpt-oss-safeguard/
+name = "GPT OSS Safeguard 20B"
+description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
+family = "gpt-oss"
+release_date = "2025-10-29"
+last_updated = "2025-10-29"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/openai/gpt-oss-safeguard-20b"
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
index 4873bf0a1d0..f1928817aaa 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
@@ -1,11 +1,12 @@
 # Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Verified from eu-west-1: returns 200 with reasoning effort set.
 # Global pricing and Runtime capabilities: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max.
 base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-5.6 Luna (Global)"
-structured_output = false
 
 [cost]
 input = 0.20
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
index 7a1098be550..4b757f188e2 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
@@ -1,6 +1,8 @@
 # Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Verified from eu-west-1: returns 200 with reasoning effort set.
 # Global pricing and Runtime capabilities: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max (none verified 2026-09-09).
 base_model = "openai/gpt-5.6-sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
index be22b54fca5..3c9cd73b7d1 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
@@ -1,6 +1,8 @@
 # Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Verified from eu-west-1: effort none|low|medium|high|xhigh|max all return 200.
 # Global pricing and Runtime capabilities: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max.
 base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
index 8fd07d08872..0e751b4ebd6 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
@@ -1,6 +1,8 @@
 # Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
-# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
-# Global CRIS pricing inferred at 1.0x OpenAI-direct from the GPT-5.6 Sol AWS card precedent.
+# Bedrock model IDs, routes, limits and Global CRIS pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = low|medium|high|xhigh|max.
+# Live check 2026-09-09 (us-east-1): outputConfig.textFormat enforces JSON schema despite a conflicting user instruction.
 # Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
 base_model = "openai/gpt-6-astra"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/in.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/in.openai.gpt-5.6-luna.toml
new file mode 100644
index 00000000000..29cceb7a2fa
--- /dev/null
+++ b/providers/amazon-bedrock/models/in.openai.gpt-5.6-luna.toml
@@ -0,0 +1,23 @@
+# India cross-Region inference profile served by Bedrock Runtime (Converse).
+# Exact ID, ap-south-1/ap-south-2 availability, Geo CRIS prices and capabilities: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
+# Bedrock 1,050,000-token context: https://developers.openai.com/api/docs/guides/amazon-bedrock
+# Effort follows the native model and existing Runtime profiles: https://developers.openai.com/api/docs/models/gpt-5.6-luna
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max; reasoning_effort is rejected.
+# Live check 2026-09-09 (ap-south-1): nested reasoning.effort=none returns 200.
+base_model = "openai/gpt-5.6-luna"
+name = "GPT-5.6 Luna (India)"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.22
+output = 1.32
+cache_read = 0.022
+cache_write = 0.275
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.44
+output = 1.98
+cache_read = 0.044
+cache_write = 0.55
diff --git a/providers/amazon-bedrock/models/in.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/in.openai.gpt-5.6-terra.toml
new file mode 100644
index 00000000000..8f07b724fa4
--- /dev/null
+++ b/providers/amazon-bedrock/models/in.openai.gpt-5.6-terra.toml
@@ -0,0 +1,24 @@
+# India cross-Region inference profile served by Bedrock Runtime (Converse).
+# Exact ID, ap-south-1/ap-south-2 availability, Geo CRIS prices and capabilities: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
+# Bedrock 1,050,000-token context: https://developers.openai.com/api/docs/guides/amazon-bedrock
+# Effort follows the native model and existing Runtime profiles: https://developers.openai.com/api/docs/models/gpt-5.6-terra
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max; reasoning_effort is rejected.
+# Live check 2026-09-09 (ap-south-1): nested reasoning.effort=none returns 200.
+base_model = "openai/gpt-5.6-terra"
+name = "GPT-5.6 Terra (India)"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+structured_output = false
+
+[cost]
+input = 2.20
+output = 13.20
+cache_read = 0.22
+cache_write = 2.75
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4.40
+output = 19.80
+cache_read = 0.44
+cache_write = 5.50
diff --git a/providers/amazon-bedrock/models/openai.gpt-5.4.toml b/providers/amazon-bedrock/models/openai.gpt-5.4.toml
index 7e72a31b62d..f36bcbbe2c8 100644
--- a/providers/amazon-bedrock/models/openai.gpt-5.4.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-5.4.toml
@@ -1,3 +1,6 @@
+# Model ID, Mantle /openai/v1/responses route, 272K cap and US pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-54.html
+# Bedrock capabilities: https://developers.openai.com/api/docs/guides/amazon-bedrock
+# Effort: reasoning.effort = none|low|medium|high|xhigh; https://developers.openai.com/api/docs/models/gpt-5.4
 base_model = "openai/gpt-5.4"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 base_model_omit = ["cost.tiers", "cost.context_over_200k", "experimental.modes.fast", "limit.input"]
@@ -10,7 +13,6 @@ cache_read = 0.275
 
 [limit]
 context = 272_000
-output = 128_000
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
diff --git a/providers/amazon-bedrock/models/openai.gpt-5.5.toml b/providers/amazon-bedrock/models/openai.gpt-5.5.toml
index 98d0b854279..596b0e1d724 100644
--- a/providers/amazon-bedrock/models/openai.gpt-5.5.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-5.5.toml
@@ -1,3 +1,6 @@
+# Model ID, Mantle /openai/v1/responses route, 272K cap and US pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-55.html
+# Bedrock capabilities: https://developers.openai.com/api/docs/guides/amazon-bedrock
+# Effort: reasoning.effort = none|low|medium|high|xhigh; https://developers.openai.com/api/docs/models/gpt-5.5
 base_model = "openai/gpt-5.5"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 base_model_omit = ["cost.tiers", "cost.context_over_200k", "experimental.modes.fast", "limit.input"]
@@ -10,7 +13,6 @@ cache_read = 0.55
 
 [limit]
 context = 272_000
-output = 128_000
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
diff --git a/providers/amazon-bedrock/models/openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/openai.gpt-5.6-luna.toml
index 6a3cb3243e5..ed675b97297 100644
--- a/providers/amazon-bedrock/models/openai.gpt-5.6-luna.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-5.6-luna.toml
@@ -1,6 +1,7 @@
-# Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
+# Mantle Responses model ID, route and pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max; https://developers.openai.com/api/docs/models/gpt-5.6-luna
 # Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing (US East N. Virginia & Ohio; also US West Oregon): https://aws.amazon.com/bedrock/pricing/
+# In-region Standard pricing for US commercial Regions; GovCloud rates differ.
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/openai.gpt-5.6-sol.toml
index c6a19c521a6..6ce552bac27 100644
--- a/providers/amazon-bedrock/models/openai.gpt-5.6-sol.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-5.6-sol.toml
@@ -1,6 +1,7 @@
-# Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
+# Mantle Responses model ID, route and pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max; https://developers.openai.com/api/docs/models/gpt-5.6-sol
 # Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
+# In-region Standard pricing for US commercial Regions.
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/openai.gpt-5.6-terra.toml
index c885ad9858a..484ca1ba031 100644
--- a/providers/amazon-bedrock/models/openai.gpt-5.6-terra.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-5.6-terra.toml
@@ -1,6 +1,7 @@
-# Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
+# Mantle Responses model ID, route and pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max; https://developers.openai.com/api/docs/models/gpt-5.6-terra
 # Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing (US East N. Virginia & Ohio; also US West Oregon): https://aws.amazon.com/bedrock/pricing/
+# In-region Standard pricing for US commercial Regions; GovCloud rates differ.
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/openai.gpt-6-astra.toml
index b6abd665d27..273249d102f 100644
--- a/providers/amazon-bedrock/models/openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-6-astra.toml
@@ -1,7 +1,6 @@
-# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
+# Bedrock model IDs, routes, limits and pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
 # Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region pricing inferred at 1.1x OpenAI-direct from the GPT-5.6 Sol precedent
-# (AWS card prices in-region and Geo CRIS at 1.1x, Global CRIS at 1.0x); no AWS Astra model card published yet.
+# In-region Standard pricing; Mantle is available in us-west-2 (Oregon) only.
 # Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-6-astra"
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-120b-1:0.toml b/providers/amazon-bedrock/models/openai.gpt-oss-120b-1:0.toml
index 8ce6c18b5ba..f300361d517 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-120b-1:0.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-120b-1:0.toml
@@ -1,15 +1,11 @@
+# Model ID, Runtime APIs and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-120b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high (Converse).
+# Request mapping: https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+# AWS documents 16K max output. Converse accepts larger maxTokens settings up to the 128K context boundary, but that is not evidence of a larger generation cap.
+base_model = "openai/gpt-oss-120b"
 name = "gpt-oss-120b"
-description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-family = "gpt-oss"
-release_date = "2025-08-05"
-last_updated = "2025-08-05"
-attachment = false
-reasoning = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.15
@@ -18,7 +14,3 @@ output = 0.60
 [limit]
 context = 128_000
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-120b.toml b/providers/amazon-bedrock/models/openai.gpt-oss-120b.toml
index 31f710d07ff..c4e515caeb1 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-120b.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-120b.toml
@@ -1,15 +1,10 @@
+# Model ID, Mantle /v1/responses route and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-120b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Effort: reasoning.effort = low|medium|high (Responses).
+# Native controls: https://huggingface.co/openai/gpt-oss-120b
+base_model = "openai/gpt-oss-120b"
 name = "gpt-oss-120b"
-description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-family = "gpt-oss"
-release_date = "2025-08-05"
-last_updated = "2025-08-05"
-attachment = false
-reasoning = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.15
@@ -19,10 +14,6 @@ output = 0.60
 context = 128_000
 output = 16_384
 
-[modalities]
-input = ["text"]
-output = ["text"]
-
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
 api = "https://bedrock-mantle.${AWS_REGION}.api.aws/v1"
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-20b-1:0.toml b/providers/amazon-bedrock/models/openai.gpt-oss-20b-1:0.toml
index 89bf253eb86..bdeca4e6772 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-20b-1:0.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-20b-1:0.toml
@@ -1,15 +1,11 @@
+# Model ID, Runtime APIs and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-20b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high (Converse).
+# Request mapping: https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+# AWS documents 16K max output. Converse accepts larger maxTokens settings up to the 128K context boundary, but that is not evidence of a larger generation cap.
+base_model = "openai/gpt-oss-20b"
 name = "gpt-oss-20b"
-description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-family = "gpt-oss"
-release_date = "2025-08-05"
-last_updated = "2025-08-05"
-attachment = false
-reasoning = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.07
@@ -18,7 +14,3 @@ output = 0.30
 [limit]
 context = 128_000
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-20b.toml b/providers/amazon-bedrock/models/openai.gpt-oss-20b.toml
index 372974d09e9..be6687cd65f 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-20b.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-20b.toml
@@ -1,15 +1,10 @@
+# Model ID, Mantle /v1/responses route and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-20b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Effort: reasoning.effort = low|medium|high (Responses).
+# Native controls: https://huggingface.co/openai/gpt-oss-20b
+base_model = "openai/gpt-oss-20b"
 name = "gpt-oss-20b"
-description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-family = "gpt-oss"
-release_date = "2025-08-05"
-last_updated = "2025-08-05"
-attachment = false
-reasoning = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.07
@@ -19,10 +14,6 @@ output = 0.30
 context = 128_000
 output = 16_384
 
-[modalities]
-input = ["text"]
-output = ["text"]
-
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
 api = "https://bedrock-mantle.${AWS_REGION}.api.aws/v1"
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-120b.toml b/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-120b.toml
index e038add877c..02cb1930634 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-120b.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-120b.toml
@@ -1,14 +1,11 @@
-name = "GPT OSS Safeguard 120B"
-description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2025-10-29"
-last_updated = "2025-10-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
+# Model ID, Runtime Converse support and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-safeguard-120b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Native reasoning controls: https://huggingface.co/openai/gpt-oss-safeguard-120b
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high, following the GPT-OSS Converse request mapping.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+# Live Converse check 2026-09-09 (us-east-1): reasoning_effort=low returns reasoningContent.
+base_model = "openai/gpt-oss-safeguard-120b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 0.15
@@ -17,7 +14,3 @@ output = 0.60
 [limit]
 context = 128_000
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-20b.toml b/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-20b.toml
index 42d743af8aa..9cd4302fbf1 100644
--- a/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-20b.toml
+++ b/providers/amazon-bedrock/models/openai.gpt-oss-safeguard-20b.toml
@@ -1,14 +1,11 @@
-name = "GPT OSS Safeguard 20B"
-description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2025-10-29"
-last_updated = "2025-10-29"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
+# Model ID, Runtime Converse support and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-safeguard-20b.html
+# Standard US East (N. Virginia) pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-east-1/index.json
+# Native reasoning controls: https://huggingface.co/openai/gpt-oss-safeguard-20b
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high, following the GPT-OSS Converse request mapping.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+# Live Converse check 2026-09-09 (us-east-1): reasoning_effort=low returns reasoningContent.
+base_model = "openai/gpt-oss-safeguard-20b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 0.07
@@ -17,7 +14,3 @@ output = 0.20
 [limit]
 context = 128_000
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-120b-1:0.toml b/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-120b-1:0.toml
new file mode 100644
index 00000000000..80e2dabaa04
--- /dev/null
+++ b/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-120b-1:0.toml
@@ -0,0 +1,15 @@
+# GovCloud cross-Region profile (Converse), exact ID, regions and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-120b.html
+# Standard us-gov-west-1 pricing (USD/1K multiplied by 1,000): https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-gov-west-1/index.json
+# Geo inference uses source-region pricing with no routing surcharge: https://docs.aws.amazon.com/bedrock/latest/userguide/cross-region-inference.html
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high; https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+base_model = "openai/gpt-oss-120b"
+name = "gpt-oss-120b (GovCloud)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.18
+output = 0.72
+
+[limit]
+context = 128_000
+output = 16_384
diff --git a/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-20b-1:0.toml b/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-20b-1:0.toml
new file mode 100644
index 00000000000..b0ae3c304d7
--- /dev/null
+++ b/providers/amazon-bedrock/models/us-gov.openai.gpt-oss-20b-1:0.toml
@@ -0,0 +1,15 @@
+# GovCloud cross-Region profile (Converse), exact ID, regions and limits: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-oss-20b.html
+# Standard us-gov-west-1 pricing (USD/1K multiplied by 1,000): https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/us-gov-west-1/index.json
+# Geo inference uses source-region pricing with no routing surcharge: https://docs.aws.amazon.com/bedrock/latest/userguide/cross-region-inference.html
+# Effort: additionalModelRequestFields.reasoning_effort = low|medium|high; https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-openai.html
+base_model = "openai/gpt-oss-20b"
+name = "gpt-oss-20b (GovCloud)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.084
+output = 0.36
+
+[limit]
+context = 128_000
+output = 16_384
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
index 98e3214639b..e0c96587bb0 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
@@ -1,12 +1,14 @@
 # US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
-# Geo CRIS pricing (= in-region) per the GPT-5.6 AWS card pattern; OpenAI-on-Bedrock is absent from the Price List API.
+# Geo CRIS pricing (= in-region) and Runtime structured outputs: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
+# Cache rates: AWS-published Responses pricing; live Converse checks 2026-09-09 reported 6,650 cache-write then cache-read tokens.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max; reasoning_effort is rejected.
+# Live check 2026-09-09 (us-east-1): outputConfig.textFormat enforces JSON schema despite a conflicting user instruction.
 # Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-luna.toml
 base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-5.6 Luna (US)"
-structured_output = false
 
 [cost]
 input = 0.22
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
index e492f79ec73..326e35c16da 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
@@ -1,6 +1,8 @@
 # US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
-# Geo CRIS pricing (= in-region) per the GPT-5.6 Sol AWS card; OpenAI-on-Bedrock is absent from the Price List API.
+# Geo CRIS pricing (= in-region): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max.
 # Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-sol.toml
 base_model = "openai/gpt-5.6-sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
index 12add6584d1..cff259b2548 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
@@ -1,6 +1,8 @@
 # US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
 # Profile ID verified via ListInferenceProfiles from us-east-1 and ca-central-1.
-# Geo CRIS pricing (= in-region) per the GPT-5.6 AWS card pattern; OpenAI-on-Bedrock is absent from the Price List API.
+# Geo CRIS pricing (= in-region): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max.
 # Effort levels incl. "none" verified on the Global twin: global.openai.gpt-5.6-terra.toml
 base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
index eabeed2aede..272d4a5ed09 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
@@ -1,6 +1,8 @@
 # US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
-# Bedrock model IDs and routes: https://github.com/openai/codex/pull/42805
-# Geo CRIS pricing inferred at 1.1x OpenAI-direct (= in-region) from the GPT-5.6 Sol AWS card precedent.
+# Bedrock model IDs, routes, limits and Geo CRIS pricing: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
+# Cache rates: AWS-published Responses pricing; the card's API coverage can lag live behavior.
+# Converse effort: additionalModelRequestFields.reasoning.effort = low|medium|high|xhigh|max.
+# Live check 2026-09-09 (us-east-1): outputConfig.textFormat enforces JSON schema despite a conflicting user instruction.
 # Effort levels match OpenAI-native Astra (Low..Max, no "none"): https://developers.openai.com/api/docs/models/gpt-6-astra
 base_model = "openai/gpt-6-astra"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]

From 408a8d0f5c99fc4274526f375a8288942ef44d1c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:00 +0000
Subject: [PATCH 1081/2012] chore(sync): update EmpirioLabs AI model catalog
 (#6844)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/gemma-3-27b.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/empiriolabs/models/gemma-3-27b.toml

diff --git a/providers/empiriolabs/models/gemma-3-27b.toml b/providers/empiriolabs/models/gemma-3-27b.toml
new file mode 100644
index 00000000000..caeb9f37cb3
--- /dev/null
+++ b/providers/empiriolabs/models/gemma-3-27b.toml
@@ -0,0 +1,12 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B"
+tool_call = false
+structured_output = true
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 128_000
+output = 8_192

From 4d533d51029784122c7366409627fc1a3afbd4b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:02 +0000
Subject: [PATCH 1082/2012] chore(sync): update Kilo model catalog (#6845)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/google/gemma-3-12b-it.toml | 15 +--------------
 providers/kilo/models/google/gemma-3-27b-it.toml | 15 +--------------
 providers/kilo/models/google/gemma-3-4b-it.toml  | 14 +-------------
 .../nvidia/nemotron-3-super-120b-a12b.toml       |  4 ++--
 .../models/openai/gpt-oss-safeguard-20b.toml     | 16 +---------------
 .../kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml      |  6 +++---
 7 files changed, 12 insertions(+), 64 deletions(-)

diff --git a/providers/kilo/models/google/gemma-3-12b-it.toml b/providers/kilo/models/google/gemma-3-12b-it.toml
index cdccfeef0ea..df1bb400e98 100644
--- a/providers/kilo/models/google/gemma-3-12b-it.toml
+++ b/providers/kilo/models/google/gemma-3-12b-it.toml
@@ -1,23 +1,10 @@
-name = "Google: Gemma 3 12B"
+base_model = "google/gemma-3-12b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-13"
-last_updated = "2025-03-13"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.05
 output = 0.15
 
 [limit]
-context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/google/gemma-3-27b-it.toml b/providers/kilo/models/google/gemma-3-27b-it.toml
index 30df275dd92..be1dacf09a9 100644
--- a/providers/kilo/models/google/gemma-3-27b-it.toml
+++ b/providers/kilo/models/google/gemma-3-27b-it.toml
@@ -1,23 +1,10 @@
-name = "Google: Gemma 3 27B"
+base_model = "google/gemma-3-27b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.08
 output = 0.16
 
 [limit]
-context = 131_072
 output = 117_964
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/google/gemma-3-4b-it.toml b/providers/kilo/models/google/gemma-3-4b-it.toml
index 469a5f8c8ea..45f5a1839e2 100644
--- a/providers/kilo/models/google/gemma-3-4b-it.toml
+++ b/providers/kilo/models/google/gemma-3-4b-it.toml
@@ -1,23 +1,11 @@
-name = "Google: Gemma 3 4B"
+base_model = "google/gemma-3-4b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-13"
-last_updated = "2025-03-13"
-attachment = true
-reasoning = false
-temperature = true
 tool_call = false
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.05
 output = 0.1
 
 [limit]
-context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
index f3a2124746d..f64d74bbe36 100644
--- a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "medium"]
 
 [cost]
-input = 0.085
-output = 0.4
+input = 0.08
+output = 0.45
 
 [limit]
 output = 16_384
diff --git a/providers/kilo/models/openai/gpt-oss-safeguard-20b.toml b/providers/kilo/models/openai/gpt-oss-safeguard-20b.toml
index f820fb3c589..1602ee7bf0a 100644
--- a/providers/kilo/models/openai/gpt-oss-safeguard-20b.toml
+++ b/providers/kilo/models/openai/gpt-oss-safeguard-20b.toml
@@ -1,14 +1,5 @@
-name = "OpenAI: gpt-oss-safeguard-20b"
+base_model = "openai/gpt-oss-safeguard-20b"
 description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2025-10-29"
-last_updated = "2025-10-29"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -20,9 +11,4 @@ output = 0.3
 cache_read = 0.0375
 
 [limit]
-context = 131_072
 output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 74831ee82da..4ff6f302821 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 6e543d6cf2b..7c4f6bc8390 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1
-output = 3.41
-cache_read = 0.2
+input = 0.9
+output = 3.069
+cache_read = 0.18
 
 [limit]
 context = 1_048_576

From d320372663434756d8c56034ae0003b33b56038b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:07 +0000
Subject: [PATCH 1083/2012] chore(sync): update Merge Gateway model catalog
 (#6847)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/gemma-3-12b-it.toml             | 12 ++++++++++++
 .../models/google/gemma-3-27b-it.toml             | 15 +++++++++++++++
 .../models/google/gemma-3-4b-it.toml              | 12 ++++++++++++
 .../models/openai/gpt-oss-safeguard-20b.toml      | 14 ++++++++++++++
 4 files changed, 53 insertions(+)
 create mode 100644 providers/merge-gateway/models/google/gemma-3-12b-it.toml
 create mode 100644 providers/merge-gateway/models/google/gemma-3-27b-it.toml
 create mode 100644 providers/merge-gateway/models/google/gemma-3-4b-it.toml
 create mode 100644 providers/merge-gateway/models/openai/gpt-oss-safeguard-20b.toml

diff --git a/providers/merge-gateway/models/google/gemma-3-12b-it.toml b/providers/merge-gateway/models/google/gemma-3-12b-it.toml
new file mode 100644
index 00000000000..3cc24f0d6df
--- /dev/null
+++ b/providers/merge-gateway/models/google/gemma-3-12b-it.toml
@@ -0,0 +1,12 @@
+base_model = "google/gemma-3-12b-it"
+name = "Gemma 3 12B"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.09
+output = 0.29
+
+[limit]
+context = 128_000
+output = 8_192
diff --git a/providers/merge-gateway/models/google/gemma-3-27b-it.toml b/providers/merge-gateway/models/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..8bbd7ddc235
--- /dev/null
+++ b/providers/merge-gateway/models/google/gemma-3-27b-it.toml
@@ -0,0 +1,15 @@
+base_model = "google/gemma-3-27b-it"
+attachment = false
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.08
+output = 0.45
+cache_read = 0.04
+
+[limit]
+output = 32_768
+
+[modalities]
+input = ["text"]
diff --git a/providers/merge-gateway/models/google/gemma-3-4b-it.toml b/providers/merge-gateway/models/google/gemma-3-4b-it.toml
new file mode 100644
index 00000000000..fb82eebfbed
--- /dev/null
+++ b/providers/merge-gateway/models/google/gemma-3-4b-it.toml
@@ -0,0 +1,12 @@
+base_model = "google/gemma-3-4b-it"
+name = "Gemma 3 4B"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.04
+output = 0.08
+
+[limit]
+context = 128_000
+output = 8_192
diff --git a/providers/merge-gateway/models/openai/gpt-oss-safeguard-20b.toml b/providers/merge-gateway/models/openai/gpt-oss-safeguard-20b.toml
new file mode 100644
index 00000000000..a60a70b9ef9
--- /dev/null
+++ b/providers/merge-gateway/models/openai/gpt-oss-safeguard-20b.toml
@@ -0,0 +1,14 @@
+base_model = "openai/gpt-oss-safeguard-20b"
+tool_call = false
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 0.07
+output = 0.2
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 4_096
+output = 4_096

From e16eb71996dac192de0aeee5859711e3b98d380f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:08 +0000
Subject: [PATCH 1084/2012] chore(sync): update Hugging Face model catalog
 (#6850)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/huggingface/models/google/gemma-3-12b-it.toml | 7 +++++++
 providers/huggingface/models/google/gemma-3-27b-it.toml | 7 +++++++
 providers/huggingface/models/google/gemma-3-4b-it.toml  | 7 +++++++
 3 files changed, 21 insertions(+)
 create mode 100644 providers/huggingface/models/google/gemma-3-12b-it.toml
 create mode 100644 providers/huggingface/models/google/gemma-3-27b-it.toml
 create mode 100644 providers/huggingface/models/google/gemma-3-4b-it.toml

diff --git a/providers/huggingface/models/google/gemma-3-12b-it.toml b/providers/huggingface/models/google/gemma-3-12b-it.toml
new file mode 100644
index 00000000000..0b84591264d
--- /dev/null
+++ b/providers/huggingface/models/google/gemma-3-12b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-12b-it"
+description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.15
diff --git a/providers/huggingface/models/google/gemma-3-27b-it.toml b/providers/huggingface/models/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..92928c254f8
--- /dev/null
+++ b/providers/huggingface/models/google/gemma-3-27b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-27b-it"
+description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
+structured_output = true
+
+[cost]
+input = 0.08
+output = 0.16
diff --git a/providers/huggingface/models/google/gemma-3-4b-it.toml b/providers/huggingface/models/google/gemma-3-4b-it.toml
new file mode 100644
index 00000000000..17d0868d8ba
--- /dev/null
+++ b/providers/huggingface/models/google/gemma-3-4b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-4b-it"
+description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.1

From c11ac22133e99c3f63be01f57d82513bd966a5e8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:11 +0000
Subject: [PATCH 1085/2012] chore(sync): update OpenRouter model catalog
 (#6848)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/google/gemma-3-12b-it.toml | 16 +---------------
 .../openrouter/models/google/gemma-3-27b-it.toml | 16 +---------------
 .../openrouter/models/google/gemma-3-4b-it.toml  | 15 +--------------
 .../openrouter/models/moonshotai/kimi-k3.toml    |  6 +++---
 .../models/openai/gpt-oss-safeguard-20b.toml     | 16 +---------------
 .../models/~moonshotai/kimi-latest.toml          |  6 +++---
 .../openrouter/models/~z-ai/glm-latest.toml      |  6 +++---
 7 files changed, 13 insertions(+), 68 deletions(-)

diff --git a/providers/openrouter/models/google/gemma-3-12b-it.toml b/providers/openrouter/models/google/gemma-3-12b-it.toml
index 8db8ed43f0b..df1bb400e98 100644
--- a/providers/openrouter/models/google/gemma-3-12b-it.toml
+++ b/providers/openrouter/models/google/gemma-3-12b-it.toml
@@ -1,24 +1,10 @@
-name = "Gemma 3 12B"
+base_model = "google/gemma-3-12b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-13"
-last_updated = "2025-03-13"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-knowledge = "2024-08-31"
-open_weights = true
 
 [cost]
 input = 0.05
 output = 0.15
 
 [limit]
-context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/google/gemma-3-27b-it.toml b/providers/openrouter/models/google/gemma-3-27b-it.toml
index f76ae331963..617f624675d 100644
--- a/providers/openrouter/models/google/gemma-3-27b-it.toml
+++ b/providers/openrouter/models/google/gemma-3-27b-it.toml
@@ -1,15 +1,6 @@
-name = "Gemma 3 27B"
+base_model = "google/gemma-3-27b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-knowledge = "2024-08-31"
-open_weights = true
 
 [cost]
 input = 0.08
@@ -17,9 +8,4 @@ output = 0.45
 cache_read = 0.04
 
 [limit]
-context = 131_072
 output = 117_964
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/google/gemma-3-4b-it.toml b/providers/openrouter/models/google/gemma-3-4b-it.toml
index aa109558138..45f5a1839e2 100644
--- a/providers/openrouter/models/google/gemma-3-4b-it.toml
+++ b/providers/openrouter/models/google/gemma-3-4b-it.toml
@@ -1,24 +1,11 @@
-name = "Gemma 3 4B"
+base_model = "google/gemma-3-4b-it"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-03-13"
-last_updated = "2025-03-13"
-attachment = true
-reasoning = false
-temperature = true
 tool_call = false
 structured_output = true
-knowledge = "2024-08-31"
-open_weights = true
 
 [cost]
 input = 0.05
 output = 0.1
 
 [limit]
-context = 131_072
 output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 975660fffb4..c831cb49220 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/openai/gpt-oss-safeguard-20b.toml b/providers/openrouter/models/openai/gpt-oss-safeguard-20b.toml
index e4898e7fedb..165dfd21c6a 100644
--- a/providers/openrouter/models/openai/gpt-oss-safeguard-20b.toml
+++ b/providers/openrouter/models/openai/gpt-oss-safeguard-20b.toml
@@ -1,14 +1,5 @@
-name = "gpt-oss-safeguard-20b"
+base_model = "openai/gpt-oss-safeguard-20b"
 description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2025-10-29"
-last_updated = "2025-10-29"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [[reasoning_options]]
 type = "effort"
@@ -20,9 +11,4 @@ output = 0.3
 cache_read = 0.0375
 
 [limit]
-context = 131_072
 output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 6eb528f593d..f78c60f8145 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 2.34
+output = 11.7
+cache_read = 0.261
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0818573f975..018aa6c317f 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1
-output = 3.41
-cache_read = 0.2
+input = 0.9
+output = 3.069
+cache_read = 0.18
 
 [limit]
 context = 1_310_720

From d5bab9cbc889e4d19844735a2f35760e5748037f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:16 +0000
Subject: [PATCH 1086/2012] chore(sync): update Venice model catalog (#6842)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/google-gemma-3-27b-it.toml | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

diff --git a/providers/venice/models/google-gemma-3-27b-it.toml b/providers/venice/models/google-gemma-3-27b-it.toml
index 0d02714f02f..a897a57a3f9 100644
--- a/providers/venice/models/google-gemma-3-27b-it.toml
+++ b/providers/venice/models/google-gemma-3-27b-it.toml
@@ -1,13 +1,9 @@
+base_model = "google/gemma-3-27b-it"
 name = "Google Gemma 3 27B Instruct"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
 release_date = "2025-11-04"
 last_updated = "2026-06-11"
-attachment = true
-reasoning = false
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.12
@@ -16,7 +12,3 @@ output = 0.2
 [limit]
 context = 198_000
 output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 5fbacacb20ff8b429ab4aca15055903b7fa34044 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:19 +0000
Subject: [PATCH 1087/2012] chore(sync): update Deep Infra model catalog
 (#6841)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/google/gemma-3-12b-it.toml | 6 ++++++
 providers/deepinfra/models/google/gemma-3-27b-it.toml | 6 ++++++
 providers/deepinfra/models/google/gemma-3-4b-it.toml  | 6 ++++++
 3 files changed, 18 insertions(+)
 create mode 100644 providers/deepinfra/models/google/gemma-3-12b-it.toml
 create mode 100644 providers/deepinfra/models/google/gemma-3-27b-it.toml
 create mode 100644 providers/deepinfra/models/google/gemma-3-4b-it.toml

diff --git a/providers/deepinfra/models/google/gemma-3-12b-it.toml b/providers/deepinfra/models/google/gemma-3-12b-it.toml
new file mode 100644
index 00000000000..0393878e0d8
--- /dev/null
+++ b/providers/deepinfra/models/google/gemma-3-12b-it.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemma-3-12b-it"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.15
diff --git a/providers/deepinfra/models/google/gemma-3-27b-it.toml b/providers/deepinfra/models/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..5ba5b627ac4
--- /dev/null
+++ b/providers/deepinfra/models/google/gemma-3-27b-it.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemma-3-27b-it"
+structured_output = true
+
+[cost]
+input = 0.08
+output = 0.16
diff --git a/providers/deepinfra/models/google/gemma-3-4b-it.toml b/providers/deepinfra/models/google/gemma-3-4b-it.toml
new file mode 100644
index 00000000000..b572440d1f7
--- /dev/null
+++ b/providers/deepinfra/models/google/gemma-3-4b-it.toml
@@ -0,0 +1,6 @@
+base_model = "google/gemma-3-4b-it"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.1

From 9c8d75b145279cb0e08a1f5c37d7d2ccd825e82e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 05:26:35 +0000
Subject: [PATCH 1088/2012] chore(sync): update NanoGPT model catalog (#6843)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-oss-safeguard-20b.toml        | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/providers/nano-gpt/models/openai/gpt-oss-safeguard-20b.toml b/providers/nano-gpt/models/openai/gpt-oss-safeguard-20b.toml
index 3e77c6ed1e7..7500d84f34d 100644
--- a/providers/nano-gpt/models/openai/gpt-oss-safeguard-20b.toml
+++ b/providers/nano-gpt/models/openai/gpt-oss-safeguard-20b.toml
@@ -1,13 +1,6 @@
-name = "GPT OSS Safeguard 20B"
-description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2026-02-23"
-last_updated = "2025-10-29"
-attachment = false
-reasoning = true
+base_model = "openai/gpt-oss-safeguard-20b"
 tool_call = false
 structured_output = false
-open_weights = true
 
 [[reasoning_options]]
 type = "effort"
@@ -21,7 +14,3 @@ output = 0.3
 context = 128_000
 input = 128_000
 output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]

From b82f83323f2a38abf33456c030ac32f29b3dd4bf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 06:39:43 +0000
Subject: [PATCH 1089/2012] chore(sync): update NanoGPT model catalog (#6853)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 9c2641248c0..fb4e3a2dd26 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 2e7355d5588..f23aac53ab8 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From 451281e16fb156d57e7b167a08e8b4959be2d8a0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 06:39:48 +0000
Subject: [PATCH 1090/2012] chore(sync): update Kilo model catalog (#6855)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-chat-v3-0324.toml       |  5 ++--
 providers/kilo/models/sakana/fugu-max.toml    | 28 +++++++++++++++++++
 .../kilo/models/sakana/fugu-ultra-v2.toml     | 28 +++++++++++++++++++
 3 files changed, 58 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/sakana/fugu-max.toml
 create mode 100644 providers/kilo/models/sakana/fugu-ultra-v2.toml

diff --git a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
index e46b521c61f..764c5575e7c 100644
--- a/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat-v3-0324.toml
@@ -11,9 +11,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.29
-output = 1.14
-cache_read = 0.11
+input = 0.25
+output = 1
 
 [limit]
 context = 163_840
diff --git a/providers/kilo/models/sakana/fugu-max.toml b/providers/kilo/models/sakana/fugu-max.toml
new file mode 100644
index 00000000000..eef3beb5bf5
--- /dev/null
+++ b/providers/kilo/models/sakana/fugu-max.toml
@@ -0,0 +1,28 @@
+name = "Sakana: Fugu Max"
+description = "Fugu Max is the cost-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to route..."
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/kilo/models/sakana/fugu-ultra-v2.toml b/providers/kilo/models/sakana/fugu-ultra-v2.toml
new file mode 100644
index 00000000000..6cc322c77c9
--- /dev/null
+++ b/providers/kilo/models/sakana/fugu-ultra-v2.toml
@@ -0,0 +1,28 @@
+name = "Sakana: Fugu Ultra v2"
+description = "Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to..."
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From 3d14ffdc320fd965cc74ca06521ac8f24c0d27d2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 06:39:53 +0000
Subject: [PATCH 1091/2012] chore(sync): update OpenRouter model catalog
 (#6854)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-chat-v3-0324.toml       |  5 ++-
 .../models/deepseek/deepseek-v4.1-flash.toml  |  6 ++--
 .../openrouter/models/sakana/fugu-max.toml    | 28 +++++++++++++++
 .../models/sakana/fugu-ultra-v2.toml          | 35 +++++++++++++++++++
 4 files changed, 68 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/sakana/fugu-max.toml
 create mode 100644 providers/openrouter/models/sakana/fugu-ultra-v2.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
index d8a154006c7..905726abf7a 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat-v3-0324.toml
@@ -12,9 +12,8 @@ knowledge = "2024-07-31"
 open_weights = true
 
 [cost]
-input = 0.29
-output = 1.14
-cache_read = 0.11
+input = 0.25
+output = 1
 
 [limit]
 context = 163_840
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/sakana/fugu-max.toml b/providers/openrouter/models/sakana/fugu-max.toml
new file mode 100644
index 00000000000..ce44a62b12b
--- /dev/null
+++ b/providers/openrouter/models/sakana/fugu-max.toml
@@ -0,0 +1,28 @@
+name = "Fugu Max"
+description = "Multi-agent model for routing expert agents across complex analytical tasks"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/openrouter/models/sakana/fugu-ultra-v2.toml b/providers/openrouter/models/sakana/fugu-ultra-v2.toml
new file mode 100644
index 00000000000..60a24159724
--- /dev/null
+++ b/providers/openrouter/models/sakana/fugu-ultra-v2.toml
@@ -0,0 +1,35 @@
+name = "Fugu Ultra v2"
+description = "Quality-first multi-agent model for hard research, analysis, and competitions"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-08-28"
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 45
+cache_read = 1
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From 0e7564eeab888db785380c358a8f5acb2b94b240 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 07:28:17 +0000
Subject: [PATCH 1092/2012] chore(sync): update Kilo model catalog (#6859)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml     | 3 ++-
 providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..fa4ef3087d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
index f64d74bbe36..a591c13e651 100644
--- a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -1,5 +1,5 @@
 base_model = "nvidia/nemotron-3-super-120b-a12b"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"

From 5347558d217d95e6d9ca23fb62f8e3b9ce291efe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 07:28:32 +0000
Subject: [PATCH 1093/2012] chore(sync): update OpenRouter model catalog
 (#6858)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 ++++---
 .../models/nvidia/nemotron-3-super-120b-a12b.toml          | 2 +-
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5bc4e6af877..30c5b0f60a5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.0494
-output = 3.1482
-cache_read = 0.03498
+input = 0.66
+output = 1.98
+cache_read = 0.066
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
index 9376472a1c3..455636bfb0a 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -1,7 +1,7 @@
 # Toggle: reasoning.enabled = true|false
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "nvidia/nemotron-3-super-120b-a12b"
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "toggle"

From b79f554fdfbe74aa1163f9c208e038dc67f5c0c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 08:33:29 +0000
Subject: [PATCH 1094/2012] chore(sync): update OpenRouter model catalog
 (#6868)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/google/gemma-4-26b-a4b-it.toml  | 3 ---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 3 files changed, 4 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index f8a013291bc..a27c6e610ef 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -9,8 +9,5 @@ type = "toggle"
 input = 0.042
 output = 0.22
 
-[limit]
-context = 1_000_000
-
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 5ab4a3ffd9b..27d5c88cb9e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,7 +20,7 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
+input = 0.05
 output = 0.16
 cache_read = 0.013
 
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 018aa6c317f..be978805f70 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3.069
-cache_read = 0.18
+input = 0.97
+output = 3.3077
+cache_read = 0.194
 
 [limit]
 context = 1_310_720

From bfb0ecdfebf5233cc9cdc8c07960bf23da2cec11 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 08:33:44 +0000
Subject: [PATCH 1095/2012] chore(sync): update Ofox model catalog (#6867)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0423.toml      |  6 ++---
 .../models/deepseek/deepseek-v4-pro-0423.toml |  2 +-
 .../models/google/gemini-2.5-flash-lite.toml  | 14 +++++++----
 providers/ofox/models/qwen/qwen-turbo.toml    |  2 +-
 providers/ofox/models/qwen/qwen-vl-max.toml   |  2 +-
 providers/ofox/models/qwen/qwen3.6-27b.toml   | 11 +++++---
 providers/ofox/models/qwen/qwen3.7-max.toml   | 15 +++++++----
 providers/ofox/models/qwen/qwen3.8-27b.toml   | 15 +++++++----
 providers/ofox/models/qwen/qwen3.8-flash.toml | 15 +++++++----
 .../ofox/models/qwen/qwen3.8-max-0902.toml    | 25 ++++++++++++-------
 providers/ofox/models/qwen/qwen3.8-max.toml   | 25 ++++++++++++-------
 providers/ofox/models/z-ai/glm-4.6.toml       |  8 +++---
 .../ofox/models/z-ai/glm-4.7-flashx.toml      |  8 +++---
 providers/ofox/models/z-ai/glm-4.7.toml       |  8 +++---
 14 files changed, 100 insertions(+), 56 deletions(-)

diff --git a/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml b/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml
index d5d592e6729..0b858dede74 100644
--- a/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml
+++ b/providers/ofox/models/deepseek/deepseek-v4-flash-0423.toml
@@ -13,6 +13,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.19
+output = 0.51
+cache_read = 0.028
diff --git a/providers/ofox/models/deepseek/deepseek-v4-pro-0423.toml b/providers/ofox/models/deepseek/deepseek-v4-pro-0423.toml
index 737b6fefef8..2480d6388a1 100644
--- a/providers/ofox/models/deepseek/deepseek-v4-pro-0423.toml
+++ b/providers/ofox/models/deepseek/deepseek-v4-pro-0423.toml
@@ -14,4 +14,4 @@ values = ["high", "max"]
 [cost]
 input = 1.32
 output = 3.96
-cache_read = 0.044
+cache_read = 0.15
diff --git a/providers/ofox/models/google/gemini-2.5-flash-lite.toml b/providers/ofox/models/google/gemini-2.5-flash-lite.toml
index 67b87706a7c..ae565244def 100644
--- a/providers/ofox/models/google/gemini-2.5-flash-lite.toml
+++ b/providers/ofox/models/google/gemini-2.5-flash-lite.toml
@@ -3,14 +3,18 @@
 # 0 disables; enabled manual budgets 512..24576
 # https://ai.google.dev/gemini-api/docs/thinking
 base_model = "google/gemini-2.5-flash-lite"
-reasoning_options = [
-  { type = "toggle" },
-  { type = "budget_tokens", min = 512, max = 24_576 },
-]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 512
+max = 24_576
 
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.025
+cache_read = 0.01
 cache_write = 1
 input_audio = 0.3
diff --git a/providers/ofox/models/qwen/qwen-turbo.toml b/providers/ofox/models/qwen/qwen-turbo.toml
index af971283622..392aca68ca5 100644
--- a/providers/ofox/models/qwen/qwen-turbo.toml
+++ b/providers/ofox/models/qwen/qwen-turbo.toml
@@ -5,7 +5,7 @@ base_model = "alibaba/qwen-turbo"
 reasoning = false
 
 [cost]
-input = 0.05
+input = 0.043
 output = 0.09
 cache_read = 0.0086
 
diff --git a/providers/ofox/models/qwen/qwen-vl-max.toml b/providers/ofox/models/qwen/qwen-vl-max.toml
index 0692c38b4b0..ba15f7556b3 100644
--- a/providers/ofox/models/qwen/qwen-vl-max.toml
+++ b/providers/ofox/models/qwen/qwen-vl-max.toml
@@ -6,7 +6,7 @@ base_model = "alibaba/qwen-vl-max"
 [cost]
 input = 0.23
 output = 0.58
-cache_read = 0.046
+cache_read = 0.023
 
 [limit]
 context = 128_000
diff --git a/providers/ofox/models/qwen/qwen3.6-27b.toml b/providers/ofox/models/qwen/qwen3.6-27b.toml
index b2ff03427c1..3309ae4d43d 100644
--- a/providers/ofox/models/qwen/qwen3.6-27b.toml
+++ b/providers/ofox/models/qwen/qwen3.6-27b.toml
@@ -4,11 +4,16 @@
 # Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
 # Budget: thinking_budget
 base_model = "alibaba/qwen3.6-27b"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 3.6
+input = 0.43
+output = 2.57
 
 [limit]
 context = 256_000
diff --git a/providers/ofox/models/qwen/qwen3.7-max.toml b/providers/ofox/models/qwen/qwen3.7-max.toml
index c071652b2c2..9a6456417eb 100644
--- a/providers/ofox/models/qwen/qwen3.7-max.toml
+++ b/providers/ofox/models/qwen/qwen3.7-max.toml
@@ -4,13 +4,18 @@
 # Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
 # Budget: thinking_budget
 base_model = "alibaba/qwen3.7-max"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 2.5
-output = 7.5
-cache_read = 0.5
-cache_write = 3.125
+input = 1.71
+output = 5.14
+cache_read = 0.17
+cache_write = 2.14
 
 [limit]
 context = 1_064_000
diff --git a/providers/ofox/models/qwen/qwen3.8-27b.toml b/providers/ofox/models/qwen/qwen3.8-27b.toml
index ea631083ca0..074148d9e8a 100644
--- a/providers/ofox/models/qwen/qwen3.8-27b.toml
+++ b/providers/ofox/models/qwen/qwen3.8-27b.toml
@@ -4,13 +4,18 @@
 # Toggle: enable_thinking true|false (Qwen/DashScope chat path through Ofox)
 # Budget: thinking_budget
 base_model = "alibaba/qwen3.8-27b"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.45
-output = 3.2
-cache_read = 0.05
-cache_write = 0.5625
+input = 0.5
+output = 1.71
+cache_read = 0.043
+cache_write = 0.63
 
 [limit]
 context = 1_131_072
diff --git a/providers/ofox/models/qwen/qwen3.8-flash.toml b/providers/ofox/models/qwen/qwen3.8-flash.toml
index 9e1592e3e53..ecd5475f9ce 100644
--- a/providers/ofox/models/qwen/qwen3.8-flash.toml
+++ b/providers/ofox/models/qwen/qwen3.8-flash.toml
@@ -6,10 +6,15 @@
 # Toggle: enable_thinking true|false (hybrid; Qwen/DashScope surface through Ofox)
 # Budget: thinking_budget
 base_model = "alibaba/qwen3.8-flash"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.15
-output = 0.47
-cache_read = 0.016
-cache_write = 0.2
+input = 0.11
+output = 0.39
+cache_read = 0.011
+cache_write = 0.14
diff --git a/providers/ofox/models/qwen/qwen3.8-max-0902.toml b/providers/ofox/models/qwen/qwen3.8-max-0902.toml
index 50d7d8338e3..fde94b3d0a8 100644
--- a/providers/ofox/models/qwen/qwen3.8-max-0902.toml
+++ b/providers/ofox/models/qwen/qwen3.8-max-0902.toml
@@ -8,14 +8,21 @@
 # Effort: reasoning_effort = low|medium|xhigh (default xhigh); high accepted as alias → xhigh
 # Budget: thinking_budget (0..262144) cannot be combined with reasoning_effort
 base_model = "alibaba/qwen3.8-max-0902"
-reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["low", "medium", "xhigh"] },
-  { type = "budget_tokens", min = 0, max = 262_144 },
-]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 0
+max = 262_144
 
 [cost]
-input = 2.00
-output = 6.00
-cache_read = 0.25
-cache_write = 2.5
+input = 1.71
+output = 5.14
+cache_read = 0.17
+cache_write = 2.14
diff --git a/providers/ofox/models/qwen/qwen3.8-max.toml b/providers/ofox/models/qwen/qwen3.8-max.toml
index 733f1266841..6c0c7e2b1eb 100644
--- a/providers/ofox/models/qwen/qwen3.8-max.toml
+++ b/providers/ofox/models/qwen/qwen3.8-max.toml
@@ -7,14 +7,21 @@
 # Effort: reasoning_effort = low|medium|xhigh (default xhigh); high accepted as alias → xhigh
 # Budget: thinking_budget (0..262144) cannot be combined with reasoning_effort
 base_model = "alibaba/qwen3.8-max"
-reasoning_options = [
-  { type = "toggle" },
-  { type = "effort", values = ["low", "medium", "xhigh"] },
-  { type = "budget_tokens", min = 0, max = 262_144 },
-]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 0
+max = 262_144
 
 [cost]
-input = 2.00
-output = 6.00
-cache_read = 0.25
-cache_write = 2.5
+input = 1.71
+output = 5.14
+cache_read = 0.17
+cache_write = 2.14
diff --git a/providers/ofox/models/z-ai/glm-4.6.toml b/providers/ofox/models/z-ai/glm-4.6.toml
index 85a42ae6558..f9182a89012 100644
--- a/providers/ofox/models/z-ai/glm-4.6.toml
+++ b/providers/ofox/models/z-ai/glm-4.6.toml
@@ -2,9 +2,11 @@
 # https://api.ofox.ai/v1/models/z-ai/glm-4.6
 # Toggle: thinking.type = enabled|disabled (Z.ai native; Ofox lists supported_parameters: reasoning)
 base_model = "zhipuai/glm-4.6"
-reasoning_options = [{ type = "toggle" }]
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.4
-output = 1.9
+input = 0.6
+output = 2.2
 cache_read = 0.11
diff --git a/providers/ofox/models/z-ai/glm-4.7-flashx.toml b/providers/ofox/models/z-ai/glm-4.7-flashx.toml
index f3bbe1f38aa..d5e52076c23 100644
--- a/providers/ofox/models/z-ai/glm-4.7-flashx.toml
+++ b/providers/ofox/models/z-ai/glm-4.7-flashx.toml
@@ -5,12 +5,14 @@
 # - https://docs.z.ai/guides/overview/pricing (lab list price reference)
 # Toggle: thinking.type = enabled|disabled
 base_model = "zhipuai/glm-4.7-flashx"
-reasoning_options = [{ type = "toggle" }]
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.072
-output = 0.43
-cache_read = 0.015
+output = 0.4
+cache_read = 0.01
 
 [limit]
 output = 128_000
diff --git a/providers/ofox/models/z-ai/glm-4.7.toml b/providers/ofox/models/z-ai/glm-4.7.toml
index 73b6b2c494b..ff7a89f7aac 100644
--- a/providers/ofox/models/z-ai/glm-4.7.toml
+++ b/providers/ofox/models/z-ai/glm-4.7.toml
@@ -1,9 +1,11 @@
 # https://ofox.ai/models/z-ai/glm-4.7
 # Toggle: thinking.type = enabled|disabled (Z.AI first-party / peers)
 base_model = "zhipuai/glm-4.7"
-reasoning_options = [{ type = "toggle" }]
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
 input = 0.4
-output = 2
-cache_read = 0.08
+output = 2.2
+cache_read = 0.11

From cc0f4f77267afd5aafe4a76c00032a99eab4d793 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 08:33:49 +0000
Subject: [PATCH 1096/2012] chore(sync): update Kilo model catalog (#6865)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 2 +-
 providers/kilo/models/~z-ai/glm-latest.toml                 | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index bd0a4217a38..80dc12b4ce8 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,7 +15,7 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
+input = 0.05
 output = 0.16
 cache_read = 0.013
 
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7c4f6bc8390..550d2e46fef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3.069
-cache_read = 0.18
+input = 0.97
+output = 3.3077
+cache_read = 0.194
 
 [limit]
 context = 1_048_576

From cf1588b90a020193d32d58544941de8f3b8272e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 09:28:34 +0000
Subject: [PATCH 1097/2012] chore(sync): update Requesty model catalog (#6869)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/deepseek-v4.1-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/requesty/models/deepseek-v4.1-flash.toml b/providers/requesty/models/deepseek-v4.1-flash.toml
index cf1db66a44a..83545096971 100644
--- a/providers/requesty/models/deepseek-v4.1-flash.toml
+++ b/providers/requesty/models/deepseek-v4.1-flash.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From f6d8ac98dad6e01b8a3c65f3a7f77acc47910742 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 09:28:40 +0000
Subject: [PATCH 1098/2012] chore(sync): update NanoGPT model catalog (#6866)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/lightonai/LightOnOCR-2-1B.toml     | 22 +++++++++++++++
 .../meta-llama/llama-3.1-405b-instruct.toml   | 23 +++++++++++++++
 .../meta-llama/llama-3.1-8b-instruct.toml     |  1 -
 .../mistral-small-24b-instruct-2501.toml      | 23 +++++++++++++++
 .../nano-gpt/models/sakana/fugu-max.toml      | 28 +++++++++++++++++++
 .../models/sakana/fugu-ultra-v1.1.toml        |  6 ++--
 .../nano-gpt/models/sakana/fugu-ultra.toml    |  6 ++--
 7 files changed, 102 insertions(+), 7 deletions(-)
 create mode 100644 providers/nano-gpt/models/lightonai/LightOnOCR-2-1B.toml
 create mode 100644 providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
 create mode 100644 providers/nano-gpt/models/mistralai/mistral-small-24b-instruct-2501.toml
 create mode 100644 providers/nano-gpt/models/sakana/fugu-max.toml

diff --git a/providers/nano-gpt/models/lightonai/LightOnOCR-2-1B.toml b/providers/nano-gpt/models/lightonai/LightOnOCR-2-1B.toml
new file mode 100644
index 00000000000..f59e8806014
--- /dev/null
+++ b/providers/nano-gpt/models/lightonai/LightOnOCR-2-1B.toml
@@ -0,0 +1,22 @@
+name = "LightOnOCR 2"
+description = "LightOnOCR 2 hosted by IONOS in Berlin, Germany. Zero data retention."
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.1785
+output = 0.3465
+
+[limit]
+context = 32_768
+input = 32_768
+output = 8_192
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml b/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
new file mode 100644
index 00000000000..7d947439c26
--- /dev/null
+++ b/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
@@ -0,0 +1,23 @@
+name = "Llama 3.1 405B"
+description = "Llama 3.1 405B hosted by IONOS in Berlin, Germany. Zero data retention."
+family = "llama"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 2.0265
+output = 2.0265
+
+[limit]
+context = 131_072
+input = 131_072
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/meta-llama/llama-3.1-8b-instruct.toml b/providers/nano-gpt/models/meta-llama/llama-3.1-8b-instruct.toml
index b17eb59e0f9..2ebca8d6c88 100644
--- a/providers/nano-gpt/models/meta-llama/llama-3.1-8b-instruct.toml
+++ b/providers/nano-gpt/models/meta-llama/llama-3.1-8b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.1-8b-instruct"
 name = "Llama 3.1 8b Instruct"
-tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/nano-gpt/models/mistralai/mistral-small-24b-instruct-2501.toml b/providers/nano-gpt/models/mistralai/mistral-small-24b-instruct-2501.toml
new file mode 100644
index 00000000000..09c4689bf49
--- /dev/null
+++ b/providers/nano-gpt/models/mistralai/mistral-small-24b-instruct-2501.toml
@@ -0,0 +1,23 @@
+name = "Mistral Small 24B"
+description = "Mistral Small 24B hosted by IONOS in Berlin, Germany. Zero data retention."
+family = "mistral-small"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.1155
+output = 0.3465
+
+[limit]
+context = 32_768
+input = 32_768
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/sakana/fugu-max.toml b/providers/nano-gpt/models/sakana/fugu-max.toml
new file mode 100644
index 00000000000..20eb479f0a7
--- /dev/null
+++ b/providers/nano-gpt/models/sakana/fugu-max.toml
@@ -0,0 +1,28 @@
+name = "Fugu Max"
+description = "Sakana AI's cost-performance Fugu model uses learned multi-agent orchestration to route tasks across expert models for reasoning, coding, and tool use."
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/sakana/fugu-ultra-v1.1.toml b/providers/nano-gpt/models/sakana/fugu-ultra-v1.1.toml
index 4e164b14478..b7918f1a026 100644
--- a/providers/nano-gpt/models/sakana/fugu-ultra-v1.1.toml
+++ b/providers/nano-gpt/models/sakana/fugu-ultra-v1.1.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["high", "xhigh", "max"]
 
 [cost]
-input = 5.25
-output = 31.5
-cache_read = 0.525
+input = 5
+output = 30
+cache_read = 0.5
 
 [limit]
 context = 1_000_000
diff --git a/providers/nano-gpt/models/sakana/fugu-ultra.toml b/providers/nano-gpt/models/sakana/fugu-ultra.toml
index d163827b235..c2429506bcc 100644
--- a/providers/nano-gpt/models/sakana/fugu-ultra.toml
+++ b/providers/nano-gpt/models/sakana/fugu-ultra.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["high", "xhigh", "max"]
 
 [cost]
-input = 5.25
-output = 31.5
-cache_read = 0.525
+input = 5
+output = 30
+cache_read = 0.5
 
 [limit]
 input = 1_000_000

From 2c2e208d2d91b6ba8a9ed26c3150d591a55bec52 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:26:52 +0000
Subject: [PATCH 1099/2012] chore(sync): update Kilo model catalog (#6871)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 6fc67901197..881a90e0d15 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384

From c9d71d756390795e6efa94bbd46635d3c3a9af40 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:26:54 +0000
Subject: [PATCH 1100/2012] chore(sync): update NanoGPT model catalog (#6870)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index fb4e3a2dd26..9c2641248c0 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index f23aac53ab8..2e7355d5588 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From b8244af8bd1ca8c0e60103f058854d5510a2534d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:27:31 +0000
Subject: [PATCH 1101/2012] chore(sync): update OpenRouter model catalog
 (#6872)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml           | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 1352c1ee599..881a90e0d15 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.72
-output = 0.72
+input = 0.4
+output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384

From deabc4b5cea3222c08d09270c264706685b59207 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 12:36:32 +0000
Subject: [PATCH 1102/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6877)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/fugu-max.toml     | 28 +++++++++++++++++++
 .../llmgateway/models/fugu-ultra-v2.0.toml    | 28 +++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 providers/llmgateway/models/fugu-max.toml
 create mode 100644 providers/llmgateway/models/fugu-ultra-v2.0.toml

diff --git a/providers/llmgateway/models/fugu-max.toml b/providers/llmgateway/models/fugu-max.toml
new file mode 100644
index 00000000000..0133ae6d67e
--- /dev/null
+++ b/providers/llmgateway/models/fugu-max.toml
@@ -0,0 +1,28 @@
+name = "Fugu Max"
+description = "Multi-agent model for routing expert agents across complex analytical tasks"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/llmgateway/models/fugu-ultra-v2.0.toml b/providers/llmgateway/models/fugu-ultra-v2.0.toml
new file mode 100644
index 00000000000..6d261e1eb2f
--- /dev/null
+++ b/providers/llmgateway/models/fugu-ultra-v2.0.toml
@@ -0,0 +1,28 @@
+name = "Fugu Ultra v2.0"
+description = "Quality-first multi-agent model for hard research, analysis, and competitions"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 0e55dfb6b126ae08c17f4c2eb108501b8423e619 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 12:37:16 +0000
Subject: [PATCH 1103/2012] chore(sync): update LLM Gateway model catalog
 (#6878)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/sakana/fugu-max.toml               | 28 +++++++++++++++++++
 .../models/sakana/fugu-ultra-v2.0.toml        | 28 +++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/sakana/fugu-max.toml
 create mode 100644 providers/llmgateway-providers/models/sakana/fugu-ultra-v2.0.toml

diff --git a/providers/llmgateway-providers/models/sakana/fugu-max.toml b/providers/llmgateway-providers/models/sakana/fugu-max.toml
new file mode 100644
index 00000000000..13a1b385939
--- /dev/null
+++ b/providers/llmgateway-providers/models/sakana/fugu-max.toml
@@ -0,0 +1,28 @@
+name = "Fugu Max (Sakana AI)"
+description = "Multi-agent model for routing expert agents across complex analytical tasks"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/llmgateway-providers/models/sakana/fugu-ultra-v2.0.toml b/providers/llmgateway-providers/models/sakana/fugu-ultra-v2.0.toml
new file mode 100644
index 00000000000..b88b4aa0fc6
--- /dev/null
+++ b/providers/llmgateway-providers/models/sakana/fugu-ultra-v2.0.toml
@@ -0,0 +1,28 @@
+name = "Fugu Ultra v2.0 (Sakana AI)"
+description = "Quality-first multi-agent model for hard research, analysis, and competitions"
+family = "fugu"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 55fa8a24a0af0115fb2287fc42581bcdbbd30e18 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 13:26:27 +0000
Subject: [PATCH 1104/2012] chore(sync): update Kilo model catalog (#6880)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/inclusionai/ling-3.0-flash-vl.toml | 28 ++++++++++++++++++
 .../kilo/models/~openai/gpt-astra-latest.toml | 29 +++++++++++++++++++
 .../kilo/models/~openai/gpt-luna-latest.toml  | 29 +++++++++++++++++++
 .../kilo/models/~openai/gpt-sol-latest.toml   | 29 +++++++++++++++++++
 .../kilo/models/~openai/gpt-terra-latest.toml | 29 +++++++++++++++++++
 5 files changed, 144 insertions(+)
 create mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
 create mode 100644 providers/kilo/models/~openai/gpt-astra-latest.toml
 create mode 100644 providers/kilo/models/~openai/gpt-luna-latest.toml
 create mode 100644 providers/kilo/models/~openai/gpt-sol-latest.toml
 create mode 100644 providers/kilo/models/~openai/gpt-terra-latest.toml

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
new file mode 100644
index 00000000000..4ea1e77c01f
--- /dev/null
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
@@ -0,0 +1,28 @@
+name = "inclusionAI: Ling 3.0 Flash VL"
+description = "Ling 3.0 Flash VL builds on Ling 3.0 Flash (124B total / 5.5B active MoE from InclusionAI), further strengthening its language capabilities while adding native visual perception and advanced visual..."
+family = "ling"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.06
+output = 0.18
+cache_read = 0.012
+
+[limit]
+context = 131_072
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/kilo/models/~openai/gpt-astra-latest.toml b/providers/kilo/models/~openai/gpt-astra-latest.toml
new file mode 100644
index 00000000000..cc33c625eda
--- /dev/null
+++ b/providers/kilo/models/~openai/gpt-astra-latest.toml
@@ -0,0 +1,29 @@
+name = "OpenAI GPT Astra Latest ($$$$)"
+description = "This model always redirects to the latest model in the OpenAI GPT Astra family."
+family = "gpt-astra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/~openai/gpt-luna-latest.toml b/providers/kilo/models/~openai/gpt-luna-latest.toml
new file mode 100644
index 00000000000..aa44302cd2e
--- /dev/null
+++ b/providers/kilo/models/~openai/gpt-luna-latest.toml
@@ -0,0 +1,29 @@
+name = "OpenAI GPT Luna Latest"
+description = "This model always redirects to the latest model in the OpenAI GPT Luna family."
+family = "gpt-luna"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/~openai/gpt-sol-latest.toml b/providers/kilo/models/~openai/gpt-sol-latest.toml
new file mode 100644
index 00000000000..264602aafc2
--- /dev/null
+++ b/providers/kilo/models/~openai/gpt-sol-latest.toml
@@ -0,0 +1,29 @@
+name = "OpenAI GPT Sol Latest"
+description = "This model always redirects to the latest model in the OpenAI GPT Sol family."
+family = "gpt-sol"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/~openai/gpt-terra-latest.toml b/providers/kilo/models/~openai/gpt-terra-latest.toml
new file mode 100644
index 00000000000..58f1f2010c2
--- /dev/null
+++ b/providers/kilo/models/~openai/gpt-terra-latest.toml
@@ -0,0 +1,29 @@
+name = "OpenAI GPT Terra Latest"
+description = "This model always redirects to the latest model in the OpenAI GPT Terra family."
+family = "gpt-terra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]

From ccd513160ed29e8586787f95d21cb08968325741 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 13:26:32 +0000
Subject: [PATCH 1105/2012] chore(sync): update OpenRouter model catalog
 (#6881)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 +--
 .../models/inclusionai/ling-3.0-flash-vl.toml | 29 +++++++++++++++
 .../models/~openai/gpt-astra-latest.toml      | 36 ++++++++++++++++++
 .../models/~openai/gpt-luna-latest.toml       | 37 +++++++++++++++++++
 .../models/~openai/gpt-sol-latest.toml        | 37 +++++++++++++++++++
 .../models/~openai/gpt-terra-latest.toml      | 37 +++++++++++++++++++
 7 files changed, 182 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
 create mode 100644 providers/openrouter/models/~openai/gpt-astra-latest.toml
 create mode 100644 providers/openrouter/models/~openai/gpt-luna-latest.toml
 create mode 100644 providers/openrouter/models/~openai/gpt-sol-latest.toml
 create mode 100644 providers/openrouter/models/~openai/gpt-terra-latest.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..8d1267ee1b5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.954738
+output = 1.909476
+cache_read = 0.079562
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
new file mode 100644
index 00000000000..0b0d75e3590
--- /dev/null
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
@@ -0,0 +1,29 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ling 3.0 Flash VL"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "ling"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.06
+output = 0.18
+cache_read = 0.012
+
+[limit]
+context = 131_072
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/openrouter/models/~openai/gpt-astra-latest.toml b/providers/openrouter/models/~openai/gpt-astra-latest.toml
new file mode 100644
index 00000000000..c5d810b7e39
--- /dev/null
+++ b/providers/openrouter/models/~openai/gpt-astra-latest.toml
@@ -0,0 +1,36 @@
+name = "OpenAI GPT Astra Latest"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt-astra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/~openai/gpt-luna-latest.toml b/providers/openrouter/models/~openai/gpt-luna-latest.toml
new file mode 100644
index 00000000000..433ee7fb1fc
--- /dev/null
+++ b/providers/openrouter/models/~openai/gpt-luna-latest.toml
@@ -0,0 +1,37 @@
+name = "OpenAI GPT Luna Latest"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt-luna"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-02-16"
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.4
+output = 1.8
+cache_read = 0.04
+cache_write = 0.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/~openai/gpt-sol-latest.toml b/providers/openrouter/models/~openai/gpt-sol-latest.toml
new file mode 100644
index 00000000000..028f2642f96
--- /dev/null
+++ b/providers/openrouter/models/~openai/gpt-sol-latest.toml
@@ -0,0 +1,37 @@
+name = "OpenAI GPT Sol Latest"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt-sol"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-02-16"
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/~openai/gpt-terra-latest.toml b/providers/openrouter/models/~openai/gpt-terra-latest.toml
new file mode 100644
index 00000000000..5d33c886c13
--- /dev/null
+++ b/providers/openrouter/models/~openai/gpt-terra-latest.toml
@@ -0,0 +1,37 @@
+name = "OpenAI GPT Terra Latest"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt-terra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-02-16"
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 18
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]

From 980e778737cd1e803bea0deaaf5525c37a4a1a03 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:27:20 +0000
Subject: [PATCH 1106/2012] chore(sync): update Charm Hyper model catalog
 (#6884)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml              | 6 +++---
 providers/hyper/models/glm-5.1.toml                         | 6 +++---
 providers/hyper/models/glm-5.toml                           | 2 +-
 providers/hyper/models/gpt-oss-120b.toml                    | 6 +++---
 .../models/llama-4-maverick-17b-128e-instruct-fp8.toml      | 6 +++---
 providers/hyper/models/minimax-m2.7.toml                    | 6 +++---
 6 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index fb39f3786c1..6c033a0804a 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.122
-output = 0.42
-cache_read = 0.061
+input = 0.102
+output = 0.356
+cache_read = 0.051
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 24bf7b877f4..2440add2e43 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.326
-output = 4.22
-cache_read = 0.663
+input = 1.318
+output = 4.308
+cache_read = 0.659
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 108e0f80af8..9882a80de94 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -5,7 +5,7 @@ reasoning_options = []
 
 [cost]
 input = 0.86
-output = 2.784
+output = 2.752
 cache_read = 0.43
 
 [limit]
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index 22c22acac3a..d1529db1b94 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.178
-output = 0.68
-cache_read = 0.089
+input = 0.168
+output = 0.66
+cache_read = 0.084
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
index 10568630e53..5a9ed7085e5 100644
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.274
-output = 0.8992
-cache_read = 0.137
+input = 0.255
+output = 0.8365
+cache_read = 0.1275
 
 [limit]
 context = 430_000
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 4a47482e34e..1489a8dde84 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.404
-output = 1.496
-cache_read = 0.202
+input = 0.396
+output = 1.464
+cache_read = 0.198
 
 [limit]
 context = 262_100

From 5d5cd46b8b4f1d5e14753d609c049c6f6365ef21 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:27:23 +0000
Subject: [PATCH 1107/2012] chore(sync): update Merge Gateway model catalog
 (#6885)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 5ef98e7a3a0..b86a0e589cd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +11,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 84592d1b8dd..0264c6bc733 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 5ce7bd210475c43d385f9c74e1e62f945f0cdba8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:27:27 +0000
Subject: [PATCH 1108/2012] chore(sync): update OpenRouter model catalog
 (#6882)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 7 +++----
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 providers/openrouter/models/minimax/minimax-m2.5.toml    | 9 ++++++---
 providers/openrouter/models/~z-ai/glm-latest.toml        | 8 ++++----
 5 files changed, 19 insertions(+), 17 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..e9b629fb28e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.0868
+output = 0.1736
+cache_read = 0.01736
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 30c5b0f60a5..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.066
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 8d1267ee1b5..944ea7f1a30 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.954738
-output = 1.909476
-cache_read = 0.079562
+input = 0.947778
+output = 1.895556
+cache_read = 0.078982
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index 1edeb5ecdfc..ca75f8e4927 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,6 +6,9 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.03
+input = 0.27
+output = 1.08
+cache_read = 0.027
+
+[limit]
+output = 128_000
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index be978805f70..2b14e7625ce 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.97
-output = 3.3077
-cache_read = 0.194
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 128_000
 
 [modalities]
 input = ["text"]

From 2d6b6d7ae7e25dd7e3296bacf0175fd88c0baeab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:27:29 +0000
Subject: [PATCH 1109/2012] chore(sync): update Kilo model catalog (#6883)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 providers/kilo/models/minimax/minimax-m2.5.toml        |  4 ++++
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index fa4ef3087d9..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index 3f15c0bc21d..e89b49a7fbe 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,3 +10,7 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
+
+[limit]
+context = 200_000
+output = 128_000
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 550d2e46fef..114538a3af7 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.97
-output = 3.3077
-cache_read = 0.194
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
+output = 128_000
 
 [modalities]
 input = ["text"]

From 5423c9dabf5cb86ca3588a3ef30686ff441fc995 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:27:37 +0000
Subject: [PATCH 1110/2012] chore(sync): update NanoGPT model catalog (#6886)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/TEE/nemotron-3.5-lightning.toml        | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/nano-gpt/models/TEE/nemotron-3.5-lightning.toml

diff --git a/providers/nano-gpt/models/TEE/nemotron-3.5-lightning.toml b/providers/nano-gpt/models/TEE/nemotron-3.5-lightning.toml
new file mode 100644
index 00000000000..8d675f6e0f2
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/nemotron-3.5-lightning.toml
@@ -0,0 +1,15 @@
+base_model = "nvidia/nemotron-3.5-lightning"
+name = "Nvidia Nemotron 3.5 Lightning TEE"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.08
+output = 0.2
+cache_read = 0.04
+
+[limit]
+input = 262_144
+output = 65_536

From 0aac80ae0b8cdfcd205a7b6e7fb58e59470c091d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:24:05 -0500
Subject: [PATCH 1111/2012] fix: [missing-model] pioneer: claude-opus-5-fast
 (#6840)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/pioneer/models/claude-opus-5-fast.toml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/pioneer/models/claude-opus-5-fast.toml

diff --git a/providers/pioneer/models/claude-opus-5-fast.toml b/providers/pioneer/models/claude-opus-5-fast.toml
new file mode 100644
index 00000000000..b6db854f608
--- /dev/null
+++ b/providers/pioneer/models/claude-opus-5-fast.toml
@@ -0,0 +1,15 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "anthropic/claude-opus-5"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5

From 7a1b25488523944a92e07f3ad066337b7b3d7bf0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:24:13 -0500
Subject: [PATCH 1112/2012] fix: [missing-model] pioneer: ministral-14b (#6833)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/mistral/ministral-14b.toml           | 26 +++++++++++++++++++++
 providers/pioneer/models/ministral-14b.toml | 12 ++++++++++
 2 files changed, 38 insertions(+)
 create mode 100644 models/mistral/ministral-14b.toml
 create mode 100644 providers/pioneer/models/ministral-14b.toml

diff --git a/models/mistral/ministral-14b.toml b/models/mistral/ministral-14b.toml
new file mode 100644
index 00000000000..28621ad6129
--- /dev/null
+++ b/models/mistral/ministral-14b.toml
@@ -0,0 +1,26 @@
+# Sources:
+# https://mistral.ai/news/mistral-3/
+# https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512
+name = "Ministral 14B"
+description = "Compact multimodal Mistral model for local assistants, edge agents, and efficient tool use"
+family = "ministral"
+release_date = "2025-12-02"
+last_updated = "2025-12-02"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = true
+license = "Apache-2.0"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512"
diff --git a/providers/pioneer/models/ministral-14b.toml b/providers/pioneer/models/ministral-14b.toml
new file mode 100644
index 00000000000..989e6e5e0bf
--- /dev/null
+++ b/providers/pioneer/models/ministral-14b.toml
@@ -0,0 +1,12 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "mistral/ministral-14b"
+
+[cost]
+input = 0.2
+output = 0.2
+cache_read = 0.2
+cache_write = 0.2
+
+[limit]
+context = 256_000
+output = 131_072

From 2730ea0dd7d1025b4d6a970aac414bd2ab57a173 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:24:19 -0500
Subject: [PATCH 1113/2012] fix: [missing-model] pioneer: devstral-small-2
 (#6834)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/mistral/devstral-small-2.toml          | 27 +++++++++++++++++++
 .../pioneer/models/devstral-small-2.toml      | 12 +++++++++
 2 files changed, 39 insertions(+)
 create mode 100644 models/mistral/devstral-small-2.toml
 create mode 100644 providers/pioneer/models/devstral-small-2.toml

diff --git a/models/mistral/devstral-small-2.toml b/models/mistral/devstral-small-2.toml
new file mode 100644
index 00000000000..f8036965c99
--- /dev/null
+++ b/models/mistral/devstral-small-2.toml
@@ -0,0 +1,27 @@
+# Sources:
+# https://mistral.ai/news/devstral-2-vibe-cli/
+# https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512
+name = "Devstral Small 2"
+description = "Compact multimodal coding model for repository exploration, file editing, and software agents"
+family = "devstral"
+release_date = "2025-12-09"
+last_updated = "2025-12-09"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2025-12"
+open_weights = true
+license = "Apache-2.0"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512"
diff --git a/providers/pioneer/models/devstral-small-2.toml b/providers/pioneer/models/devstral-small-2.toml
new file mode 100644
index 00000000000..9372e8cbe89
--- /dev/null
+++ b/providers/pioneer/models/devstral-small-2.toml
@@ -0,0 +1,12 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "mistral/devstral-small-2"
+
+[cost]
+input = 0.1
+output = 0.3
+cache_read = 0.1
+cache_write = 0.1
+
+[limit]
+context = 256_000
+output = 131_072

From 90b7d665cb7359fc720f55da12edd1b11882a9c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:24:25 -0500
Subject: [PATCH 1114/2012] fix: [missing-model] pioneer: magistral-medium
 (#6835)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/pioneer/models/magistral-medium.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/pioneer/models/magistral-medium.toml

diff --git a/providers/pioneer/models/magistral-medium.toml b/providers/pioneer/models/magistral-medium.toml
new file mode 100644
index 00000000000..24fe9ebbc91
--- /dev/null
+++ b/providers/pioneer/models/magistral-medium.toml
@@ -0,0 +1,18 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "mistral/magistral-medium-latest"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 2
+output = 5
+cache_read = 2
+cache_write = 2
+
+[limit]
+output = 64_000

From effbdaa4bdd4083ff6e1479d7b284a1b80654127 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:25:38 +0000
Subject: [PATCH 1115/2012] fix: [missing-model] pioneer:
 thinkingmachines/inkling-small (#6812)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../thinkingmachines/inkling-small.toml       | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/pioneer/models/thinkingmachines/inkling-small.toml

diff --git a/providers/pioneer/models/thinkingmachines/inkling-small.toml b/providers/pioneer/models/thinkingmachines/inkling-small.toml
new file mode 100644
index 00000000000..ab657339645
--- /dev/null
+++ b/providers/pioneer/models/thinkingmachines/inkling-small.toml
@@ -0,0 +1,19 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "thinkingmachines/inkling-small"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.5
+output = 1.2
+cache_read = 0.1
+cache_write = 0.5
+
+[limit]
+context = 1_000_000
+output = 131_072

From 51d3abc05c65e68e73f10484549ec546f46de1a9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:25:44 +0000
Subject: [PATCH 1116/2012] fix: [missing-model] pioneer: ministral-3b (#6825)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/mistral/ministral-3b.toml           | 22 ++++++++++++++++++++++
 providers/pioneer/models/ministral-3b.toml | 11 +++++++++++
 2 files changed, 33 insertions(+)
 create mode 100644 models/mistral/ministral-3b.toml
 create mode 100644 providers/pioneer/models/ministral-3b.toml

diff --git a/models/mistral/ministral-3b.toml b/models/mistral/ministral-3b.toml
new file mode 100644
index 00000000000..ab4a0a28366
--- /dev/null
+++ b/models/mistral/ministral-3b.toml
@@ -0,0 +1,22 @@
+# Sources:
+# https://mistral.ai/news/ministraux/
+# https://ai.azure.com/catalog/models/Ministral-3B
+name = "Ministral 3B"
+description = "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads"
+family = "ministral"
+release_date = "2024-10-16"
+last_updated = "2024-10-16"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2024-03"
+open_weights = true
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/pioneer/models/ministral-3b.toml b/providers/pioneer/models/ministral-3b.toml
new file mode 100644
index 00000000000..1048a637647
--- /dev/null
+++ b/providers/pioneer/models/ministral-3b.toml
@@ -0,0 +1,11 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "mistral/ministral-3b"
+
+[cost]
+input = 0.1
+output = 0.1
+cache_read = 0.1
+cache_write = 0.1
+
+[limit]
+output = 4_000

From cdf2b6c51d42c20dae3a35911c3efdab3f8ec245 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:25:48 +0000
Subject: [PATCH 1117/2012] fix: [missing-model] pioneer:
 moonshotai/Kimi-K3-Fast (#6826)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../models/moonshotai/Kimi-K3-Fast.toml        | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/pioneer/models/moonshotai/Kimi-K3-Fast.toml

diff --git a/providers/pioneer/models/moonshotai/Kimi-K3-Fast.toml b/providers/pioneer/models/moonshotai/Kimi-K3-Fast.toml
new file mode 100644
index 00000000000..ff9cdc8daf4
--- /dev/null
+++ b/providers/pioneer/models/moonshotai/Kimi-K3-Fast.toml
@@ -0,0 +1,18 @@
+# Source: https://api.pioneer.ai/v1/models (accessed 2026-09-11)
+base_model = "moonshotai/kimi-k3"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 4.5
+output = 22.5
+cache_read = 0.45
+cache_write = 4.5
+
+[limit]
+context = 1_000_000

From 86d0c9618b30efab8c42cba883b6ca2146a99b09 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:26:07 +0000
Subject: [PATCH 1118/2012] chore(sync): update Kilo model catalog (#6888)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~openai/gpt-latest.toml | 30 -------------------
 1 file changed, 30 deletions(-)
 delete mode 100644 providers/kilo/models/~openai/gpt-latest.toml

diff --git a/providers/kilo/models/~openai/gpt-latest.toml b/providers/kilo/models/~openai/gpt-latest.toml
deleted file mode 100644
index 7c94513eab8..00000000000
--- a/providers/kilo/models/~openai/gpt-latest.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# Tracks the current GPT frontier alias target on Kilo (GPT-5.5).
-name = "OpenAI GPT Latest"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-04-27"
-last_updated = "2026-04-27"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]

From 1a3836368ecb83b5780246ec6142eb4d83c83c6a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:26:13 +0000
Subject: [PATCH 1119/2012] chore(sync): update EmpirioLabs AI model catalog
 (#6887)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../empiriolabs/models/fugu-ultra-v2-0.toml   | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/empiriolabs/models/fugu-ultra-v2-0.toml

diff --git a/providers/empiriolabs/models/fugu-ultra-v2-0.toml b/providers/empiriolabs/models/fugu-ultra-v2-0.toml
new file mode 100644
index 00000000000..e4461632bc0
--- /dev/null
+++ b/providers/empiriolabs/models/fugu-ultra-v2-0.toml
@@ -0,0 +1,20 @@
+base_model = "sakana/fugu-ultra"
+name = "Fugu Ultra v2.0"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 45
+cache_read = 1
+
+[limit]
+output = 131_072

From 17b8008818291c9537841a6bfdcf2ac6b50cc6d1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 15:28:40 +0000
Subject: [PATCH 1120/2012] chore(sync): update OpenRouter model catalog
 (#6889)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +--
 .../openrouter/models/~openai/gpt-latest.toml | 39 -------------------
 2 files changed, 3 insertions(+), 42 deletions(-)
 delete mode 100644 providers/openrouter/models/~openai/gpt-latest.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e9b629fb28e..713b581e8d8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0868
-output = 0.1736
-cache_read = 0.01736
+input = 0.08652
+output = 0.17304
+cache_read = 0.017304
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~openai/gpt-latest.toml b/providers/openrouter/models/~openai/gpt-latest.toml
deleted file mode 100644
index 10fad96622b..00000000000
--- a/providers/openrouter/models/~openai/gpt-latest.toml
+++ /dev/null
@@ -1,39 +0,0 @@
-# OpenRouter alias currently redirects to GPT-5.6 Sol.
-# https://openrouter.ai/~openai/gpt-latest
-name = "OpenAI GPT Latest"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-04-27"
-last_updated = "2026-04-27"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-knowledge = "2026-02-16"
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
-input = 4
-output = 15
-cache_read = 0.4
-cache_write = 5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]

From 2427c8d0b4da97b5af55bee3fb22c93bd248cc23 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:34:03 -0500
Subject: [PATCH 1121/2012] fix(amazon-bedrock): correct DeepSeek and Qwen
 catalog (#6823)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 .../models/deepseek.r1-v1:0.toml              | 28 ++++++-------------
 .../models/deepseek.v3-v1:0.toml              | 22 +++++----------
 .../amazon-bedrock/models/deepseek.v3.2.toml  | 24 ++++++----------
 .../qwen.qwen3-235b-a22b-2507-v1:0.toml       | 19 +++----------
 .../models/qwen.qwen3-32b-v1:0.toml           | 25 ++++++-----------
 .../models/qwen.qwen3-coder-30b-a3b-v1:0.toml | 19 +++----------
 .../qwen.qwen3-coder-480b-a35b-v1:0.toml      | 21 ++++----------
 .../models/qwen.qwen3-coder-next.toml         | 27 ++++--------------
 .../models/qwen.qwen3-next-80b-a3b.toml       | 24 ++++++----------
 .../models/qwen.qwen3-vl-235b-a22b.toml       | 24 +++++-----------
 .../models/us.deepseek.r1-v1:0.toml           |  6 ++++
 11 files changed, 72 insertions(+), 167 deletions(-)

diff --git a/providers/amazon-bedrock/models/deepseek.r1-v1:0.toml b/providers/amazon-bedrock/models/deepseek.r1-v1:0.toml
index 9874c883ca3..0bd83c97e60 100644
--- a/providers/amazon-bedrock/models/deepseek.r1-v1:0.toml
+++ b/providers/amazon-bedrock/models/deepseek.r1-v1:0.toml
@@ -1,24 +1,14 @@
-name = "DeepSeek-R1"
-description = "DeepSeek reasoning model for multi-step analysis, math, coding, and tools"
-family = "deepseek-thinking"
-release_date = "2025-01-20"
-last_updated = "2025-05-29"
-attachment = false
-reasoning = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-r1.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-deepseek.html
+# Runtime accepts 32,768 output tokens; 8,192 is the quality recommendation.
+# R1 has always-on reasoning and no reasoning toggle, effort, or budget parameter.
+# US Standard pricing; use the us.deepseek.r1-v1:0 profile for on-demand Converse.
+# Live US profile (us-east-1, 2026-09-09): toolConfig rejected with HTTP 400,
+# "This model doesn't support tool use." with AUTO or omitted toolChoice.
+base_model = "deepseek/deepseek-r1"
 reasoning_options = []
-temperature = true
-knowledge = "2024-07"
-tool_call = true
-open_weights = false
+tool_call = false
 
 [cost]
 input = 1.35
 output = 5.4
-
-[limit]
-context = 128_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/deepseek.v3-v1:0.toml b/providers/amazon-bedrock/models/deepseek.v3-v1:0.toml
index 0853474205b..4b3edd237d3 100644
--- a/providers/amazon-bedrock/models/deepseek.v3-v1:0.toml
+++ b/providers/amazon-bedrock/models/deepseek.v3-v1:0.toml
@@ -1,16 +1,12 @@
-name = "DeepSeek-V3.1"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-family = "deepseek"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-v3-1.html
+# Runtime ID; Mantle uses deepseek.v3.1. Prices are US Standard.
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default mode, matching V3.2.
+# Live 2026-09-09: high returns reasoningContent; maxTokens=81920 accepted (card's 8K is not the cap).
+base_model = "deepseek/deepseek-v3.1"
 last_updated = "2025-09-18"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-knowledge = "2024-07"
-tool_call = true
+reasoning_options = [{ type = "toggle" }]
+interleaved = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.58
@@ -19,7 +15,3 @@ output = 1.68
 [limit]
 context = 163_840
 output = 81_920
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/deepseek.v3.2.toml b/providers/amazon-bedrock/models/deepseek.v3.2.toml
index d58574ca721..da189e05685 100644
--- a/providers/amazon-bedrock/models/deepseek.v3.2.toml
+++ b/providers/amazon-bedrock/models/deepseek.v3.2.toml
@@ -1,16 +1,12 @@
-name = "DeepSeek-V3.2"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-family = "deepseek"
-release_date = "2026-02-06"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-v3-2.html
+# https://github.com/aws-samples/bedrock-claude-chatbot/blob/eeb2fa35339081d34ea9caa4e4d39cf1c189770a/agent/chat_agent.py
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default non-thinking mode.
+# Live 2026-09-09: high/omitted returns/omits reasoningContent; maxTokens=81920 accepted.
+# Runtime Converse; US Standard pricing.
+base_model = "deepseek/deepseek-v3.2"
 last_updated = "2026-02-06"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-knowledge = "2024-07"
-tool_call = true
-structured_output = true
-open_weights = true
+reasoning_options = [{ type = "toggle" }]
+interleaved = true
 
 [cost]
 input = 0.62
@@ -19,7 +15,3 @@ output = 1.85
 [limit]
 context = 163_840
 output = 81_920
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-235b-a22b-2507-v1:0.toml b/providers/amazon-bedrock/models/qwen.qwen3-235b-a22b-2507-v1:0.toml
index 038ba7f6d4c..ffe6eb7e0e0 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-235b-a22b-2507-v1:0.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-235b-a22b-2507-v1:0.toml
@@ -1,24 +1,13 @@
-name = "Qwen3 235B A22B 2507"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-235b-a22b-2507.html
+# https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507 (non-thinking Instruct variant)
+# Runtime ID; US Standard pricing.
+base_model = "alibaba/qwen3-235b-a22b-instruct-2507"
 last_updated = "2025-09-18"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2024-04"
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.22
 output = 0.88
 
 [limit]
-context = 262_144
 output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-32b-v1:0.toml b/providers/amazon-bedrock/models/qwen.qwen3-32b-v1:0.toml
index 702692c56d2..ca7b6b3325d 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-32b-v1:0.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-32b-v1:0.toml
@@ -1,25 +1,16 @@
-name = "Qwen3 32B (dense)"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-32b.html
+# Toggle: additionalModelRequestFields.reasoning_config = low|high (off|on).
+# Live 2026-09-09: high returns reasoningContent, low omits it; maxTokens=16384 accepted.
+# Runtime ID; US Standard pricing. No verified Bedrock reasoning-token budget.
+base_model = "alibaba/qwen3-32b"
 last_updated = "2025-09-18"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-knowledge = "2024-04"
-tool_call = true
+reasoning_options = [{ type = "toggle" }]
+interleaved = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.15
 output = 0.6
 
 [limit]
-context = 16_384
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
+context = 32_768
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-coder-30b-a3b-v1:0.toml b/providers/amazon-bedrock/models/qwen.qwen3-coder-30b-a3b-v1:0.toml
index 32a8e76ffe4..d21e3849085 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-coder-30b-a3b-v1:0.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-coder-30b-a3b-v1:0.toml
@@ -1,24 +1,13 @@
-name = "Qwen3 Coder 30B A3B Instruct"
-description = "Qwen coding model for software agents, repository edits, and code reasoning"
-family = "qwen"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-coder-30b-a3b-instruct.html
+# Runtime ID; US Standard pricing.
+base_model = "alibaba/qwen3-coder-30b-a3b-instruct"
+release_date = "2025-07-31"
 last_updated = "2025-09-18"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2024-04"
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.15
 output = 0.6
 
 [limit]
-context = 262_144
 output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-coder-480b-a35b-v1:0.toml b/providers/amazon-bedrock/models/qwen.qwen3-coder-480b-a35b-v1:0.toml
index 80708802a88..3a127506eff 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-coder-480b-a35b-v1:0.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-coder-480b-a35b-v1:0.toml
@@ -1,24 +1,13 @@
-name = "Qwen3 Coder 480B A35B Instruct"
-description = "Qwen coding model for software agents, repository edits, and code reasoning"
-family = "qwen"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-coder-480b-a35b-instruct.html
+# AWS Price List: USW2-Qwen3Coder-480B-A35B-{input,output}-tokens (Standard).
+base_model = "alibaba/qwen3-coder-480b-a35b-instruct"
+release_date = "2025-07-23"
 last_updated = "2025-09-18"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2024-04"
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
-input = 0.22
+input = 0.45
 output = 1.8
 
 [limit]
 context = 131_072
-output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-coder-next.toml b/providers/amazon-bedrock/models/qwen.qwen3-coder-next.toml
index 3720c3587b7..41c27b2ec86 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-coder-next.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-coder-next.toml
@@ -1,24 +1,9 @@
-name = "Qwen3 Coder Next"
-description = "Qwen coding model for software agents, repository edits, and code reasoning"
-family = "qwen"
-release_date = "2026-02-06"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-coder-next.html
+# https://huggingface.co/Qwen/Qwen3-Coder-Next (explicitly non-thinking)
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+base_model = "alibaba/qwen3-coder-next"
 last_updated = "2026-02-06"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
-input = 0.22
-output = 1.8
-
-[limit]
-context = 131_072
-output = 65_536
-
-[modalities]
-input = ["text"]
-output = ["text"]
+input = 0.50
+output = 1.20
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-next-80b-a3b.toml b/providers/amazon-bedrock/models/qwen.qwen3-next-80b-a3b.toml
index b6a011d69de..c0eea165451 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-next-80b-a3b.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-next-80b-a3b.toml
@@ -1,23 +1,15 @@
-name = "Qwen/Qwen3-Next-80B-A3B-Instruct"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-09-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-next-80b-a3b.html
+# https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct (non-thinking Instruct variant)
+# https://aws.amazon.com/bedrock/pricing/ (US Standard: 0.15/1.20; Mantle has separate SKUs)
+base_model = "alibaba/qwen3-next-80b-a3b-instruct"
+release_date = "2025-09-11"
 last_updated = "2025-11-25"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
-input = 0.14
-output = 1.4
+input = 0.15
+output = 1.20
 
 [limit]
-context = 262_000
+context = 262_144
 output = 262_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/qwen.qwen3-vl-235b-a22b.toml b/providers/amazon-bedrock/models/qwen.qwen3-vl-235b-a22b.toml
index e6f3a529a73..49ec3f9287f 100644
--- a/providers/amazon-bedrock/models/qwen.qwen3-vl-235b-a22b.toml
+++ b/providers/amazon-bedrock/models/qwen.qwen3-vl-235b-a22b.toml
@@ -1,23 +1,13 @@
-name = "Qwen/Qwen3-VL-235B-A22B-Instruct"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-10-04"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-qwen-qwen3-vl-235b-a22b.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live Converse 2026-09-11: outputConfig.textFormat enforced a JSON schema.
+base_model = "alibaba/qwen3-vl-235b-a22b-instruct"
 last_updated = "2025-11-25"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [cost]
-input = 0.3
-output = 1.5
+input = 0.53
+output = 2.66
 
 [limit]
-context = 262_000
+context = 262_144
 output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/us.deepseek.r1-v1:0.toml b/providers/amazon-bedrock/models/us.deepseek.r1-v1:0.toml
index e933c53c6ce..4a9e1a667f2 100644
--- a/providers/amazon-bedrock/models/us.deepseek.r1-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.deepseek.r1-v1:0.toml
@@ -1,6 +1,12 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-deepseek.html
+# R1 has always-on reasoning; the API accepts 32,768 output tokens (8,192 recommended).
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-r1.html
+# Live Converse (us-east-1, 2026-09-09): toolConfig rejected with HTTP 400,
+# "This model doesn't support tool use." with AUTO or omitted toolChoice.
 base_model = "deepseek/deepseek-r1"
 reasoning_options = []
 name = "DeepSeek-R1 (US)"
+tool_call = false
 
 [cost]
 input = 1.35

From d3d050f22b7bbb761734fb94f4ca849831d04405 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:34:12 -0500
Subject: [PATCH 1122/2012] fix: Basten: DeepSeek v4.1 Flash is available in
 Basten (#6875)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../deepseek-ai/DeepSeek-V4.1-Flash.toml      | 25 +++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/baseten/models/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/baseten/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/baseten/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..5ec2411504b
--- /dev/null
+++ b/providers/baseten/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,25 @@
+# Baseten documents top-level reasoning_effort for DeepSeek V4.1 Flash with
+# values none | low | high (default) | max. Reasoning is enabled by default;
+# reasoning_effort: "none" disables it. Reasoning output is returned in
+# reasoning_content.
+# Baseten serves at most 32k output tokens for this model on Model APIs.
+# https://www.baseten.co/library/deepseek-v41-flash/
+# https://docs.baseten.co/inference/model-apis/overview
+# https://docs.baseten.co/inference/model-apis/reasoning
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 32_768

From 45b70963502c436ca457667c8bd9efa7b6efc1c4 Mon Sep 17 00:00:00 2001
From: Robert Keus <robert@brothersmedia.nl>
Date: Fri, 11 Sep 2026 17:34:25 +0200
Subject: [PATCH 1123/2012] feat(greenpt): add DeepSeek V4.1 Flash (#6860)

---
 providers/greenpt/models/deepseek-v4.1-flash.toml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/greenpt/models/deepseek-v4.1-flash.toml

diff --git a/providers/greenpt/models/deepseek-v4.1-flash.toml b/providers/greenpt/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..28bd8e54523
--- /dev/null
+++ b/providers/greenpt/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,13 @@
+# Cost: GreenPT EUR list prices converted at 1.1616 USD/EUR (ECB, 2026-09-10).
+# Sources: https://docs.greenpt.ai/deepseek-v4-1-flash and https://docs.greenpt.ai/model-cards
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# GreenPT accepts reasoning_effort = none|minimal|low|medium|high.
+# Source: https://docs.greenpt.ai/reasoning
+# Cache writes are not charged: https://docs.greenpt.ai/prompt-caching
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high"] }]
+
+[cost]
+input = 0.255552
+cache_read = 0.0127776
+output = 1.27776

From a1f9d53fa1d3c9cccf3de345779a2f8ed261a122 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Fri, 11 Sep 2026 22:36:48 +0700
Subject: [PATCH 1124/2012] feat(bothub): add Muse Spark 1.3 Contributor
 (#6830)

* feat(bothub): add Muse Spark 1.3 Contributor

Add Meta Muse Spark 1.3 Contributor (contributor tier) routed via bothub.

- New lab entry models/meta/muse-spark-1.3-contributor.toml
- Override-only provider file with base_model, reasoning efforts
  (minimal/low/medium/high/xhigh; 'max' not supported by the router)
  and live-verified pricing 0.10/0.20 USD per 1M input/output tokens

Sources: research.meta.ai blog, openrouter.ai, cloudprice.net, bothub.ru/models.
Pricing and reasoning_effort support verified against the bothub API.

* fix(models): drop parallel lab identity for Muse Spark 1.3 Contributor

* fix(bothub): point Contributor tier at meta/muse-spark-1.3 base model

---------

Co-authored-by: 4ebuRushka <4ebuRushka@users.noreply.github.com>
---
 .../models/muse-spark-1.3-contributor.toml    | 25 +++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/bothub/models/muse-spark-1.3-contributor.toml

diff --git a/providers/bothub/models/muse-spark-1.3-contributor.toml b/providers/bothub/models/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..8875137a496
--- /dev/null
+++ b/providers/bothub/models/muse-spark-1.3-contributor.toml
@@ -0,0 +1,25 @@
+# Pricing live-verified 2026-09-11 via a chat completion usage.cost
+# (11 prompt + 200 completion tokens => exactly $0.10/$0.20 per 1M).
+# Listed on https://bothub.ru/muse-spark-1.3-contributor as 11,79/23,57 RUB per 1M.
+# Effort live-verified 2026-09-11: reasoning_effort accepts
+# [minimal, low, medium, high, xhigh]; "max" is rejected upstream
+# (Meta holds max back pending safety testing:
+# https://research.meta.ai/blog/introducing-muse-spark-1-3).
+# Wire: reasoning in message.reasoning, mirrored in reasoning_details (encrypted).
+# Output override: live test 2026-09-11 accepted >131k completion tokens on
+# bothub (lab file's 131_072 is stale); 943 718 per OpenRouter/llm-stats.
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 Contributor"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.2
+
+[limit]
+output = 943_718

From 0ba0c45c64a829e27ef9bf6d0067f734d4c44205 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=AD=A4=E6=96=B9=E4=B8=80=E6=B3=89?= <lg_329@163.com>
Date: Fri, 11 Sep 2026 23:36:57 +0800
Subject: [PATCH 1125/2012] fix(volcengine-coding-plan): add glm-5.3-flash and
 kimi-k3 (#6831)

* fix(volcengine-coding-plan): add glm-5.3-flash and kimi-k3

The model list in https://www.volcengine.com/docs/82379/1928261 (Model Name
configuration) includes glm-5.3-flash and kimi-k3, but the provider only
modeled glm-5.3 and kimi-k2.7-code. Both are zero-cost subscription
entitlements like the other eight entries.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* review: author kimi-k3 reasoning baseline from lab/peers

Per review: relay reasoning_options must not be [] from unmeasured
uncertainty. Mirrors moonshotai/kimi-k3 and kimi-for-coding (toggle +
effort low/high/max) with the /api/coding/v3 wire fields documented;
endpoints-specific level behavior flagged as unmeasured.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* review: document this relay's real wire fields for kimi-k3

reasoning_effort (not output_config.effort) on the OpenAI chat path;
thinking.type is enabled|disabled without the lab-only adaptive value.
output_config.effort noted for the separate Anthropic path. Option set
unchanged (toggle + low/high/max).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: guanchen.lin <guanchen.lin@kunlun-inc.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
---
 .../models/glm-5.3-flash.toml                 | 17 +++++++++++
 .../models/kimi-k3.toml                       | 28 +++++++++++++++++++
 2 files changed, 45 insertions(+)
 create mode 100644 providers/volcengine-coding-plan/models/glm-5.3-flash.toml
 create mode 100644 providers/volcengine-coding-plan/models/kimi-k3.toml

diff --git a/providers/volcengine-coding-plan/models/glm-5.3-flash.toml b/providers/volcengine-coding-plan/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..45dd0b497e0
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/glm-5.3-flash.toml
@@ -0,0 +1,17 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-09-11): the model list includes both glm-5.3 (glm-latest)
+# and glm-5.3-flash. Subscription tier: no per-token price.
+# Effort levels mirror providers/zai-coding-plan/glm-5.3-flash.toml.
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/volcengine-coding-plan/models/kimi-k3.toml b/providers/volcengine-coding-plan/models/kimi-k3.toml
new file mode 100644
index 00000000000..18f097450bd
--- /dev/null
+++ b/providers/volcengine-coding-plan/models/kimi-k3.toml
@@ -0,0 +1,28 @@
+# Coding Plan model name from https://www.volcengine.com/docs/82379/1928261
+# (accessed 2026-09-11), listed alongside kimi-k2.7-code. Subscription tier:
+# no per-token price.
+# reasoning_options mirror the lab (providers/moonshotai/models/kimi-k3.toml)
+# and the kimi-for-coding relay peer. Wire fields on this provider's OpenAI
+# path (provider.toml): POST /api/coding/v3/chat/completions takes
+# thinking.type = enabled|disabled plus reasoning_effort = low|high|max
+# (`adaptive` is the Moonshot lab surface, not accepted here); the separate
+# Anthropic path (/api/coding/v1/messages) uses output_config.effort.
+# Endpoint-specific effort behavior on this relay was not measured; the
+# baseline is copied per relay convention — adjust if wire probing shows
+# levels collapsing as they do on kimi-k2.7-code (see that file).
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0

From 55a3d7a73f7c59e9d93c2adab2e5f98025f2fdc8 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:37:14 -0500
Subject: [PATCH 1126/2012] fix(amazon-bedrock): correct hosted reasoner
 catalog (#6822)

* fix(amazon-bedrock): correct hosted reasoner catalog

* docs(amazon-bedrock): clarify Nemotron context

---------

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 .../models/global.xai.grok-4.6.toml           |  4 ++-
 .../models/minimax.minimax-m2.1.toml          | 26 ++++-----------
 .../models/minimax.minimax-m2.5.toml          | 22 +++++--------
 .../models/minimax.minimax-m2.toml            | 22 ++++---------
 .../models/moonshot.kimi-k2-thinking.toml     | 18 +++-------
 .../models/moonshotai.kimi-k2.5.toml          | 33 +++++++++----------
 .../models/nvidia.nemotron-nano-12b-v2.toml   | 14 +++-----
 .../models/nvidia.nemotron-nano-3-30b.toml    | 18 +++-------
 .../models/nvidia.nemotron-nano-9b-v2.toml    | 17 +++-------
 .../models/nvidia.nemotron-super-3-120b.toml  | 16 ++-------
 .../models/us.xai.grok-4.6.toml               |  5 ++-
 .../amazon-bedrock/models/xai.grok-4.3.toml   |  4 ++-
 .../amazon-bedrock/models/xai.grok-4.6.toml   |  2 ++
 .../models/zai.glm-4.7-flash.toml             | 27 ++++-----------
 .../amazon-bedrock/models/zai.glm-4.7.toml    | 31 +++++------------
 .../amazon-bedrock/models/zai.glm-5.toml      | 25 ++++----------
 16 files changed, 91 insertions(+), 193 deletions(-)

diff --git a/providers/amazon-bedrock/models/global.xai.grok-4.6.toml b/providers/amazon-bedrock/models/global.xai.grok-4.6.toml
index 552f860bcc3..0be8c0e821e 100644
--- a/providers/amazon-bedrock/models/global.xai.grok-4.6.toml
+++ b/providers/amazon-bedrock/models/global.xai.grok-4.6.toml
@@ -1,4 +1,6 @@
-# Source: AWS Bedrock Grok 4.6 model card — Runtime profiles, pricing, and capabilities.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-6.html
+# Runtime Converse profile; Global CRIS Standard pricing. Structured outputs are Mantle-only.
+# Effort: additionalModelRequestFields.reasoning.effort = low|medium|high|xhigh, matching the US profile.
 base_model = "xai/grok-4.6"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
 name = "Grok 4.6 (Global)"
diff --git a/providers/amazon-bedrock/models/minimax.minimax-m2.1.toml b/providers/amazon-bedrock/models/minimax.minimax-m2.1.toml
index 7e184c29483..5261a4e1407 100644
--- a/providers/amazon-bedrock/models/minimax.minimax-m2.1.toml
+++ b/providers/amazon-bedrock/models/minimax.minimax-m2.1.toml
@@ -1,24 +1,12 @@
-name = "MiniMax M2.1"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-family = "minimax"
-release_date = "2025-12-23"
-last_updated = "2025-12-23"
-attachment = false
-reasoning = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-minimax-minimax-m2-1.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# MiniMax M2.1 has always-on thinking, matching the first-party API.
+# Live Converse 2026-09-11: reasoningContent returned and outputConfig.textFormat enforced a JSON schema.
+base_model = "minimax/MiniMax-M2.1"
 reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
+interleaved = true
+structured_output = true
 
 [cost]
 input = 0.30
 output = 1.20
-
-[limit]
-context = 204_800
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/minimax.minimax-m2.5.toml b/providers/amazon-bedrock/models/minimax.minimax-m2.5.toml
index 12503198854..67efaa691ef 100644
--- a/providers/amazon-bedrock/models/minimax.minimax-m2.5.toml
+++ b/providers/amazon-bedrock/models/minimax.minimax-m2.5.toml
@@ -1,15 +1,13 @@
-name = "MiniMax M2.5"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-family = "minimax"
-release_date = "2026-03-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-minimax-minimax-m2-5.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# MiniMax M2.5 has always-on thinking, matching the first-party API.
+# Live 2026-09-09: maxTokens=98304 accepted and reasoningContent returned with no reasoning configuration.
+# Live Converse 2026-09-11: outputConfig.textFormat enforced a JSON schema.
+base_model = "minimax/MiniMax-M2.5"
 last_updated = "2026-03-18"
-attachment = false
-reasoning = true
 reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
+interleaved = true
+structured_output = true
 
 [cost]
 input = 0.30
@@ -18,7 +16,3 @@ output = 1.20
 [limit]
 context = 196_608
 output = 98_304
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/minimax.minimax-m2.toml b/providers/amazon-bedrock/models/minimax.minimax-m2.toml
index a4e86eb782e..b5c56da3d6a 100644
--- a/providers/amazon-bedrock/models/minimax.minimax-m2.toml
+++ b/providers/amazon-bedrock/models/minimax.minimax-m2.toml
@@ -1,15 +1,11 @@
-name = "MiniMax M2"
-description = "MiniMax model for chat, coding, office work, and agentic tasks"
-family = "minimax"
-release_date = "2025-10-27"
-last_updated = "2025-10-27"
-attachment = false
-reasoning = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-minimax-minimax-m2.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# MiniMax M2 has always-on thinking, matching the first-party API.
+# Live Converse 2026-09-11: reasoningContent returned and outputConfig.textFormat enforced a JSON schema.
+base_model = "minimax/MiniMax-M2"
 reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
+interleaved = true
+structured_output = true
 
 [cost]
 input = 0.30
@@ -18,7 +14,3 @@ output = 1.20
 [limit]
 context = 204_608
 output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/moonshot.kimi-k2-thinking.toml b/providers/amazon-bedrock/models/moonshot.kimi-k2-thinking.toml
index c617d0d1066..628707af2ad 100644
--- a/providers/amazon-bedrock/models/moonshot.kimi-k2-thinking.toml
+++ b/providers/amazon-bedrock/models/moonshot.kimi-k2-thinking.toml
@@ -1,17 +1,11 @@
-# See https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-thinking.html
-name = "Kimi K2 Thinking"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
-release_date = "2025-12-02"
-family = "kimi-thinking"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-thinking.html
+# https://platform.moonshot.ai/docs/guide/use-kimi-k2-thinking-model (always-on thinking)
+# Runtime ID; Mantle uses moonshotai.kimi-k2-thinking. US Standard pricing.
+base_model = "moonshotai/kimi-k2-thinking"
 last_updated = "2025-12-02"
-attachment = false
-reasoning = true
 reasoning_options = []
-temperature = true
-tool_call = true
 structured_output = true
 interleaved = true
-open_weights = true
 
 [cost]
 input = 0.6
@@ -20,7 +14,3 @@ output = 2.5
 [limit]
 context = 262_143
 output = 16_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/moonshotai.kimi-k2.5.toml b/providers/amazon-bedrock/models/moonshotai.kimi-k2.5.toml
index 0610dfe9866..35ca4efa6c4 100644
--- a/providers/amazon-bedrock/models/moonshotai.kimi-k2.5.toml
+++ b/providers/amazon-bedrock/models/moonshotai.kimi-k2.5.toml
@@ -1,26 +1,23 @@
-# See https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html
-name = "Kimi K2.5"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi"
-release_date = "2026-02-06"
-last_updated = "2026-02-06"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = true
-interleaved = true
-open_weights = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html
+# https://github.com/aws-samples/bedrock-claude-chatbot/blob/eeb2fa35339081d34ea9caa4e4d39cf1c189770a/agent/chat_agent.py
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default non-thinking mode.
+# Live 2026-09-09: high/omitted returns/omits reasoningContent; maxTokens=16384 accepted.
+# Live Converse 2026-09-09: inferenceConfig.temperature accepts both 0 and 1.
+# Runtime Converse; US Standard pricing.
+base_model = "moonshotai/kimi-k2.5"
+release_date = "2026-01-27"
+last_updated = "2026-02-06"
+reasoning_options = [{ type = "toggle" }]
+temperature = true
+interleaved = true
 
 [cost]
 input = 0.6
 output = 3
 
-[limit]
-context = 262_143
-output = 16_000
+[limit]
+context = 262_143
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/nvidia.nemotron-nano-12b-v2.toml b/providers/amazon-bedrock/models/nvidia.nemotron-nano-12b-v2.toml
index 81a0370149e..9963a83da9a 100644
--- a/providers/amazon-bedrock/models/nvidia.nemotron-nano-12b-v2.toml
+++ b/providers/amazon-bedrock/models/nvidia.nemotron-nano-12b-v2.toml
@@ -1,23 +1,17 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-nvidia-nvidia-nemotron-nano-12b-v2-vl-bf16.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live 2026-09-09: maxTokens=8192 accepted; reasoning_config=high returned no reasoningContent.
 name = "NVIDIA Nemotron Nano 12B v2 VL BF16"
 base_model = "nvidia/nemotron-nano-12b-v2-vl"
-family = "nemotron"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
 reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.20
 output = 0.60
 
 [limit]
-context = 128_000
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/nvidia.nemotron-nano-3-30b.toml b/providers/amazon-bedrock/models/nvidia.nemotron-nano-3-30b.toml
index 309f93bca5a..a18ed947c6c 100644
--- a/providers/amazon-bedrock/models/nvidia.nemotron-nano-3-30b.toml
+++ b/providers/amazon-bedrock/models/nvidia.nemotron-nano-3-30b.toml
@@ -1,24 +1,16 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-nvidia-nemotron-nano-3-30b.html
+# AWS documents a 256K Bedrock context window; inherit the matching 262,144-token lab context.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live 2026-09-09: maxTokens=8192 accepted; chat_template_kwargs.enable_thinking had no visible effect.
 name = "NVIDIA Nemotron Nano 3 30B"
 base_model = "nvidia/nemotron-3-nano-30b-a3b"
-family = "nemotron"
-release_date = "2025-12-23"
 last_updated = "2025-12-23"
-attachment = false
-reasoning = true
 reasoning_options = []
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.06
 output = 0.24
 
 [limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/nvidia.nemotron-nano-9b-v2.toml b/providers/amazon-bedrock/models/nvidia.nemotron-nano-9b-v2.toml
index 7edf18dc4a5..2d5d07d252d 100644
--- a/providers/amazon-bedrock/models/nvidia.nemotron-nano-9b-v2.toml
+++ b/providers/amazon-bedrock/models/nvidia.nemotron-nano-9b-v2.toml
@@ -1,23 +1,14 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-nvidia-nvidia-nemotron-nano-9b-v2.html
+# Live 2026-09-09: maxTokens=8192 accepted; tested JSON thinking controls had no visible effect.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
 name = "NVIDIA Nemotron Nano 9B v2"
 base_model = "nvidia/nemotron-nano-9b-v2"
-family = "nemotron"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
 reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.06
 output = 0.23
 
 [limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/nvidia.nemotron-super-3-120b.toml b/providers/amazon-bedrock/models/nvidia.nemotron-super-3-120b.toml
index ae0f780cdd4..7b54c08aa50 100644
--- a/providers/amazon-bedrock/models/nvidia.nemotron-super-3-120b.toml
+++ b/providers/amazon-bedrock/models/nvidia.nemotron-super-3-120b.toml
@@ -1,24 +1,14 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-nvidia-nemotron-super-3-120b.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Live 2026-09-09: maxTokens=131072 accepted; chat_template_kwargs.enable_thinking had no visible effect.
 name = "NVIDIA Nemotron 3 Super 120B A12B"
 base_model = "nvidia/nemotron-3-super-120b-a12b"
-family = "nemotron"
-release_date = "2026-03-11"
-last_updated = "2026-03-11"
-attachment = false
-reasoning = true
 reasoning_options = []
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.15
 output = 0.65
 
 [limit]
-context = 262_144
 output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/us.xai.grok-4.6.toml b/providers/amazon-bedrock/models/us.xai.grok-4.6.toml
index 44fb743ba04..e550aa0a2b9 100644
--- a/providers/amazon-bedrock/models/us.xai.grok-4.6.toml
+++ b/providers/amazon-bedrock/models/us.xai.grok-4.6.toml
@@ -1,4 +1,7 @@
-# Source: AWS Bedrock Grok 4.6 model card — Runtime profiles, pricing, and capabilities.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-6.html
+# Runtime Converse profile; Geo CRIS Standard pricing. Structured outputs are Mantle-only.
+# Effort: additionalModelRequestFields.reasoning.effort = low|medium|high|xhigh.
+# Live 2026-09-09: nested effort=high returns reasoningContent; maxTokens=500000 accepted.
 base_model = "xai/grok-4.6"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
 name = "Grok 4.6 (US)"
diff --git a/providers/amazon-bedrock/models/xai.grok-4.3.toml b/providers/amazon-bedrock/models/xai.grok-4.3.toml
index 1f2e66ea32e..1c1f2e3b190 100644
--- a/providers/amazon-bedrock/models/xai.grok-4.3.toml
+++ b/providers/amazon-bedrock/models/xai.grok-4.3.toml
@@ -1,3 +1,6 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html
+# Mantle Responses only for this catalog route; US Standard pricing.
+# Effort: reasoning.effort = none|low|medium|high.
 base_model = "xai/grok-4.3"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 last_updated = "2026-06-28"
@@ -8,7 +11,6 @@ output = 2.50
 cache_read = 0.20
 
 [limit]
-context = 1_000_000
 output = 131_072
 
 [modalities]
diff --git a/providers/amazon-bedrock/models/xai.grok-4.6.toml b/providers/amazon-bedrock/models/xai.grok-4.6.toml
index 41d52dab1e6..2080832f78d 100644
--- a/providers/amazon-bedrock/models/xai.grok-4.6.toml
+++ b/providers/amazon-bedrock/models/xai.grok-4.6.toml
@@ -1,4 +1,6 @@
 # Source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-6.html
+# Mantle Responses; US Standard in-Region pricing.
+# Effort: reasoning.effort = low|medium|high|xhigh (always on).
 base_model = "xai/grok-4.6"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
 last_updated = "2026-08-18"
diff --git a/providers/amazon-bedrock/models/zai.glm-4.7-flash.toml b/providers/amazon-bedrock/models/zai.glm-4.7-flash.toml
index e98bb6a9c7f..97c3e0eed1e 100644
--- a/providers/amazon-bedrock/models/zai.glm-4.7-flash.toml
+++ b/providers/amazon-bedrock/models/zai.glm-4.7-flash.toml
@@ -1,25 +1,12 @@
-name = "GLM-4.7-Flash"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm-flash"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-zai-glm-4-7-flash.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default mode, matching GLM-4.7.
+# Live 2026-09-09: high returns reasoningContent; maxTokens=131072 accepted (card's 4K is not the cap).
+base_model = "zhipuai/glm-4.7-flash"
+reasoning_options = [{ type = "toggle" }]
+interleaved = true
 structured_output = true
-knowledge = "2025-04"
-open_weights = true
 
 [cost]
 input = 0.07
 output = 0.40
-
-[limit]
-context = 200_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/zai.glm-4.7.toml b/providers/amazon-bedrock/models/zai.glm-4.7.toml
index 2831936322a..b560d79ab91 100644
--- a/providers/amazon-bedrock/models/zai.glm-4.7.toml
+++ b/providers/amazon-bedrock/models/zai.glm-4.7.toml
@@ -1,28 +1,13 @@
-name = "GLM-4.7"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2025-12-22"
-last_updated = "2025-12-22"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-zai-glm-4-7.html
+# https://github.com/danny-avila/LibreChat/issues/11974 (Bedrock Converse request)
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default non-thinking mode.
+# Live 2026-09-09: high/omitted returns/omits reasoningContent; maxTokens=131072 accepted.
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+base_model = "zhipuai/glm-4.7"
+reasoning_options = [{ type = "toggle" }]
 structured_output = true
-knowledge = "2025-04"
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
+interleaved = true
 
 [cost]
 input = 0.60
 output = 2.20
-
-[limit]
-context = 204_800
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/zai.glm-5.toml b/providers/amazon-bedrock/models/zai.glm-5.toml
index 701a386a083..dacdb75b72d 100644
--- a/providers/amazon-bedrock/models/zai.glm-5.toml
+++ b/providers/amazon-bedrock/models/zai.glm-5.toml
@@ -1,18 +1,12 @@
-name = "GLM-5"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2026-03-18"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-zai-glm-5.html
+# https://aws.amazon.com/bedrock/pricing/ (US Standard pricing)
+# Toggle: additionalModelRequestFields.reasoning_config = "high"; omit for default mode, matching GLM-4.7.
+# Live 2026-09-09: high returns reasoningContent; maxTokens=131072 accepted.
+base_model = "zhipuai/glm-5"
 last_updated = "2026-03-18"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }]
 structured_output = true
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
+interleaved = true
 
 [cost]
 input = 1.00
@@ -20,8 +14,3 @@ output = 3.20
 
 [limit]
 context = 202_752
-output = 101_376
-
-[modalities]
-input = ["text"]
-output = ["text"]

From d8d896a9a9ee909d92c5de7d82eeae4aa1379fd9 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:37:54 -0500
Subject: [PATCH 1127/2012] fix: cline-pass: deepseek-v4.1-flash missing from
 provider data (#6863)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/cline-pass/deepseek-v4.1-flash.toml        | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml

diff --git a/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml b/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..cdc87207c83
--- /dev/null
+++ b/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml
@@ -0,0 +1,11 @@
+# https://api.cline.bot/api/v1/ai/cline/recommended-models (clinePass id cline-pass/deepseek-v4.1-flash, accessed 2026-09-11)
+# https://docs.cline.bot/getting-started/clinepass (ClinePass reference pricing cites DeepSeek API pricing for Flash quota)
+# https://api-docs.deepseek.com/quick_start/pricing (DeepSeek-V4.1-Flash off-peak USD/MTok; ClinePass docs table still lists deepseek-v4-flash only)
+# Reasoning effort values match other ClinePass DeepSeek entries on this host.
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003

From 2f5d8ff4b85aec82b318b2d12a0504742ff7a3af Mon Sep 17 00:00:00 2001
From: YongYuH <larry830326@hotmail.com>
Date: Fri, 11 Sep 2026 23:38:54 +0800
Subject: [PATCH 1128/2012] fix: [missing-model] ollama-cloud:
 deepseek-v4.1-flash (#6851)

Co-authored-by: larry.huang <larry.huang@scmp.com>
---
 .../ollama-cloud/models/deepseek-v4.1-flash.toml   | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/ollama-cloud/models/deepseek-v4.1-flash.toml

diff --git a/providers/ollama-cloud/models/deepseek-v4.1-flash.toml b/providers/ollama-cloud/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..84ecfb2ff46
--- /dev/null
+++ b/providers/ollama-cloud/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+# Toggle: native POST /api/chat and /api/generate use `think = true|false`.
+# Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
+# or `reasoning.effort` with low|high|max (Flash maps requested low -> low).
+# Sources: https://docs.ollama.com/capabilities/thinking, https://docs.ollama.com/openai
+# API-verified context: https://ollama.com/api/show (deepseek_v41.context_length = 1048576)
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+context = 1048576
\ No newline at end of file

From 4c581f109512443fce1009d0f54590c6cf40fe59 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 10:39:54 -0500
Subject: [PATCH 1129/2012] chore(sync): update Eden AI model catalog (#6849)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/amazon/google.gemma-3-12b-it.toml      | 11 +++++++++++
 .../models/amazon/google.gemma-3-12b-it@us.toml   | 11 +++++++++++
 .../models/amazon/google.gemma-3-27b-it.toml      | 11 +++++++++++
 .../models/amazon/google.gemma-3-27b-it@us.toml   | 11 +++++++++++
 .../models/amazon/google.gemma-3-4b-it.toml       | 11 +++++++++++
 .../models/amazon/google.gemma-3-4b-it@us.toml    | 11 +++++++++++
 .../amazon/openai.gpt-oss-safeguard-20b.toml      | 15 +++++++++++++++
 .../amazon/openai.gpt-oss-safeguard-20b@us.toml   | 15 +++++++++++++++
 .../deepseek-ai/DeepSeek-V4.1-Flash.toml          |  4 ++--
 .../models/deepinfra/google/gemma-3-12b-it.toml   |  7 +++++++
 .../models/deepinfra/google/gemma-3-27b-it.toml   |  7 +++++++
 .../models/deepinfra/google/gemma-3-4b-it.toml    |  7 +++++++
 .../models/groq/openai/gpt-oss-safeguard-20b.toml | 11 +++++++++++
 .../ionos/meta-llama/Llama-3.3-70B-Instruct.toml  |  4 ++--
 .../edenai/models/ionos/openai/gpt-oss-120b.toml  |  4 ++--
 .../models/nebius/google/gemma-3-27b-it.toml      | 11 +++++++++++
 .../models/scaleway/deepseek-v4-flash-0731.toml   |  4 ++--
 .../edenai/models/scaleway/gemma-3-27b-it.toml    | 11 +++++++++++
 .../edenai/models/scaleway/gpt-oss-120b.toml      |  4 ++--
 .../models/scaleway/llama-3.3-70b-instruct.toml   |  4 ++--
 20 files changed, 162 insertions(+), 12 deletions(-)
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-12b-it.toml
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-27b-it.toml
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-4b-it.toml
 create mode 100644 providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
 create mode 100644 providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
 create mode 100644 providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
 create mode 100644 providers/edenai/models/deepinfra/google/gemma-3-12b-it.toml
 create mode 100644 providers/edenai/models/deepinfra/google/gemma-3-27b-it.toml
 create mode 100644 providers/edenai/models/deepinfra/google/gemma-3-4b-it.toml
 create mode 100644 providers/edenai/models/groq/openai/gpt-oss-safeguard-20b.toml
 create mode 100644 providers/edenai/models/nebius/google/gemma-3-27b-it.toml
 create mode 100644 providers/edenai/models/scaleway/gemma-3-27b-it.toml

diff --git a/providers/edenai/models/amazon/google.gemma-3-12b-it.toml b/providers/edenai/models/amazon/google.gemma-3-12b-it.toml
new file mode 100644
index 00000000000..76ea0c2327a
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-12b-it.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-12b-it"
+name = "Gemma 3 12B IT (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.09
+output = 0.29
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
new file mode 100644
index 00000000000..91ebef21c30
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-12b-it"
+name = "Gemma 3 12B IT (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.09
+output = 0.29
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/google.gemma-3-27b-it.toml b/providers/edenai/models/amazon/google.gemma-3-27b-it.toml
new file mode 100644
index 00000000000..b99b2b38ac2
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-27b-it.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B IT (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.23
+output = 0.38
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
new file mode 100644
index 00000000000..4c9f67f300e
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B IT (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.23
+output = 0.38
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/google.gemma-3-4b-it.toml b/providers/edenai/models/amazon/google.gemma-3-4b-it.toml
new file mode 100644
index 00000000000..f502a9a96cd
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-4b-it.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-4b-it"
+name = "Gemma 3 4B IT (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.04
+output = 0.08
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
new file mode 100644
index 00000000000..6d658417b4f
--- /dev/null
+++ b/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-4b-it"
+name = "Gemma 3 4B IT (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.04
+output = 0.08
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
new file mode 100644
index 00000000000..536758dfd0e
--- /dev/null
+++ b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-oss-safeguard-20b"
+name = "GPT OSS Safeguard 20B (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.07
+output = 0.2
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
new file mode 100644
index 00000000000..7992978de7d
--- /dev/null
+++ b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-oss-safeguard-20b"
+name = "GPT OSS Safeguard 20B (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.07
+output = 0.2
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
index 3a219aa35fb..d741e86c6d5 100644
--- a/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
+++ b/providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
+input = 0.2
+output = 0.6
 cache_read = 0.006
 
 [limit]
diff --git a/providers/edenai/models/deepinfra/google/gemma-3-12b-it.toml b/providers/edenai/models/deepinfra/google/gemma-3-12b-it.toml
new file mode 100644
index 00000000000..5e45c8c3afc
--- /dev/null
+++ b/providers/edenai/models/deepinfra/google/gemma-3-12b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-12b-it"
+name = "Gemma 3 12B IT (Deep Infra)"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.15
diff --git a/providers/edenai/models/deepinfra/google/gemma-3-27b-it.toml b/providers/edenai/models/deepinfra/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..01fdf25ec18
--- /dev/null
+++ b/providers/edenai/models/deepinfra/google/gemma-3-27b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B IT (Deep Infra)"
+structured_output = true
+
+[cost]
+input = 0.08
+output = 0.16
diff --git a/providers/edenai/models/deepinfra/google/gemma-3-4b-it.toml b/providers/edenai/models/deepinfra/google/gemma-3-4b-it.toml
new file mode 100644
index 00000000000..a8895d2adb3
--- /dev/null
+++ b/providers/edenai/models/deepinfra/google/gemma-3-4b-it.toml
@@ -0,0 +1,7 @@
+base_model = "google/gemma-3-4b-it"
+name = "Gemma 3 4B IT (Deep Infra)"
+structured_output = true
+
+[cost]
+input = 0.05
+output = 0.1
diff --git a/providers/edenai/models/groq/openai/gpt-oss-safeguard-20b.toml b/providers/edenai/models/groq/openai/gpt-oss-safeguard-20b.toml
new file mode 100644
index 00000000000..bceeac7daec
--- /dev/null
+++ b/providers/edenai/models/groq/openai/gpt-oss-safeguard-20b.toml
@@ -0,0 +1,11 @@
+base_model = "openai/gpt-oss-safeguard-20b"
+name = "GPT OSS Safeguard 20B (Groq)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.075
+output = 0.3
+cache_read = 0.0375
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 646d2146fb0..1930d4b5b28 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -5,5 +5,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75504
-output = 0.75504
+input = 0.75348
+output = 0.75348
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 32722586cef..9f8263abe1f 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17424
-output = 0.75504
+input = 0.17388
+output = 0.75348
diff --git a/providers/edenai/models/nebius/google/gemma-3-27b-it.toml b/providers/edenai/models/nebius/google/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..3d6e1bdc836
--- /dev/null
+++ b/providers/edenai/models/nebius/google/gemma-3-27b-it.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B IT (Nebius)"
+structured_output = true
+
+[cost]
+input = 0.1
+output = 0.3
+cache_read = 0.1
+
+[limit]
+context = 110_000
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index eb910dd9b4a..cedc6b730a8 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46464
-output = 0.92928
+input = 0.46368
+output = 0.92736
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gemma-3-27b-it.toml b/providers/edenai/models/scaleway/gemma-3-27b-it.toml
new file mode 100644
index 00000000000..48792411115
--- /dev/null
+++ b/providers/edenai/models/scaleway/gemma-3-27b-it.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemma-3-27b-it"
+name = "Gemma 3 27B IT (Scaleway)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.287125
+output = 0.57425
+
+[limit]
+context = 40_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 55e8357eee7..d2d829b7f23 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17424
-output = 0.69696
+input = 0.17388
+output = 0.69552
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 1d9a07d1077..078fcc3eabf 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -4,5 +4,5 @@ attachment = false
 structured_output = false
 
 [cost]
-input = 1.045439
-output = 1.045439
+input = 1.04328
+output = 1.04328

From b76a180ec92e873b3f35c17130e9e0580c38b54a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 16:27:39 +0000
Subject: [PATCH 1130/2012] chore(sync): update NanoGPT model catalog (#6891)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-astra-latest.toml       | 29 +++++++++++++++++++
 .../models/openai/gpt-luna-latest.toml        | 29 +++++++++++++++++++
 .../models/openai/gpt-sol-latest.toml         | 29 +++++++++++++++++++
 .../models/openai/gpt-terra-latest.toml       | 29 +++++++++++++++++++
 4 files changed, 116 insertions(+)
 create mode 100644 providers/nano-gpt/models/openai/gpt-astra-latest.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-luna-latest.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-sol-latest.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-terra-latest.toml

diff --git a/providers/nano-gpt/models/openai/gpt-astra-latest.toml b/providers/nano-gpt/models/openai/gpt-astra-latest.toml
new file mode 100644
index 00000000000..35d4f3bcd9f
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-astra-latest.toml
@@ -0,0 +1,29 @@
+name = "GPT Astra Latest"
+description = "Compatibility alias that routes to GPT 6 Astra, the latest supported GPT Astra model."
+family = "gpt-astra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-luna-latest.toml b/providers/nano-gpt/models/openai/gpt-luna-latest.toml
new file mode 100644
index 00000000000..c5990d9e154
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-luna-latest.toml
@@ -0,0 +1,29 @@
+name = "GPT Luna Latest"
+description = "Compatibility alias that routes to GPT 5.6 Luna, the latest supported GPT Luna model."
+family = "gpt-luna"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-sol-latest.toml b/providers/nano-gpt/models/openai/gpt-sol-latest.toml
new file mode 100644
index 00000000000..5a374a5f702
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-sol-latest.toml
@@ -0,0 +1,29 @@
+name = "GPT Sol Latest"
+description = "Compatibility alias that routes to GPT 5.6 Sol, the latest supported GPT Sol model."
+family = "gpt-sol"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-terra-latest.toml b/providers/nano-gpt/models/openai/gpt-terra-latest.toml
new file mode 100644
index 00000000000..5cabdf7794c
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-terra-latest.toml
@@ -0,0 +1,29 @@
+name = "GPT Terra Latest"
+description = "Compatibility alias that routes to GPT 5.6 Terra, the latest supported GPT Terra model."
+family = "gpt-terra"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

From 16fe1b0a4a9f8f91459926c19c9a3c63db0c1111 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 16:27:49 +0000
Subject: [PATCH 1131/2012] chore(sync): update Merge Gateway model catalog
 (#6890)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 providers/merge-gateway/models/sakana/fugu-ultra.toml      | 7 +++++--
 3 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index d7d8d42decc..f0885b3822b 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +10,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 0264c6bc733..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/sakana/fugu-ultra.toml b/providers/merge-gateway/models/sakana/fugu-ultra.toml
index d1a2d6507ff..cdb82eedfdf 100644
--- a/providers/merge-gateway/models/sakana/fugu-ultra.toml
+++ b/providers/merge-gateway/models/sakana/fugu-ultra.toml
@@ -1,7 +1,10 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=sakana%2Ffugu-ultra (accessed 2026-07-21)
 base_model = "sakana/fugu-ultra"
 structured_output = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
 
 [cost]
 input = 5
@@ -9,4 +12,4 @@ output = 30
 cache_read = 0.5
 
 [limit]
-output = 250_000
+output = 128_000

From cb755db040fd7c49cdb1519f4d5ee4616b709ed3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 16:27:57 +0000
Subject: [PATCH 1132/2012] chore(sync): update Kilo model catalog (#6893)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml      |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 114538a3af7..2829ed5b448 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 0.8727
+output = 3.36
+cache_read = 0.1639
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 14c48b3489809164502593aa3033ed265e4d2d94 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 16:27:59 +0000
Subject: [PATCH 1133/2012] chore(sync): update OpenRouter model catalog
 (#6892)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 713b581e8d8..0a16c6a0710 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08652
-output = 0.17304
-cache_read = 0.017304
+input = 0.08624
+output = 0.17248
+cache_read = 0.017248
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 944ea7f1a30..206243d372f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.947778
-output = 1.895556
-cache_read = 0.078982
+input = 0.859908
+output = 1.719816
+cache_read = 0.071659
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 2b14e7625ce..11d35e9f1e7 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 0.8727
+output = 3.36
+cache_read = 0.1639
 
 [limit]
 context = 1_310_720
-output = 128_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 351c57e1760933796ab83c2f5949af5439b5582a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 17:23:52 +0000
Subject: [PATCH 1134/2012] chore(sync): update OpenRouter model catalog
 (#6895)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml              | 7 ++++---
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0a16c6a0710..02b6932bc45 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08624
-output = 0.17248
-cache_read = 0.017248
+input = 0.08596
+output = 0.17192
+cache_read = 0.017192
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 64652242a0e..f6e49b1cd7a 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.966
-output = 3.036
-cache_read = 0.1932
+input = 0.6
+output = 2
+cache_read = 0.15
 
 [limit]
 context = 1_048_576
+output = 182_476

From 561617877b178656deca8ff49900de3695a3f75f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 17:23:56 +0000
Subject: [PATCH 1135/2012] chore(sync): update Kilo model catalog (#6897)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.2.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index 521ea40a372..d1193203e34 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -9,3 +9,7 @@ values = ["none", "high", "xhigh"]
 input = 1.4
 output = 4.4
 cache_read = 0.26
+
+[limit]
+context = 202_752
+output = 182_476

From d0680accdbda2389322d38b5e507501c15f20442 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 17:24:26 +0000
Subject: [PATCH 1136/2012] chore(sync): update Venice model catalog (#6896)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/deepseek-v4-pro-0813.toml           | 5 ++++-
 providers/venice/models/mistral-small-3-2-24b-instruct.toml | 4 ++--
 2 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/providers/venice/models/deepseek-v4-pro-0813.toml b/providers/venice/models/deepseek-v4-pro-0813.toml
index 84bd6185cc1..df6b2765085 100644
--- a/providers/venice/models/deepseek-v4-pro-0813.toml
+++ b/providers/venice/models/deepseek-v4-pro-0813.toml
@@ -2,7 +2,10 @@ base_model = "deepseek/deepseek-v4-pro-0813"
 description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
 release_date = "2026-08-14"
 last_updated = "2026-08-14"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 1.65
diff --git a/providers/venice/models/mistral-small-3-2-24b-instruct.toml b/providers/venice/models/mistral-small-3-2-24b-instruct.toml
index ab8412dcf3b..5407b754356 100644
--- a/providers/venice/models/mistral-small-3-2-24b-instruct.toml
+++ b/providers/venice/models/mistral-small-3-2-24b-instruct.toml
@@ -3,7 +3,7 @@ description = "Efficient Mistral model for fast chat, extraction, and production
 family = "mistral-small"
 release_date = "2026-01-15"
 last_updated = "2026-06-11"
-attachment = false
+attachment = true
 reasoning = false
 tool_call = true
 structured_output = true
@@ -18,5 +18,5 @@ context = 256_000
 output = 16_384
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]

From c05362a20c299d80c710311d5c7f1c096eb3a706 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 18:29:59 +0000
Subject: [PATCH 1137/2012] chore(sync): update OpenRouter model catalog
 (#6900)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 02b6932bc45..a1e3802eb9d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08596
-output = 0.17192
-cache_read = 0.017192
+input = 0.08568
+output = 0.17136
+cache_read = 0.017136
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 206243d372f..944ea7f1a30 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.859908
-output = 1.719816
-cache_read = 0.071659
+input = 0.947778
+output = 1.895556
+cache_read = 0.078982
 
 [limit]
 context = 1_048_576

From c9ee1de53dd91233595aebad50eb0f75414950c3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 18:30:01 +0000
Subject: [PATCH 1138/2012] chore(sync): update Requesty model catalog (#6901)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/gpt-5.6-sol@eu.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/requesty/models/gpt-5.6-sol@eu.toml b/providers/requesty/models/gpt-5.6-sol@eu.toml
index ae290093a78..ba13b9db3e1 100644
--- a/providers/requesty/models/gpt-5.6-sol@eu.toml
+++ b/providers/requesty/models/gpt-5.6-sol@eu.toml
@@ -9,6 +9,6 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 5.5
-output = 33
-cache_read = 0.55
+input = 4.4
+output = 22
+cache_read = 0.44

From 3d77c0ccc4e4a259fba38470fad73451d9116b58 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 18:30:06 +0000
Subject: [PATCH 1139/2012] chore(sync): update Kilo model catalog (#6903)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/thinkingmachines/inkling:free.toml     | 15 ---------------
 1 file changed, 15 deletions(-)
 delete mode 100644 providers/kilo/models/thinkingmachines/inkling:free.toml

diff --git a/providers/kilo/models/thinkingmachines/inkling:free.toml b/providers/kilo/models/thinkingmachines/inkling:free.toml
deleted file mode 100644
index f9daa5b4864..00000000000
--- a/providers/kilo/models/thinkingmachines/inkling:free.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "thinkingmachines/inkling"
-name = "Thinking Machines: Inkling (free)"
-description = "Inkling is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 41B active parameters out of 975B total. It is designed for general-purpose reasoning, coding, agentic and tool-use systems,..."
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "max"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-output = 262_144

From 030afe2234c656d38be76bfdd0b0d34bd7e950af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 19:22:58 +0000
Subject: [PATCH 1140/2012] chore(sync): update Merge Gateway model catalog
 (#6906)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml            | 7 +++++--
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 5 files changed, 20 insertions(+), 15 deletions(-)

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
index d4498af2eee..91f9e45dc56 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-flash-0731.toml
@@ -8,6 +8,9 @@ type = "effort"
 values = ["none", "low", "medium", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
+input = 0.035
+output = 0.07
 cache_read = 0.007
+
+[limit]
+context = 1_048_576
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index b86a0e589cd..5ef98e7a3a0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index f0885b3822b..d7d8d42decc 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -10,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 42aa3ae2951..f966f69d01c 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 84592d1b8dd..0264c6bc733 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From 25f0eec44f29df2ca2514664417be03794fc0498 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 19:23:00 +0000
Subject: [PATCH 1141/2012] chore(sync): update Kilo model catalog (#6907)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 4ff6f302821..ab70d74a6d7 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 context = 1_048_576

From ca1b5e01544718602230a7d939f9fa976606e532 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 19:23:12 +0000
Subject: [PATCH 1142/2012] chore(sync): update OpenRouter model catalog
 (#6905)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a1e3802eb9d..4dca8a54662 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08568
-output = 0.17136
-cache_read = 0.017136
+input = 0.0854
+output = 0.1708
+cache_read = 0.01708
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 944ea7f1a30..206243d372f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.947778
-output = 1.895556
-cache_read = 0.078982
+input = 0.859908
+output = 1.719816
+cache_read = 0.071659
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index c831cb49220..af500f205e8 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index f78c60f8145..66be7b84546 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.34
-output = 11.7
-cache_read = 0.261
+input = 2.1
+output = 10.53
+cache_read = 0.235
 
 [limit]
 context = 1_048_576

From bfd3681e7577bf041628d783e4bf6c03f8c9f946 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rub=C3=A9n=20Solano?=
 <41839257+rsolanodev@users.noreply.github.com>
Date: Fri, 11 Sep 2026 21:50:54 +0200
Subject: [PATCH 1143/2012] fix(nan): align model catalog with nan.builders
 docs (#6908)

* fix(nan): align model catalog with nan.builders docs

Source: https://nan.builders/docs/models (accessed 2026-09-11)

- deepseek-v4-flash: base on deepseek/deepseek-v4-flash-vision-exp; the host serves the Vision-Exp variant, so it takes images (text+image).

- Replace glm5.2 with glm5.3: the docs list GLM 5.3 as the premium model (~753B, text-only, 1M context, 3,000M/billing period, 400M per rolling 4h).

- glm5.3-flash: correct monthly quota to 2B, keep text+image input.

- qwen3.8-flash: correct monthly quota to 500M.

- gemma4/qwen3.6/mimo-v2.5: comment-only corrections to match the docs.

* fix(nan): point deepseek-v4-flash at the DeepSeek V4.1 redirect

The docs describe this id as the Vision-Exp variant, but the host redirects deepseek-v4-flash to DeepSeek V4.1 Flash. Both are text+image, so the resolved capabilities are unchanged.

* fix(nan): evidence the deepseek-v4-flash redirect and align its name

The id deepseek-v4-flash serves DeepSeek V4.1 Flash: forced-choice self-identification with randomized order returned deepseek/deepseek-v4.1-flash 8/8, deepseek-v4.1-flash is not directly callable (401), and it takes images, matching DeepSeek's own legacy alias. Drop the name override so the display name matches the base.
---
 providers/nan/models/deepseek-v4-flash.toml | 25 ++++++++++++---------
 providers/nan/models/gemma4.toml            |  6 ++---
 providers/nan/models/glm5.2.toml            | 18 ---------------
 providers/nan/models/glm5.3-flash.toml      | 11 +++++----
 providers/nan/models/glm5.3.toml            | 14 ++++++++++++
 providers/nan/models/mimo-v2.5.toml         | 15 +++++--------
 providers/nan/models/qwen3.6.toml           | 11 +++++----
 providers/nan/models/qwen3.8-flash.toml     |  8 +++----
 providers/nan/provider.toml                 |  2 +-
 9 files changed, 51 insertions(+), 59 deletions(-)
 delete mode 100644 providers/nan/models/glm5.2.toml
 create mode 100644 providers/nan/models/glm5.3.toml

diff --git a/providers/nan/models/deepseek-v4-flash.toml b/providers/nan/models/deepseek-v4-flash.toml
index f232549beb2..3a5176c1d84 100644
--- a/providers/nan/models/deepseek-v4-flash.toml
+++ b/providers/nan/models/deepseek-v4-flash.toml
@@ -1,14 +1,17 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# 284B-21B MoE, FP8, 1M context. 2B token monthly quota per member, 1.5M tpm.
-# Served from the shared community GPU cluster, so no per-token price: included
-# in the membership.
-# Reasoning: no caller control on this surface. NaN lists a reasoning mode but
-# exposes no request field for it anywhere in its docs (only gemma4/qwen3.6 show
-# `reasoning_config={}`) and no graded effort, so neither the DeepSeek toggle nor
-# its low|high|max levels reach this host.
-# The docs/examples opencode.json snippet caps contextWindow at 500000; the
-# models reference publishes 1M, which is what is recorded here.
-base_model = "deepseek/deepseek-v4-flash"
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# NaN documents this id as the Vision-Exp variant, but the endpoint serves
+# DeepSeek V4.1 Flash. Evidence (2026-09-11): forced-choice self-identification
+# with randomized option order returned deepseek/deepseek-v4.1-flash 8/8; the id
+# deepseek-v4.1-flash is not directly callable here (401) while deepseek-v4-flash
+# is, matching DeepSeek's own legacy alias
+# (providers/deepseek/models/deepseek-v4-flash.toml). The model takes images
+# (text+image), consistent with V4.1 Flash; NaN serves open-weight models, which
+# V4.1 Flash is (MIT) and the experimental Vision-Exp entry is not.
+# 1M token context, 3B token monthly quota per member, 1.5M tpm. Included in the
+# membership, so no per-token price.
+# Reasoning: no caller control on this surface; NaN's request examples document
+# no reasoning field and reasoning_effort/thinking produced no measurable effect.
+base_model = "deepseek/deepseek-v4.1-flash"
 reasoning_options = []
 
 [cost]
diff --git a/providers/nan/models/gemma4.toml b/providers/nan/models/gemma4.toml
index 35f7fb1e841..b457c6aca22 100644
--- a/providers/nan/models/gemma4.toml
+++ b/providers/nan/models/gemma4.toml
@@ -1,6 +1,6 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# 26B-A4B MoE, FP8, 256K context, vision, XML tool calling. No token counter,
-# 1.5M tpm. Sampling defaults temp=0.6, top_p=0.95.
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# 26B-A4B MoE, FP8, 256K token context, multimodal with vision, XML tool
+# calling, 1.5M tpm. Sampling defaults temp=0.6, top_p=0.95.
 # Included in the membership, so no per-token price.
 # Toggle: `reasoning_config={}` in the request body enables reasoning mode --
 # the only reasoning field NaN documents. No graded effort.
diff --git a/providers/nan/models/glm5.2.toml b/providers/nan/models/glm5.2.toml
deleted file mode 100644
index 876a00ec212..00000000000
--- a/providers/nan/models/glm5.2.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# ~753B MoE, FP8, sparse attention, text only. Callable only with a key on the
-# 200 EUR/month GLM 5.2 premium membership: 3,000M tokens per billing period and
-# 400M tokens per rolling 4h window, 5 concurrent requests. Context is 500K on
-# this host, not the 1M of the weights. Included in the tier, so no per-token price.
-# Reasoning: no caller control on this surface. NaN documents a reasoning trace
-# but no request field for it, and the high|max effort levels of the first-party
-# Z.ai API are not exposed here.
-base_model = "zhipuai/glm-5.2"
-description = "Open flagship GLM for long-horizon coding agents, served at 500K context on the NaN premium tier"
-reasoning_options = []
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 500_000
diff --git a/providers/nan/models/glm5.3-flash.toml b/providers/nan/models/glm5.3-flash.toml
index 550bda3827e..7a7e51240b7 100644
--- a/providers/nan/models/glm5.3-flash.toml
+++ b/providers/nan/models/glm5.3-flash.toml
@@ -1,10 +1,9 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# 320B-18B MoE, FP8, natively multimodal. 1M context. 500M token monthly quota
-# per member. MIT license. No video or PDF input on this host.
-# Included in the membership, so no per-token price.
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# 320B-18B MoE, FP8, natively multimodal. 1M token context. 2B token monthly
+# quota per member. MIT license. Input is text and image on this host: no video
+# or PDF. Included in the membership, so no per-token price.
 # Reasoning: no caller control on this surface. NaN lists a reasoning mode but
-# exposes no request field for it, and the low|high|max effort levels of the
-# first-party Z.ai API are not exposed here.
+# exposes no request field for it.
 base_model = "zhipuai/glm-5.3-flash"
 reasoning_options = []
 
diff --git a/providers/nan/models/glm5.3.toml b/providers/nan/models/glm5.3.toml
new file mode 100644
index 00000000000..0bdc1abc4ad
--- /dev/null
+++ b/providers/nan/models/glm5.3.toml
@@ -0,0 +1,14 @@
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# ~753B MoE, FP8, sparse attention, text only: it does not take images.
+# Premium tier: callable only with a key on the GLM 5.3 premium membership.
+# 1M token context, 3,000M token quota per billing period, capped at 400M tokens
+# per rolling 4h window, 5 concurrent requests. Included in the tier, so no
+# per-token price.
+# Reasoning: no caller control on this surface. NaN documents a reasoning trace
+# but no request field for it.
+base_model = "zhipuai/glm-5.3"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/nan/models/mimo-v2.5.toml b/providers/nan/models/mimo-v2.5.toml
index 882287acf30..e14eac593f0 100644
--- a/providers/nan/models/mimo-v2.5.toml
+++ b/providers/nan/models/mimo-v2.5.toml
@@ -1,15 +1,10 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
 # 310B-15B MoE, FP8, natively omnimodal with dedicated vision and audio encoders.
-# 1M context. 1.0B token monthly quota per member, 1.5M tpm. MIT license.
-# No video input on this host. Included in the membership, so no per-token price.
+# 1M token context, 1.0B token monthly quota per member, 1.5M tpm. MIT license.
+# Input is text, image and audio on this host: no video. Included in the
+# membership, so no per-token price.
 # Reasoning: no caller control on this surface. NaN lists a reasoning mode
-# (recommends max_tokens >= 300) but exposes no request field for it and no
-# graded effort.
-# Context is inherited: NaN publishes "1M tokens", the same wording it uses for
-# deepseek-v4-flash and glm5.3-flash, with no documented trim of the model's
-# native 1_048_576 window. The docs/examples opencode.json snippet caps
-# contextWindow at 500000, but that snippet also trims maxTokens to a
-# per-request value rather than the ceiling.
+# (recommends max_tokens >= 300) but exposes no request field for it.
 base_model = "xiaomi/mimo-v2.5"
 reasoning_options = []
 
diff --git a/providers/nan/models/qwen3.6.toml b/providers/nan/models/qwen3.6.toml
index fb47cc0b56a..72da1d4c294 100644
--- a/providers/nan/models/qwen3.6.toml
+++ b/providers/nan/models/qwen3.6.toml
@@ -1,9 +1,8 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# 35B total MoE, 3B active per token, FP8, 256K context, vision, XML tool
-# calling. MTP speculative decoding (~2x throughput). No token counter, 1.5M tpm.
-# Sampling defaults temp=0.6, top_p=0.95. NaN's own client configs confirm
-# maxTokens 65536 as the model ceiling. Included in the membership, so no
-# per-token price.
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# Flagship model. 35B total MoE, 3B active per token, FP8, 256K token context,
+# multimodal with vision, XML tool calling, MTP speculative decoding
+# (~2x throughput), 1.5M tpm. Sampling defaults temp=0.6, top_p=0.95.
+# Included in the membership, so no per-token price.
 # Toggle: `reasoning_config={}` in the request body enables reasoning mode --
 # the only reasoning field NaN documents. No graded effort.
 base_model = "alibaba/qwen3.6-35b-a3b"
diff --git a/providers/nan/models/qwen3.8-flash.toml b/providers/nan/models/qwen3.8-flash.toml
index dd94ab609d8..072346dc409 100644
--- a/providers/nan/models/qwen3.8-flash.toml
+++ b/providers/nan/models/qwen3.8-flash.toml
@@ -1,9 +1,9 @@
-# Source: https://nan.builders/docs/models (accessed 2026-09-03)
-# 125B-6B MoE with vision, XML tool calling. 262K context, the model's native
-# window. 1B token monthly quota per member. License qwen-community-1.0.
+# Source: https://nan.builders/docs/models (accessed 2026-09-11)
+# 125B-6B MoE, multimodal with vision, 262K token context (the model's native
+# window), 500M token monthly quota per member. qwen-community-1.0 license.
 # Included in the membership, so no per-token price.
 # Reasoning: on by default, with no caller control on this surface. NaN exposes
-# no request field to disable it and no graded effort.
+# no request field to disable it.
 base_model = "alibaba/qwen3.8-flash"
 reasoning_options = []
 
diff --git a/providers/nan/provider.toml b/providers/nan/provider.toml
index 6359c2fb7e6..c47082dd0d3 100644
--- a/providers/nan/provider.toml
+++ b/providers/nan/provider.toml
@@ -1,6 +1,6 @@
 # NaN is a builders community that shares dedicated GPUs to serve open-weight models.
 # Access is via LiteLLM with an OpenAI-compatible API at https://api.nan.builders/v1.
-# Membership based (70 EUR/month, plus a 200 EUR/month tier that unlocks glm5.2), so
+# Membership based (70 EUR/month, plus a GLM 5.3 premium tier), so
 # per-token cost is 0: usage is metered as monthly token quotas, not billed per token.
 # Rate limits: 60 rpm and 5 concurrent requests per API key.
 # https://nan.builders/docs/getting-started (accessed 2026-09-03)

From 6e5cc5628e63bd41419bb067d661e096d7d19a08 Mon Sep 17 00:00:00 2001
From: realtimegb <support@realtimecomms.co.uk>
Date: Fri, 11 Sep 2026 20:51:14 +0100
Subject: [PATCH 1144/2012] above: DeepSeek V4 Flash price and base after the
 V4.1 Flash launch (#6904)

DeepSeek released V4.1 Flash on 2026-09-10 (04:00 UTC) and lowered the
Flash price. The deepseek-v4-flash id now serves V4.1 Flash, so the
base_model moves to deepseek/deepseek-v4.1-flash (image input verified
through above.dev) and cost moves to the new upstream rate x1.10:
$0.15 in / $0.003 cache-hit / $0.60 out per million, off-peak.

Sources are cited in the file header: DeepSeek pricing page, DeepSeek
thinking-mode guide, and above.dev docs.

deepseek-v4-pro is deliberately unchanged: DeepSeek announced its
retirement in favour of V4.1 Flash, then cancelled that on 2026-09-11
"in response to user demand", with billing unchanged.

Co-authored-by: realtimegb <pkreil3@gmail.com>
---
 providers/above/models/deepseek-v4-flash.toml | 34 +++++++++++--------
 1 file changed, 19 insertions(+), 15 deletions(-)

diff --git a/providers/above/models/deepseek-v4-flash.toml b/providers/above/models/deepseek-v4-flash.toml
index d8ff403d9ea..9b61b977bf8 100644
--- a/providers/above/models/deepseek-v4-flash.toml
+++ b/providers/above/models/deepseek-v4-flash.toml
@@ -1,14 +1,18 @@
-# Passthrough to the official DeepSeek API.
-# Toggle: `thinking.type = enabled|disabled`; effort: `reasoning_effort = low|high|max`
-# (Flash maps requested low→low). https://api-docs.deepseek.com/guides/thinking_mode/
-# Off-peak rate from the CURRENT DeepSeek price sheet effective 2026-08-16
-# (flash $0.22/$0.007/$0.66, pro $0.66/$0.022/$1.98 off-peak) plus a flat 10%;
-# peak (weekdays 01:00-04:00 and 06:00-10:00 UTC, Beijing-time weekends
-# off-peak) is 2x. https://api-docs.deepseek.com/quick_start/pricing/
-# (accessed 2026-08-29; note first-party catalog entries predate 2026-08-16).
-# Toggle verified end-to-end on this host 2026-08-29.
-# Reasoning tokens billed at the output rate.
-base_model = "deepseek/deepseek-v4-flash-0731"
+# Passthrough to the official DeepSeek API. Since 2026-09-10 04:00 UTC the
+# `deepseek-v4-flash` id is served by DeepSeek-V4.1-Flash and billed at the
+# Flash price, so the base is the V4.1 lab entry. Image input verified
+# end-to-end through above.dev on 2026-09-11.
+# Sources:
+#   https://api-docs.deepseek.com/quick_start/pricing  (upstream Flash rate:
+#     $0.15 in / $0.003 cache-hit / $0.60 out per MTok, off-peak; peak is 2x on
+#     weekdays 01:00-04:00 and 06:00-10:00 UTC; Beijing-time weekends off-peak)
+#   https://api-docs.deepseek.com/guides/thinking_mode/  (toggle
+#     `thinking.type = enabled|disabled`, effort `reasoning_effort = low|high|max`)
+#   https://above.dev/docs  (above resells every upstream rate at a flat x1.10,
+#     listed here at the off-peak rate; cost headers on each response)
+# V4 Pro is deliberately unchanged: DeepSeek cancelled its retirement on
+# 2026-09-11 "in response to user demand". Reasoning billed at the output rate.
+base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4 Flash"
 
 [[reasoning_options]]
@@ -22,7 +26,7 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.242
-output = 0.726
-reasoning = 0.726
-cache_read = 0.0077
+input = 0.165
+output = 0.66
+reasoning = 0.66
+cache_read = 0.0033

From 9a29841349fa4191a02b2a899e688ebfec679a2e Mon Sep 17 00:00:00 2001
From: rakshith1928 <rakshithn6485@gmail.com>
Date: Sat, 12 Sep 2026 01:21:52 +0530
Subject: [PATCH 1145/2012] feat(aihubmix): add deepseek-v4.1-flash (#6899)

* feat(aihubmix): add deepseek-v4.1-flash (lab + provider)

* chore(deepseek): remove deepseek-v4.1-flash lab metadata

---------

Co-authored-by: rakshith1928 <rakshith1928@users.noreply.github.com>
---
 .../aihubmix/models/deepseek-v4.1-flash.toml     | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/aihubmix/models/deepseek-v4.1-flash.toml

diff --git a/providers/aihubmix/models/deepseek-v4.1-flash.toml b/providers/aihubmix/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..026ceb4058b
--- /dev/null
+++ b/providers/aihubmix/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,16 @@
+# AIHubMix pricing per https://aihubmix.com/model/deepseek-v4.1-flash (accessed 2026-09-11).
+# Base/off-peak rate: input $0.155, output $0.62, cache $0.0031 per 1M tokens.
+# Peak rate (01:00-04:00, 06:00-10:00 UTC) is 2x: input $0.3098, output $1.2392, cache $0.0062.
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort = low|high|max
+# AIHubMix effort levels returned HTTP 200 (validated 2026-08-31T04:17:24Z).
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.155
+output = 0.62
+cache_read = 0.0031
\ No newline at end of file

From e14f58d376d544b2c75cc7a321803f4f34458921 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:52:15 -0500
Subject: [PATCH 1146/2012] chore(sync): update Cortecs model catalog (#6846)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/gemma-3-27b-it.toml | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

diff --git a/providers/cortecs/models/gemma-3-27b-it.toml b/providers/cortecs/models/gemma-3-27b-it.toml
index ee11f1ec4c3..b205d78270c 100644
--- a/providers/cortecs/models/gemma-3-27b-it.toml
+++ b/providers/cortecs/models/gemma-3-27b-it.toml
@@ -1,13 +1,9 @@
-name = "gemma-3-27b-it"
+base_model = "google/gemma-3-27b-it"
 description = "Gemma 3 is a family of lightweight, multimodal models from Google, supporting text and image inputs, multilingual capabilities, and a 131K context window."
-release_date = "2025-03-12"
-last_updated = "2025-03-12"
-attachment = true
 reasoning = true
 temperature = false
 tool_call = false
 structured_output = true
-open_weights = false
 reasoning_options = []
 
 [cost]
@@ -17,7 +13,3 @@ output = 0.299
 [limit]
 context = 131_000
 output = 110_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 931a4b8418944cd15f24a287eed7d061b8ad182b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 20:26:06 +0000
Subject: [PATCH 1147/2012] chore(sync): update Kilo model catalog (#6911)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index ab70d74a6d7..74831ee82da 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 context = 1_048_576

From 10747c54f8c788f7420293b01639087f69cde53f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 20:26:09 +0000
Subject: [PATCH 1148/2012] chore(sync): update Cortecs model catalog (#6910)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/deepseek-v4-flash-0731.toml | 6 +++---
 providers/cortecs/models/glm-5.3-flash.toml          | 6 +++---
 providers/cortecs/models/qwen3.8-27b.toml            | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/cortecs/models/deepseek-v4-flash-0731.toml b/providers/cortecs/models/deepseek-v4-flash-0731.toml
index cbbf634182a..a6037a9e2fe 100644
--- a/providers/cortecs/models/deepseek-v4-flash-0731.toml
+++ b/providers/cortecs/models/deepseek-v4-flash-0731.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.13
-output = 0.28
-cache_read = 0.03
+input = 0.09
+output = 0.17
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/cortecs/models/glm-5.3-flash.toml b/providers/cortecs/models/glm-5.3-flash.toml
index 6f2582b9c88..7a151f91d7a 100644
--- a/providers/cortecs/models/glm-5.3-flash.toml
+++ b/providers/cortecs/models/glm-5.3-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.201
-output = 0.5
-cache_read = 0.05
+input = 0.1
+output = 0.35
+cache_read = 0.018
 
 [limit]
 context = 1_048_576
diff --git a/providers/cortecs/models/qwen3.8-27b.toml b/providers/cortecs/models/qwen3.8-27b.toml
index c1b61cfe923..49807837ad1 100644
--- a/providers/cortecs/models/qwen3.8-27b.toml
+++ b/providers/cortecs/models/qwen3.8-27b.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.334
-output = 2.451
-cache_read = 0.111
+input = 0.1
+output = 0.4
+cache_read = 0.04
 
 [limit]
 output = 262_144

From 460ff5ded884d6070f9fdb2a2770b7d2d9a72c22 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 20:26:29 +0000
Subject: [PATCH 1149/2012] chore(sync): update OpenRouter model catalog
 (#6912)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 4dca8a54662..ac4930e9112 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0854
-output = 0.1708
-cache_read = 0.01708
+input = 0.08512
+output = 0.17024
+cache_read = 0.017024
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 206243d372f..944ea7f1a30 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.859908
-output = 1.719816
-cache_read = 0.071659
+input = 0.947778
+output = 1.895556
+cache_read = 0.078982
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index af500f205e8..975660fffb4 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 66be7b84546..6eb528f593d 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.53
-cache_read = 0.235
+input = 1.89
+output = 9.48
+cache_read = 0.216
 
 [limit]
 context = 1_048_576

From 6f0079acd6c2af10cc7608ec38c12b42a733a1a0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 21:22:50 +0000
Subject: [PATCH 1150/2012] chore(sync): update Merge Gateway model catalog
 (#6916)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-27b.toml       | 7 +++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 5ef98e7a3a0..b86a0e589cd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +11,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index d7d8d42decc..f0885b3822b 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,6 +1,5 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +10,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From d42f3cdd108955fb4f3ae709498fe69e436ebdec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 21:22:54 +0000
Subject: [PATCH 1151/2012] chore(sync): update LLM Gateway model catalog
 (#6914)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/deepseek-v4-flash.toml              | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
index 5e6b181be08..20a72c32212 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4-flash.toml
@@ -14,4 +14,5 @@ output = 0.1
 cache_read = 0.01
 
 [limit]
+context = 1_050_000
 output = 393_216

From 797920734b6eeb6b96e4a265b068f08f4ce61555 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 21:23:05 +0000
Subject: [PATCH 1152/2012] chore(sync): update OpenRouter model catalog
 (#6913)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../models/mistralai/mistral-small-3.2-24b-instruct.toml    | 2 +-
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 5 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ac4930e9112..a878d771be8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08512
-output = 0.17024
-cache_read = 0.017024
+input = 0.08484
+output = 0.16968
+cache_read = 0.016968
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 944ea7f1a30..206243d372f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.947778
-output = 1.895556
-cache_read = 0.078982
+input = 0.859908
+output = 1.719816
+cache_read = 0.071659
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml b/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
index e56ed9bddb4..a05f6600220 100644
--- a/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
+++ b/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
@@ -16,7 +16,7 @@ input = 0.075
 output = 0.2
 
 [limit]
-context = 131_072
+context = 256_000
 output = 16_384
 
 [modalities]
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 975660fffb4..275f3bc4fab 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 1.7955
+output = 9.006
+cache_read = 0.2052
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 6eb528f593d..7e4c3b665b7 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 1.7955
+output = 9.006
+cache_read = 0.2052
 
 [limit]
 context = 1_048_576

From 531915dcf7f83ebbfb79aaa6fe77ec287617ae4e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 21:23:09 +0000
Subject: [PATCH 1153/2012] chore(sync): update Kilo model catalog (#6915)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3.5-lightning.toml | 5 ++---
 providers/kilo/models/~moonshotai/kimi-latest.toml       | 6 +++---
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index f6b9541a819..78104d7319f 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -6,9 +6,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.08
-output = 0.2
-cache_read = 0.04
+input = 0.065
+output = 0.18
 
 [limit]
 output = 131_072
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 74831ee82da..19dc926e329 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.89
-output = 9.48
-cache_read = 0.216
+input = 1.7955
+output = 9.006
+cache_read = 0.2052
 
 [limit]
 context = 1_048_576

From fd5f2a2a6fbd5a60010b0454f95be8f2085e10d4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 22:24:23 +0000
Subject: [PATCH 1154/2012] chore(sync): update OpenRouter model catalog
 (#6918)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a878d771be8..96acdd95652 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08484
-output = 0.16968
-cache_read = 0.016968
+input = 0.08456
+output = 0.16912
+cache_read = 0.016912
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 206243d372f..944ea7f1a30 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.859908
-output = 1.719816
-cache_read = 0.071659
+input = 0.947778
+output = 1.895556
+cache_read = 0.078982
 
 [limit]
 context = 1_048_576

From 8a5e9b711e3f782298c9edc4084b7edac4600d6e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 23:22:57 +0000
Subject: [PATCH 1155/2012] chore(sync): update Kilo model catalog (#6919)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 19dc926e329..2230214cdf5 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.7955
-output = 9.006
-cache_read = 0.2052
+input = 1.61595
+output = 8.1054
+cache_read = 0.18468
 
 [limit]
 context = 1_048_576

From e5ce1ab981509a84a286ddd451325655b4eb6df7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 23:23:03 +0000
Subject: [PATCH 1156/2012] chore(sync): update OpenRouter model catalog
 (#6920)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 96acdd95652..0695342727c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08456
-output = 0.16912
-cache_read = 0.016912
+input = 0.06762
+output = 0.13524
+cache_read = 0.013524
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 944ea7f1a30..1b4ffd8a683 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.947778
-output = 1.895556
-cache_read = 0.078982
+input = 0.849816
+output = 1.699632
+cache_read = 0.070818
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 275f3bc4fab..b0dbfcb86c8 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7955
-output = 9.006
-cache_read = 0.2052
+input = 1.61595
+output = 8.1054
+cache_read = 0.18468
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 7e4c3b665b7..3e99f503b4b 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7955
-output = 9.006
-cache_read = 0.2052
+input = 1.61595
+output = 8.1054
+cache_read = 0.18468
 
 [limit]
 context = 1_048_576

From ddbbb931a81e660b0ae535ffb3512bd556d85bf8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 00:57:19 +0000
Subject: [PATCH 1157/2012] chore(sync): update OpenRouter model catalog
 (#6926)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 7 ++++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml        | 6 +++---
 providers/openrouter/models/minimax/minimax-m2.5.toml    | 9 +++------
 providers/openrouter/models/moonshotai/kimi-k3.toml      | 6 +++---
 .../openrouter/models/qwen/qwen3-235b-a22b-2507.toml     | 7 ++++---
 providers/openrouter/models/qwen/qwen3.8-27b.toml        | 6 +++---
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 9 files changed, 29 insertions(+), 30 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0695342727c..cb7e272671c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06762
-output = 0.13524
-cache_read = 0.013524
+input = 0.06734
+output = 0.13468
+cache_read = 0.013468
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..7c435d69ac6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 0.57816
+output = 1.73448
+cache_read = 0.018396
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..1352c1ee599 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.4
-output = 0.4
+input = 0.72
+output = 0.72
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index ca75f8e4927..1edeb5ecdfc 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,9 +6,6 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.27
-output = 1.08
-cache_read = 0.027
-
-[limit]
-output = 128_000
+input = 0.3
+output = 1.2
+cache_read = 0.03
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index b0dbfcb86c8..d013f6a38a2 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.61595
-output = 8.1054
-cache_read = 0.18468
+input = 1.535153
+output = 7.70013
+cache_read = 0.175446
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
index 0711c94269c..0d691163001 100644
--- a/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-235b-a22b-2507.toml
@@ -12,12 +12,13 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.22
-output = 0.88
+input = 0.0875
+output = 0.35
+cache_read = 0.0175
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index d1ddcffa2ed..488923c2abf 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.42
-output = 3
-cache_read = 0.085
+input = 0.214
+output = 2.55
+cache_read = 0.15
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 3e99f503b4b..2111cf6b352 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.61595
-output = 8.1054
-cache_read = 0.18468
+input = 1.535153
+output = 7.70013
+cache_read = 0.175446
 
 [limit]
 context = 1_048_576

From 13824cbc7c3a4e082dc154d8ee1c4305a2e4733a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 00:57:24 +0000
Subject: [PATCH 1158/2012] chore(sync): update Kilo model catalog (#6928)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml    | 3 ++-
 .../kilo/models/meta-llama/llama-3.1-70b-instruct.toml      | 2 +-
 providers/kilo/models/minimax/minimax-m2.5.toml             | 4 ----
 providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml        | 2 +-
 providers/kilo/models/qwen/qwen3.8-27b.toml                 | 1 -
 providers/kilo/models/tencent/hy3.toml                      | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml          | 6 +++---
 7 files changed, 10 insertions(+), 14 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..6fc67901197 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index e89b49a7fbe..3f15c0bc21d 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,7 +10,3 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
-
-[limit]
-context = 200_000
-output = 128_000
diff --git a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
index 3f37ded3629..b356342358d 100644
--- a/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-235b-a22b-2507.toml
@@ -16,7 +16,7 @@ output = 0.598
 
 [limit]
 context = 262_144
-output = 16_384
+output = 235_929
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 820f69fb194..9b18fb98b33 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,5 +12,4 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
-context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 2230214cdf5..77de40dc16d 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.61595
-output = 8.1054
-cache_read = 0.18468
+input = 1.535153
+output = 7.70013
+cache_read = 0.175446
 
 [limit]
 context = 1_048_576

From 82193b54ac5616cd631143a19069dd27da149170 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 00:57:26 +0000
Subject: [PATCH 1159/2012] chore(sync): update DigitalOcean model catalog
 (#6927)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/glm-5.3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/digitalocean/models/glm-5.3.toml b/providers/digitalocean/models/glm-5.3.toml
index b736e699d4c..3325586991b 100644
--- a/providers/digitalocean/models/glm-5.3.toml
+++ b/providers/digitalocean/models/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 0.95
+output = 3.4
+cache_read = 0.2
 
 [limit]
 context = 1_048_576

From 4d2479914ad69f8af35ce02235e60d0630bdc763 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Fri, 11 Sep 2026 20:25:56 -0500
Subject: [PATCH 1160/2012] fix(amazon-bedrock): correct Meta, Mistral, and
 Writer catalog (#6791)

* fix(amazon-bedrock): correct Meta, Mistral, and Writer catalog

* fix(amazon-bedrock): preserve Palmyra X4 context

* fix(writer): preserve Palmyra update date

---------

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 models/meta/llama-3.3-70b-instruct.toml       |  2 +-
 models/meta/llama-4-scout-17b-instruct.toml   |  3 +-
 models/mistral/magistral-small-2509.toml      | 27 +++++++++++++++++
 .../ministral-3-14b-instruct-2512.toml        | 29 +++++++++++++++++++
 .../mistral/ministral-3-3b-instruct-2512.toml | 29 +++++++++++++++++++
 .../mistral/ministral-3-8b-instruct-2512.toml | 29 +++++++++++++++++++
 models/mistral/mistral-large-2512.toml        |  3 +-
 models/mistral/voxtral-mini-3b-2507.toml      | 29 +++++++++++++++++++
 models/mistral/voxtral-small-24b-2507.toml    | 29 +++++++++++++++++++
 models/writer/palmyra-x4.toml                 | 10 ++++---
 models/writer/palmyra-x5.toml                 |  9 ++++--
 .../eu.mistral.pixtral-large-2502-v1:0.toml   |  1 +
 .../meta.llama3-1-70b-instruct-v1:0.toml      | 21 ++------------
 .../meta.llama3-1-8b-instruct-v1:0.toml       | 21 ++------------
 .../meta.llama3-3-70b-instruct-v1:0.toml      | 21 ++------------
 ...eta.llama4-maverick-17b-instruct-v1:0.toml | 22 ++++----------
 .../meta.llama4-scout-17b-instruct-v1:0.toml  | 23 ++++-----------
 .../models/mistral.devstral-2-123b.toml       | 16 ++--------
 .../models/mistral.magistral-small-2509.toml  | 18 +++---------
 .../mistral.ministral-3-14b-instruct.toml     | 18 +++---------
 .../mistral.ministral-3-3b-instruct.toml      | 18 ++----------
 .../mistral.ministral-3-8b-instruct.toml      | 19 +++---------
 ...mistral.mistral-large-3-675b-instruct.toml | 17 ++---------
 .../mistral.pixtral-large-2502-v1:0.toml      | 21 ++------------
 .../models/mistral.voxtral-mini-3b-2507.toml  | 18 ++----------
 .../mistral.voxtral-small-24b-2507.toml       | 24 ++++-----------
 .../us.meta.llama3-3-70b-instruct-v1:0.toml   |  1 -
 ...eta.llama4-maverick-17b-instruct-v1:0.toml |  6 ++++
 ...s.meta.llama4-scout-17b-instruct-v1:0.toml |  7 +++++
 .../us.mistral.pixtral-large-2502-v1:0.toml   |  1 +
 .../models/us.writer.palmyra-x4-v1:0.toml     |  9 ++++--
 .../models/us.writer.palmyra-x5-v1:0.toml     | 14 +++++++--
 .../models/writer.palmyra-x4-v1:0.toml        | 18 +++---------
 .../models/writer.palmyra-x5-v1:0.toml        | 17 ++++-------
 34 files changed, 286 insertions(+), 264 deletions(-)
 create mode 100644 models/mistral/magistral-small-2509.toml
 create mode 100644 models/mistral/ministral-3-14b-instruct-2512.toml
 create mode 100644 models/mistral/ministral-3-3b-instruct-2512.toml
 create mode 100644 models/mistral/ministral-3-8b-instruct-2512.toml
 create mode 100644 models/mistral/voxtral-mini-3b-2507.toml
 create mode 100644 models/mistral/voxtral-small-24b-2507.toml

diff --git a/models/meta/llama-3.3-70b-instruct.toml b/models/meta/llama-3.3-70b-instruct.toml
index 2f48b022bda..246ab54c9ef 100644
--- a/models/meta/llama-3.3-70b-instruct.toml
+++ b/models/meta/llama-3.3-70b-instruct.toml
@@ -3,7 +3,7 @@ description = "Popular open Llama workhorse for multilingual chat, coding, and s
 family = "llama"
 release_date = "2024-12-06"
 last_updated = "2024-12-06"
-attachment = true
+attachment = false
 reasoning = false
 temperature = true
 tool_call = true
diff --git a/models/meta/llama-4-scout-17b-instruct.toml b/models/meta/llama-4-scout-17b-instruct.toml
index e381e973ce9..e129a0ceaac 100644
--- a/models/meta/llama-4-scout-17b-instruct.toml
+++ b/models/meta/llama-4-scout-17b-instruct.toml
@@ -1,3 +1,4 @@
+# https://github.com/meta-llama/llama-models/blob/main/models/llama4/MODEL_CARD.md
 name = "Llama 4 Scout 17B Instruct"
 description = "Open Llama with long-context vision for efficient multimodal agents"
 family = "llama"
@@ -11,7 +12,7 @@ knowledge = "2024-08"
 open_weights = true
 
 [limit]
-context = 3_500_000
+context = 10_000_000
 output = 16_384
 
 [modalities]
diff --git a/models/mistral/magistral-small-2509.toml b/models/mistral/magistral-small-2509.toml
new file mode 100644
index 00000000000..f4dcc38c9e9
--- /dev/null
+++ b/models/mistral/magistral-small-2509.toml
@@ -0,0 +1,27 @@
+# https://docs.mistral.ai/models/magistral-small-1-2-25-09
+# https://huggingface.co/mistralai/Magistral-Small-2509
+# https://huggingface.co/mistralai/Magistral-Small-2509/blob/main/generation_config.json
+# Native generation config sets max_new_tokens = 131072, matching the model card's max_tokens setting.
+name = "Magistral Small 1.2"
+description = "Open multimodal reasoning model for transparent analysis of text and images"
+family = "magistral"
+release_date = "2025-09-18"
+last_updated = "2025-09-18"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Magistral-Small-2509"
diff --git a/models/mistral/ministral-3-14b-instruct-2512.toml b/models/mistral/ministral-3-14b-instruct-2512.toml
new file mode 100644
index 00000000000..f39bfa4efe3
--- /dev/null
+++ b/models/mistral/ministral-3-14b-instruct-2512.toml
@@ -0,0 +1,29 @@
+# https://docs.mistral.ai/models/ministral-3-14b-25-12
+# https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512
+# https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512/blob/main/generation_config.json
+# Native examples request max_tokens = 262144; generation_config.max_length is also 262144.
+# This is a shared prompt/generation ceiling, not a host-specific independent output cap.
+name = "Ministral 3 14B"
+description = "Open vision-language model for efficient local deployment, instruction following, and tool use"
+family = "ministral"
+release_date = "2025-12-02"
+last_updated = "2025-12-02"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512"
diff --git a/models/mistral/ministral-3-3b-instruct-2512.toml b/models/mistral/ministral-3-3b-instruct-2512.toml
new file mode 100644
index 00000000000..e3b6fdfeea8
--- /dev/null
+++ b/models/mistral/ministral-3-3b-instruct-2512.toml
@@ -0,0 +1,29 @@
+# https://mistral.ai/news/mistral-3
+# https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512
+# https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512/blob/main/generation_config.json
+# Native examples request max_tokens = 262144; generation_config.max_length is also 262144.
+# This is a shared prompt/generation ceiling, not a host-specific independent output cap.
+name = "Ministral 3 3B"
+description = "Compact open vision-language model for edge deployment, instruction following, and tool use"
+family = "ministral"
+release_date = "2025-12-02"
+last_updated = "2025-12-02"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512"
diff --git a/models/mistral/ministral-3-8b-instruct-2512.toml b/models/mistral/ministral-3-8b-instruct-2512.toml
new file mode 100644
index 00000000000..493dc85c43c
--- /dev/null
+++ b/models/mistral/ministral-3-8b-instruct-2512.toml
@@ -0,0 +1,29 @@
+# https://mistral.ai/news/mistral-3
+# https://huggingface.co/mistralai/Ministral-3-8B-Instruct-2512
+# https://huggingface.co/mistralai/Ministral-3-8B-Instruct-2512/blob/main/generation_config.json
+# Native examples request max_tokens = 262144; generation_config.max_length is also 262144.
+# This is a shared prompt/generation ceiling, not a host-specific independent output cap.
+name = "Ministral 3 8B"
+description = "Compact open vision-language model for edge deployment, instruction following, and tool use"
+family = "ministral"
+release_date = "2025-12-02"
+last_updated = "2025-12-02"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Ministral-3-8B-Instruct-2512"
diff --git a/models/mistral/mistral-large-2512.toml b/models/mistral/mistral-large-2512.toml
index 7efdb8119a5..9905e0e1c5b 100644
--- a/models/mistral/mistral-large-2512.toml
+++ b/models/mistral/mistral-large-2512.toml
@@ -1,7 +1,8 @@
+# https://mistral.ai/news/mistral-3
 name = "Mistral Large 3"
 description = "Mistral's largest general model for enterprise agents, coding, and multilingual reasoning"
 family = "mistral-large"
-release_date = "2024-11-01"
+release_date = "2025-12-02"
 last_updated = "2025-12-02"
 attachment = true
 reasoning = false
diff --git a/models/mistral/voxtral-mini-3b-2507.toml b/models/mistral/voxtral-mini-3b-2507.toml
new file mode 100644
index 00000000000..af016cdd17d
--- /dev/null
+++ b/models/mistral/voxtral-mini-3b-2507.toml
@@ -0,0 +1,29 @@
+# https://mistral.ai/news/voxtral/
+# https://huggingface.co/mistralai/Voxtral-Mini-3B-2507
+# https://huggingface.co/mistralai/Voxtral-Mini-3B-2507/blob/main/params.json
+# https://docs.mistral.ai/api/endpoint/chat
+# Native params specify 32768 positions, matching the card's 32k context.
+# Output is this shared-window ceiling: prompt + max_tokens must fit; no separate native output cap is published.
+name = "Voxtral Mini 3B 2507"
+description = "Open audio-language model for speech transcription, audio understanding, and voice-driven tool use"
+family = "voxtral"
+release_date = "2025-07-15"
+last_updated = "2025-07-15"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text", "audio"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Voxtral-Mini-3B-2507"
diff --git a/models/mistral/voxtral-small-24b-2507.toml b/models/mistral/voxtral-small-24b-2507.toml
new file mode 100644
index 00000000000..c42b7c07608
--- /dev/null
+++ b/models/mistral/voxtral-small-24b-2507.toml
@@ -0,0 +1,29 @@
+# https://mistral.ai/news/voxtral/
+# https://huggingface.co/mistralai/Voxtral-Small-24B-2507
+# https://huggingface.co/mistralai/Voxtral-Small-24B-2507/blob/main/params.json
+# https://docs.mistral.ai/api/endpoint/chat
+# Native params specify 32768 positions, matching the card's 32k context.
+# Output is this shared-window ceiling: prompt + max_tokens must fit; no separate native output cap is published.
+name = "Voxtral Small 24B 2507"
+description = "Open audio-language model for speech transcription, audio understanding, and voice-driven tool use"
+family = "voxtral"
+release_date = "2025-07-15"
+last_updated = "2025-07-15"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text", "audio"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/mistralai/Voxtral-Small-24B-2507"
diff --git a/models/writer/palmyra-x4.toml b/models/writer/palmyra-x4.toml
index 265d3ce5855..2e4fc6728c4 100644
--- a/models/writer/palmyra-x4.toml
+++ b/models/writer/palmyra-x4.toml
@@ -1,7 +1,9 @@
+# https://writer.com/blog/actions-with-palmyra-x4/
+# https://dev.writer.com/home/models
 name = "Palmyra X4"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+description = "Enterprise language model for workflow automation, coding, data analysis, and tool use"
 family = "palmyra"
-release_date = "2025-04-28"
+release_date = "2024-10-09"
 last_updated = "2025-04-28"
 attachment = false
 reasoning = true
@@ -10,8 +12,8 @@ tool_call = true
 open_weights = false
 
 [limit]
-context = 122_880
-output = 8_192
+context = 128_000
+output = 4_096
 
 [modalities]
 input = ["text"]
diff --git a/models/writer/palmyra-x5.toml b/models/writer/palmyra-x5.toml
index 328c9dc5011..502ff35e2ec 100644
--- a/models/writer/palmyra-x5.toml
+++ b/models/writer/palmyra-x5.toml
@@ -1,18 +1,21 @@
+# https://writer.com/blog/long-context-palmyra-x5/
+# https://dev.writer.com/home/models
+# https://dev.writer.com/home/chat-with-images
 name = "Palmyra X5"
 description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
 family = "palmyra"
 release_date = "2025-04-28"
 last_updated = "2025-04-28"
-attachment = false
+attachment = true
 reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
 [limit]
-context = 1_040_000
+context = 1_000_000
 output = 8_192
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
 output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml b/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
index 50089982857..da36a93c809 100644
--- a/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.mistral.pixtral-large-2502-v1:0.toml
@@ -1,3 +1,4 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-pixtral-large.html
 # EU cross-Region inference profile, verified via ListInferenceProfiles from eu-west-1.
 # EU geo pricing (Price List eu-west-1 standard): input/output USD per 1M tokens.
 base_model = "mistral/pixtral-large-2502"
diff --git a/providers/amazon-bedrock/models/meta.llama3-1-70b-instruct-v1:0.toml b/providers/amazon-bedrock/models/meta.llama3-1-70b-instruct-v1:0.toml
index 24e05e58f65..30c8a50cd36 100644
--- a/providers/amazon-bedrock/models/meta.llama3-1-70b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/meta.llama3-1-70b-instruct-v1:0.toml
@@ -1,23 +1,8 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-3-1-70b-instruct.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "meta/llama-3.1-70b-instruct"
 name = "Llama 3.1 70B Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2023-12"
-tool_call = true
-open_weights = true
 
 [cost]
 input = 0.72
 output = 0.72
-
-[limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/meta.llama3-1-8b-instruct-v1:0.toml b/providers/amazon-bedrock/models/meta.llama3-1-8b-instruct-v1:0.toml
index 0ecb525add1..7d5a8aa6ca7 100644
--- a/providers/amazon-bedrock/models/meta.llama3-1-8b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/meta.llama3-1-8b-instruct-v1:0.toml
@@ -1,23 +1,8 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-3-1-8b-instruct.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "meta/llama-3.1-8b-instruct"
 name = "Llama 3.1 8B Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2023-12"
-tool_call = true
-open_weights = true
 
 [cost]
 input = 0.22
 output = 0.22
-
-[limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/meta.llama3-3-70b-instruct-v1:0.toml b/providers/amazon-bedrock/models/meta.llama3-3-70b-instruct-v1:0.toml
index 80ce307a3e5..49b655c547a 100644
--- a/providers/amazon-bedrock/models/meta.llama3-3-70b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/meta.llama3-3-70b-instruct-v1:0.toml
@@ -1,23 +1,8 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-3-3-70b-instruct.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-12-06"
-last_updated = "2024-12-06"
-attachment = false
-reasoning = false
-temperature = true
-knowledge = "2023-12"
-tool_call = true
-open_weights = true
 
 [cost]
 input = 0.72
 output = 0.72
-
-[limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/meta.llama4-maverick-17b-instruct-v1:0.toml b/providers/amazon-bedrock/models/meta.llama4-maverick-17b-instruct-v1:0.toml
index 708c0850c6b..c4c1ab3773d 100644
--- a/providers/amazon-bedrock/models/meta.llama4-maverick-17b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/meta.llama4-maverick-17b-instruct-v1:0.toml
@@ -1,23 +1,11 @@
-name = "Llama 4 Maverick 17B Instruct"
-description = "Open multimodal Llama model for strong reasoning and fast responses"
-family = "llama"
-release_date = "2025-04-05"
-last_updated = "2025-04-05"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-08"
-tool_call = true
-open_weights = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-4-maverick-17b-instruct.html
+# Converse US profile verified 2026-09-09: maxTokens=8192 succeeds; 16384 exceeds the explicit 8192 limit.
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "meta/llama-4-maverick-17b-instruct"
 
 [cost]
 input = 0.24
 output = 0.97
 
 [limit]
-context = 1_000_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/meta.llama4-scout-17b-instruct-v1:0.toml b/providers/amazon-bedrock/models/meta.llama4-scout-17b-instruct-v1:0.toml
index 7f2c62bb743..4dc7fc487db 100644
--- a/providers/amazon-bedrock/models/meta.llama4-scout-17b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/meta.llama4-scout-17b-instruct-v1:0.toml
@@ -1,23 +1,12 @@
-name = "Llama 4 Scout 17B Instruct"
-description = "Open multimodal Llama model for long-context analysis and efficient agents"
-family = "llama"
-release_date = "2025-04-05"
-last_updated = "2025-04-05"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-08"
-tool_call = true
-open_weights = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-4-scout-17b-instruct.html
+# AWS documents the Bedrock context window as 10M tokens; inherit the matching lab context.
+# Converse US profile verified 2026-09-09: maxTokens=8192 succeeds; 16384 exceeds the explicit 8192 limit.
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "meta/llama-4-scout-17b-instruct"
 
 [cost]
 input = 0.17
 output = 0.66
 
 [limit]
-context = 3_500_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/mistral.devstral-2-123b.toml b/providers/amazon-bedrock/models/mistral.devstral-2-123b.toml
index 63e02383c7b..a56c3b4be30 100644
--- a/providers/amazon-bedrock/models/mistral.devstral-2-123b.toml
+++ b/providers/amazon-bedrock/models/mistral.devstral-2-123b.toml
@@ -1,14 +1,8 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-devstral-2-123b.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/devstral-2512"
 name = "Devstral 2 123B"
-description = "Mistral coding agent model for repository tasks and software engineering workflows"
-family = "devstral"
-release_date = "2026-02-17"
-last_updated = "2026-02-17"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.40
@@ -17,7 +11,3 @@ output = 2.00
 [limit]
 context = 256_000
 output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.magistral-small-2509.toml b/providers/amazon-bedrock/models/mistral.magistral-small-2509.toml
index 136c31600fd..8e4e75887b1 100644
--- a/providers/amazon-bedrock/models/mistral.magistral-small-2509.toml
+++ b/providers/amazon-bedrock/models/mistral.magistral-small-2509.toml
@@ -1,15 +1,9 @@
-name = "Magistral Small 1.2"
-description = "Mistral reasoning model for transparent analysis, math, and complex decisions"
-family = "magistral"
-release_date = "2025-12-02"
-last_updated = "2025-12-02"
-attachment = false
-reasoning = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-magistral-small-2509.html
+# https://huggingface.co/mistralai/Magistral-Small-2509 — reasoning via the model's think tokens, no effort/budget field.
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/magistral-small-2509"
 reasoning_options = []
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.50
@@ -18,7 +12,3 @@ output = 1.50
 [limit]
 context = 128_000
 output = 40_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.ministral-3-14b-instruct.toml b/providers/amazon-bedrock/models/mistral.ministral-3-14b-instruct.toml
index c5a71d714e4..bf7801837dc 100644
--- a/providers/amazon-bedrock/models/mistral.ministral-3-14b-instruct.toml
+++ b/providers/amazon-bedrock/models/mistral.ministral-3-14b-instruct.toml
@@ -1,14 +1,8 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-ministral-14b-3-0.html
+# Converse image input verified in us-east-1 on 2026-09-09: a yellow PNG was identified as "Yellow.".
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/ministral-3-14b-instruct-2512"
 name = "Ministral 14B 3.0"
-description = "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads"
-family = "ministral"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
 
 [cost]
 input = 0.20
@@ -17,7 +11,3 @@ output = 0.20
 [limit]
 context = 128_000
 output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.ministral-3-3b-instruct.toml b/providers/amazon-bedrock/models/mistral.ministral-3-3b-instruct.toml
index 8905f5dee8a..09cde0e1d1e 100644
--- a/providers/amazon-bedrock/models/mistral.ministral-3-3b-instruct.toml
+++ b/providers/amazon-bedrock/models/mistral.ministral-3-3b-instruct.toml
@@ -1,14 +1,6 @@
-name = "Ministral 3 3B"
-description = "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads"
-family = "ministral"
-release_date = "2025-12-02"
-last_updated = "2025-12-02"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-ministral-3b.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/ministral-3-3b-instruct-2512"
 
 [cost]
 input = 0.10
@@ -17,7 +9,3 @@ output = 0.10
 [limit]
 context = 256_000
 output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.ministral-3-8b-instruct.toml b/providers/amazon-bedrock/models/mistral.ministral-3-8b-instruct.toml
index e0e28477ebf..3358e8edd66 100644
--- a/providers/amazon-bedrock/models/mistral.ministral-3-8b-instruct.toml
+++ b/providers/amazon-bedrock/models/mistral.ministral-3-8b-instruct.toml
@@ -1,14 +1,7 @@
-name = "Ministral 3 8B"
-description = "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads"
-family = "ministral"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-ministral-3-8b.html
+# Converse image input verified in us-east-1 on 2026-09-09: a blue PNG was identified as "Blue".
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/ministral-3-8b-instruct-2512"
 
 [cost]
 input = 0.15
@@ -17,7 +10,3 @@ output = 0.15
 [limit]
 context = 128_000
 output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.mistral-large-3-675b-instruct.toml b/providers/amazon-bedrock/models/mistral.mistral-large-3-675b-instruct.toml
index e48e415d27f..ca133c0ad46 100644
--- a/providers/amazon-bedrock/models/mistral.mistral-large-3-675b-instruct.toml
+++ b/providers/amazon-bedrock/models/mistral.mistral-large-3-675b-instruct.toml
@@ -1,14 +1,7 @@
-name = "Mistral Large 3"
-description = "Flagship Mistral model for advanced reasoning, coding, and multilingual work"
-family = "mistral"
-release_date = "2025-12-02"
-last_updated = "2025-12-02"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-mistral-large-3.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/mistral-large-2512"
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.50
@@ -17,7 +10,3 @@ output = 1.50
 [limit]
 context = 256_000
 output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.pixtral-large-2502-v1:0.toml b/providers/amazon-bedrock/models/mistral.pixtral-large-2502-v1:0.toml
index 18325d7d856..20ae752c7b9 100644
--- a/providers/amazon-bedrock/models/mistral.pixtral-large-2502-v1:0.toml
+++ b/providers/amazon-bedrock/models/mistral.pixtral-large-2502-v1:0.toml
@@ -1,22 +1,7 @@
-name = "Pixtral Large (25.02)"
-description = "Mistral vision-language model for image understanding and multimodal chat"
-family = "mistral"
-release_date = "2025-04-08"
-last_updated = "2025-04-08"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-pixtral-large.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/pixtral-large-2502"
 
 [cost]
 input = 2.00
 output = 6.00
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.voxtral-mini-3b-2507.toml b/providers/amazon-bedrock/models/mistral.voxtral-mini-3b-2507.toml
index db22b79733b..e6c47fef162 100644
--- a/providers/amazon-bedrock/models/mistral.voxtral-mini-3b-2507.toml
+++ b/providers/amazon-bedrock/models/mistral.voxtral-mini-3b-2507.toml
@@ -1,23 +1,11 @@
-name = "Voxtral Mini 3B 2507"
-description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "mistral"
-release_date = "2024-12-01"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-voxtral-mini-3b-2507.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/voxtral-mini-3b-2507"
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.04
 output = 0.04
 
 [limit]
-context = 128_000
 output = 4_096
-
-[modalities]
-input = ["audio", "text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/mistral.voxtral-small-24b-2507.toml b/providers/amazon-bedrock/models/mistral.voxtral-small-24b-2507.toml
index aecea5edbfb..f1ac2ee6791 100644
--- a/providers/amazon-bedrock/models/mistral.voxtral-small-24b-2507.toml
+++ b/providers/amazon-bedrock/models/mistral.voxtral-small-24b-2507.toml
@@ -1,23 +1,11 @@
-name = "Voxtral Small 24B 2507"
-description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "mistral"
-release_date = "2025-07-01"
-last_updated = "2025-07-01"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-voxtral-small-24b-2507.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+base_model = "mistral/voxtral-small-24b-2507"
 structured_output = true
-open_weights = true
 
 [cost]
-input = 0.15
-output = 0.35
+input = 0.10
+output = 0.30
 
 [limit]
-context = 32_000
-output = 8192
-
-[modalities]
-input = ["text", "audio"]
-output = ["text"]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
index c0cdccb6f37..c6596a8c72d 100644
--- a/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.meta.llama3-3-70b-instruct-v1:0.toml
@@ -2,7 +2,6 @@
 # US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct (US)"
-attachment = false
 
 [cost]
 input = 0.72
diff --git a/providers/amazon-bedrock/models/us.meta.llama4-maverick-17b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama4-maverick-17b-instruct-v1:0.toml
index 5030ef4621b..f0d7497b272 100644
--- a/providers/amazon-bedrock/models/us.meta.llama4-maverick-17b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.meta.llama4-maverick-17b-instruct-v1:0.toml
@@ -1,6 +1,12 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-4-maverick-17b-instruct.html
+# Converse verified 2026-09-09: maxTokens=8192 succeeds; 16384 exceeds the explicit 8192 limit.
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
 base_model = "meta/llama-4-maverick-17b-instruct"
 name = "Llama 4 Maverick 17B Instruct (US)"
 
 [cost]
 input = 0.24
 output = 0.97
+
+[limit]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/us.meta.llama4-scout-17b-instruct-v1:0.toml b/providers/amazon-bedrock/models/us.meta.llama4-scout-17b-instruct-v1:0.toml
index b0633670585..583f68f426b 100644
--- a/providers/amazon-bedrock/models/us.meta.llama4-scout-17b-instruct-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.meta.llama4-scout-17b-instruct-v1:0.toml
@@ -1,6 +1,13 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-meta-llama-4-scout-17b-instruct.html
+# AWS documents the Bedrock context window as 10M tokens; inherit the matching lab context.
+# Converse verified 2026-09-09: maxTokens=8192 succeeds; 16384 exceeds the explicit 8192 limit.
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
 base_model = "meta/llama-4-scout-17b-instruct"
 name = "Llama 4 Scout 17B Instruct (US)"
 
 [cost]
 input = 0.17
 output = 0.66
+
+[limit]
+output = 8_192
diff --git a/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml b/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
index d0e81bd25a7..15903157777 100644
--- a/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.mistral.pixtral-large-2502-v1:0.toml
@@ -1,3 +1,4 @@
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-mistral-ai-pixtral-large.html
 # US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
 # US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
 base_model = "mistral/pixtral-large-2502"
diff --git a/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml b/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
index f2b5bfb7981..98c43dce244 100644
--- a/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.writer.palmyra-x4-v1:0.toml
@@ -1,5 +1,6 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
-# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-palmyra-x4.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+# Bedrock's documented request has no caller-selectable reasoning control.
 base_model = "writer/palmyra-x4"
 reasoning_options = []
 name = "Palmyra X4 (US)"
@@ -7,3 +8,7 @@ name = "Palmyra X4 (US)"
 [cost]
 input = 2.50
 output = 10.00
+
+[limit]
+context = 122_880
+output = 8_192
diff --git a/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml b/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml
index b56d9601c9f..0be9e353cb1 100644
--- a/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.writer.palmyra-x5-v1:0.toml
@@ -1,9 +1,19 @@
-# US cross-Region inference profile, verified via ListInferenceProfiles from us-east-1.
-# US geo pricing (Price List us-east-1 standard): input/output USD per 1M tokens.
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-palmyra-x5.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-writer-palmyra-x5.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+# Bedrock accepts text only and documents no caller-selectable reasoning control.
 base_model = "writer/palmyra-x5"
+attachment = false
 reasoning_options = []
 name = "Palmyra X5 (US)"
 
 [cost]
 input = 0.60
 output = 6.00
+
+[limit]
+context = 1_040_000
+input = 1_040_000
+
+[modalities]
+input = ["text"]
diff --git a/providers/amazon-bedrock/models/writer.palmyra-x4-v1:0.toml b/providers/amazon-bedrock/models/writer.palmyra-x4-v1:0.toml
index d532f6fab3f..faf256a92d1 100644
--- a/providers/amazon-bedrock/models/writer.palmyra-x4-v1:0.toml
+++ b/providers/amazon-bedrock/models/writer.palmyra-x4-v1:0.toml
@@ -1,14 +1,8 @@
-name = "Palmyra X4"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "palmyra"
-release_date = "2025-04-28"
-last_updated = "2025-04-28"
-attachment = false
-reasoning = true
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-palmyra-x4.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+# Bedrock's documented request has no caller-selectable reasoning control.
+base_model = "writer/palmyra-x4"
 reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
 
 [cost]
 input = 2.5
@@ -17,7 +11,3 @@ output = 10
 [limit]
 context = 122_880
 output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/writer.palmyra-x5-v1:0.toml b/providers/amazon-bedrock/models/writer.palmyra-x5-v1:0.toml
index db7bae7cbcc..51974afff22 100644
--- a/providers/amazon-bedrock/models/writer.palmyra-x5-v1:0.toml
+++ b/providers/amazon-bedrock/models/writer.palmyra-x5-v1:0.toml
@@ -1,14 +1,10 @@
-name = "Palmyra X5"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-family = "palmyra"
-release_date = "2025-04-28"
-last_updated = "2025-04-28"
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-palmyra-x5.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-writer-palmyra-x5.html
+# https://aws.amazon.com/bedrock/pricing/ — us-east-1 standard, USD per 1M tokens.
+# Bedrock accepts text only and documents no caller-selectable reasoning control.
+base_model = "writer/palmyra-x5"
 attachment = false
-reasoning = true
 reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
 
 [cost]
 input = 0.6
@@ -16,8 +12,7 @@ output = 6
 
 [limit]
 context = 1_040_000
-output = 8_192
+input = 1_040_000
 
 [modalities]
 input = ["text"]
-output = ["text"]

From ee7eadfce305015f4b323e0d7a3db7201fae8e33 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:15 +0000
Subject: [PATCH 1161/2012] chore(sync): update Kilo model catalog (#6933)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta-llama/llama-3.3-70b-instruct.toml    |  1 -
 .../models/mistralai/voxtral-small-24b-2507.toml     | 12 +-----------
 providers/kilo/models/~moonshotai/kimi-latest.toml   |  6 +++---
 3 files changed, 4 insertions(+), 15 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
index 878106e6be1..6ec2fdc3aa4 100644
--- a/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-attachment = false
 structured_output = true
 
 [cost]
diff --git a/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml b/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
index e73d3ea91b9..fec9c4575b2 100644
--- a/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
+++ b/providers/kilo/models/mistralai/voxtral-small-24b-2507.toml
@@ -1,14 +1,6 @@
-name = "Mistral: Voxtral Small 24B 2507"
+base_model = "mistral/voxtral-small-24b-2507"
 description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "mistral"
-release_date = "2025-10-30"
-last_updated = "2025-10-30"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [cost]
 input = 0.1
@@ -16,9 +8,7 @@ output = 0.3
 cache_read = 0.01
 
 [limit]
-context = 32_768
 output = 26_214
 
 [modalities]
 input = ["text", "audio", "pdf"]
-output = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 77de40dc16d..b5d891762ad 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.535153
-output = 7.70013
-cache_read = 0.175446
+input = 2.302729
+output = 11.550195
+cache_read = 0.263169
 
 [limit]
 context = 1_048_576

From 9e88a33c3a0e03da884caca2e9c1cb68d8045a60 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:18 +0000
Subject: [PATCH 1162/2012] chore(sync): update Weights & Biases model catalog
 (#6932)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/meta-llama/Llama-3.3-70B-Instruct.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/wandb/models/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/wandb/models/meta-llama/Llama-3.3-70B-Instruct.toml
index 4ff37a1bc2c..bbef1cd7465 100644
--- a/providers/wandb/models/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/wandb/models/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -3,7 +3,6 @@ name = "Llama 3.3 70B"
 description = "Multilingual model excelling in conversational tasks, detailed instruction-following, and coding."
 release_date = "2024-12-01"
 last_updated = "2024-12-01"
-attachment = false
 structured_output = true
 
 [cost]

From 661609c74f6d1df57a4387462b67c6925f29f431 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:19 +0000
Subject: [PATCH 1163/2012] chore(sync): update Tinfoil model catalog (#6935)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/tinfoil/models/deepseek-v4-flash.toml b/providers/tinfoil/models/deepseek-v4-flash.toml
index b2b050f3391..c36b1fdc555 100644
--- a/providers/tinfoil/models/deepseek-v4-flash.toml
+++ b/providers/tinfoil/models/deepseek-v4-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.3
-output = 0.7
-cache_read = 0.06
+input = 0.65
+output = 1.45
+cache_read = 0.13
 
 [limit]
 context = 1_048_576

From 78912a3628f47042b2235878d36dc96e26633927 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:21 +0000
Subject: [PATCH 1164/2012] chore(sync): update Charm Hyper model catalog
 (#6929)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/llama-3.3-70b-instruct.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/hyper/models/llama-3.3-70b-instruct.toml b/providers/hyper/models/llama-3.3-70b-instruct.toml
index bb98c102871..96943440fab 100644
--- a/providers/hyper/models/llama-3.3-70b-instruct.toml
+++ b/providers/hyper/models/llama-3.3-70b-instruct.toml
@@ -1,7 +1,6 @@
 base_model = "meta/llama-3.3-70b-instruct"
 release_date = "2026-04-30"
 last_updated = "2026-07-22"
-attachment = false
 
 [cost]
 input = 0.6066

From d5a790a674e3def7db3ba7f2aa4308c4c097f71e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:23 +0000
Subject: [PATCH 1165/2012] chore(sync): update Eden AI model catalog (#6930)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../amazon/mistral.voxtral-mini-3b-2507.toml      | 11 +++++++++++
 .../amazon/mistral.voxtral-mini-3b-2507@us.toml   | 11 +++++++++++
 .../amazon/mistral.voxtral-small-24b-2507.toml    | 11 +++++++++++
 .../amazon/mistral.voxtral-small-24b-2507@us.toml | 11 +++++++++++
 .../meta-llama/Llama-3.3-70B-Instruct.toml        |  1 -
 .../mistralai/Ministral-3-14B-Instruct-2512.toml  | 15 +++++++++++++++
 .../ionos/meta-llama/Llama-3.3-70B-Instruct.toml  |  1 -
 .../nebius/meta-llama/Llama-3.3-70B-Instruct.toml |  1 -
 .../models/scaleway/llama-3.3-70b-instruct.toml   |  1 -
 9 files changed, 59 insertions(+), 4 deletions(-)
 create mode 100644 providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507.toml
 create mode 100644 providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507@us.toml
 create mode 100644 providers/edenai/models/amazon/mistral.voxtral-small-24b-2507.toml
 create mode 100644 providers/edenai/models/amazon/mistral.voxtral-small-24b-2507@us.toml
 create mode 100644 providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml

diff --git a/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507.toml b/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507.toml
new file mode 100644
index 00000000000..ac2969a8bd4
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/voxtral-mini-3b-2507"
+name = "Voxtral Mini 3B 2507 (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.04
+output = 0.04
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507@us.toml b/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507@us.toml
new file mode 100644
index 00000000000..2fcf5b2bbae
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.voxtral-mini-3b-2507@us.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/voxtral-mini-3b-2507"
+name = "Voxtral Mini 3B 2507 (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.04
+output = 0.04
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507.toml b/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507.toml
new file mode 100644
index 00000000000..99068522d25
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/voxtral-small-24b-2507"
+name = "Voxtral Small 24B 2507 (Amazon Bedrock)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.1
+output = 0.3
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507@us.toml b/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507@us.toml
new file mode 100644
index 00000000000..df20d10ce8e
--- /dev/null
+++ b/providers/edenai/models/amazon/mistral.voxtral-small-24b-2507@us.toml
@@ -0,0 +1,11 @@
+base_model = "mistral/voxtral-small-24b-2507"
+name = "Voxtral Small 24B 2507 (Amazon Bedrock, US)"
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.1
+output = 0.3
+
+[limit]
+context = 128_000
diff --git a/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
index b55af51c0bb..bf88f7ab5c4 100644
--- a/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/deepinfra/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama-3.3-70B-Instruct (Deep Infra)"
-attachment = false
 structured_output = true
 
 [cost]
diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
new file mode 100644
index 00000000000..bd3c18bf6a6
--- /dev/null
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -0,0 +1,15 @@
+base_model = "mistral/ministral-3-14b-instruct-2512"
+name = "Ministral 3 14B (Infomaniak)"
+attachment = false
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.34776
+output = 0.46368
+
+[limit]
+context = 100_000
+
+[modalities]
+input = ["text"]
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 1930d4b5b28..6c331de7ca0 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama-3.3-70B-Instruct (IONOS)"
-attachment = false
 tool_call = false
 structured_output = false
 
diff --git a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
index 9485ca66aaf..808d6b6ba5f 100644
--- a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama-3.3-70B-Instruct (Nebius)"
-attachment = false
 structured_output = true
 
 [cost]
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 078fcc3eabf..efaf5da1374 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama-3.3-70B-Instruct (Scaleway)"
-attachment = false
 structured_output = false
 
 [cost]

From fa2733d5fac735e0731e8ce81aaaa4946f70bffc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:26 +0000
Subject: [PATCH 1166/2012] chore(sync): update Merge Gateway model catalog
 (#6931)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/meta/llama-3.3-70b-instruct.toml | 2 --
 1 file changed, 2 deletions(-)

diff --git a/providers/merge-gateway/models/meta/llama-3.3-70b-instruct.toml b/providers/merge-gateway/models/meta/llama-3.3-70b-instruct.toml
index 424d70a90d0..34420f94fab 100644
--- a/providers/merge-gateway/models/meta/llama-3.3-70b-instruct.toml
+++ b/providers/merge-gateway/models/meta/llama-3.3-70b-instruct.toml
@@ -1,8 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=meta%2Fllama-3.3-70b-instruct (accessed 2026-07-21)
-
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct"
-attachment = false
 structured_output = false
 
 [cost]

From 43911c4cef2cc2072eae71b87b33ccc6ee0c6f4e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:42 +0000
Subject: [PATCH 1167/2012] chore(sync): update LLM Gateway model catalog
 (#6937)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/cerebras/llama-3.3-70b-instruct.toml                  | 1 -
 .../models/novita/llama-3.3-70b-instruct.toml                    | 1 -
 2 files changed, 2 deletions(-)

diff --git a/providers/llmgateway-providers/models/cerebras/llama-3.3-70b-instruct.toml b/providers/llmgateway-providers/models/cerebras/llama-3.3-70b-instruct.toml
index 34696489bf7..c333bbf720c 100644
--- a/providers/llmgateway-providers/models/cerebras/llama-3.3-70b-instruct.toml
+++ b/providers/llmgateway-providers/models/cerebras/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct (Cerebras)"
-attachment = false
 structured_output = false
 
 [cost]
diff --git a/providers/llmgateway-providers/models/novita/llama-3.3-70b-instruct.toml b/providers/llmgateway-providers/models/novita/llama-3.3-70b-instruct.toml
index 3f6dcf12ade..fdf0390b4cf 100644
--- a/providers/llmgateway-providers/models/novita/llama-3.3-70b-instruct.toml
+++ b/providers/llmgateway-providers/models/novita/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct (NovitaAI)"
-attachment = false
 structured_output = false
 
 [cost]

From 19ff4120a5d9a04a869f122475ef3e0976c85821 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:31:44 +0000
Subject: [PATCH 1168/2012] chore(sync): update Cortecs model catalog (#6938)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/llama-3.3-70b-instruct.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/cortecs/models/llama-3.3-70b-instruct.toml b/providers/cortecs/models/llama-3.3-70b-instruct.toml
index cb054b806d6..4b2b5f7f4d9 100644
--- a/providers/cortecs/models/llama-3.3-70b-instruct.toml
+++ b/providers/cortecs/models/llama-3.3-70b-instruct.toml
@@ -1,5 +1,4 @@
 base_model = "meta/llama-3.3-70b-instruct"
-attachment = false
 structured_output = true
 
 [cost]

From e9dd91860fba73653832b43e70be50308b79c9c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:32:03 +0000
Subject: [PATCH 1169/2012] chore(sync): update OpenRouter model catalog
 (#6936)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/meta-llama/llama-3.3-70b-instruct.toml    |  1 -
 .../models/mistralai/voxtral-small-24b-2507.toml     | 12 +-----------
 providers/openrouter/models/moonshotai/kimi-k3.toml  |  6 +++---
 .../openrouter/models/~moonshotai/kimi-latest.toml   |  6 +++---
 4 files changed, 7 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
index 5080de4489c..7440afe7d67 100644
--- a/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.3-70b-instruct.toml
@@ -1,5 +1,4 @@
 base_model = "meta/llama-3.3-70b-instruct"
-attachment = false
 structured_output = true
 
 [cost]
diff --git a/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml b/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
index a6827d692ac..fec9c4575b2 100644
--- a/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
+++ b/providers/openrouter/models/mistralai/voxtral-small-24b-2507.toml
@@ -1,14 +1,6 @@
-name = "Voxtral Small 24B 2507"
+base_model = "mistral/voxtral-small-24b-2507"
 description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "mistral"
-release_date = "2025-10-30"
-last_updated = "2025-10-30"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = true
 
 [cost]
 input = 0.1
@@ -16,9 +8,7 @@ output = 0.3
 cache_read = 0.01
 
 [limit]
-context = 32_768
 output = 26_214
 
 [modalities]
 input = ["text", "audio", "pdf"]
-output = ["text"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index d013f6a38a2..22846abd378 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.535153
-output = 7.70013
-cache_read = 0.175446
+input = 2.302729
+output = 11.550195
+cache_read = 0.263169
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 2111cf6b352..6d2db7673f9 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.535153
-output = 7.70013
-cache_read = 0.175446
+input = 2.302729
+output = 11.550195
+cache_read = 0.263169
 
 [limit]
 context = 1_048_576

From a0b4d1bfad87f0bec94b624fb8a0323df4076fe1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 01:32:04 +0000
Subject: [PATCH 1170/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#6934)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/@cf/meta/llama-3.3-70b-instruct-fp8-fast.toml         | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/meta/llama-3.3-70b-instruct-fp8-fast.toml b/providers/cloudflare-workers-ai/models/@cf/meta/llama-3.3-70b-instruct-fp8-fast.toml
index d12dc71b7f6..74c200b2614 100644
--- a/providers/cloudflare-workers-ai/models/@cf/meta/llama-3.3-70b-instruct-fp8-fast.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/meta/llama-3.3-70b-instruct-fp8-fast.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct fp8 Fast"
-attachment = false
 structured_output = false
 
 [cost]

From 0102e86758cb3cc24fe1e7b4fa03ac0c333dcaab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 02:33:35 +0000
Subject: [PATCH 1171/2012] chore(sync): update OpenRouter model catalog
 (#6939)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../inference-net/schematron-v2-small.toml    | 22 +++++++++++++++++++
 .../inference-net/schematron-v2-turbo.toml    | 22 +++++++++++++++++++
 3 files changed, 47 insertions(+), 3 deletions(-)
 create mode 100644 providers/openrouter/models/inference-net/schematron-v2-small.toml
 create mode 100644 providers/openrouter/models/inference-net/schematron-v2-turbo.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index cb7e272671c..1a10a9ce83c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06734
-output = 0.13468
-cache_read = 0.013468
+input = 0.06706
+output = 0.13412
+cache_read = 0.013412
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inference-net/schematron-v2-small.toml b/providers/openrouter/models/inference-net/schematron-v2-small.toml
new file mode 100644
index 00000000000..90109228329
--- /dev/null
+++ b/providers/openrouter/models/inference-net/schematron-v2-small.toml
@@ -0,0 +1,22 @@
+name = "Schematron V2 Small"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.05
+output = 0.23
+
+[limit]
+context = 128_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/inference-net/schematron-v2-turbo.toml b/providers/openrouter/models/inference-net/schematron-v2-turbo.toml
new file mode 100644
index 00000000000..1ecb2bb230d
--- /dev/null
+++ b/providers/openrouter/models/inference-net/schematron-v2-turbo.toml
@@ -0,0 +1,22 @@
+name = "Schematron V2 Turbo"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.03
+output = 0.15
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From bf0e61334df4b19bdb369f9ebe8d15181921c5e0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 02:34:01 +0000
Subject: [PATCH 1172/2012] chore(sync): update Kilo model catalog (#6940)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inference-net/schematron-v2-small.toml    | 22 +++++++++++++++++++
 .../inference-net/schematron-v2-turbo.toml    | 22 +++++++++++++++++++
 providers/kilo/models/tencent/hy3.toml        |  6 ++---
 3 files changed, 47 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/inference-net/schematron-v2-small.toml
 create mode 100644 providers/kilo/models/inference-net/schematron-v2-turbo.toml

diff --git a/providers/kilo/models/inference-net/schematron-v2-small.toml b/providers/kilo/models/inference-net/schematron-v2-small.toml
new file mode 100644
index 00000000000..b3551182e5f
--- /dev/null
+++ b/providers/kilo/models/inference-net/schematron-v2-small.toml
@@ -0,0 +1,22 @@
+name = "Inference.net: Schematron V2 Small"
+description = "Schematron V2 Small is a 3B-parameter HTML-to-JSON extraction model from Inference.net. It prioritizes extraction quality for complex schemas and long pages. Extraction instructions must be supplied through a JSON schema..."
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.05
+output = 0.23
+
+[limit]
+context = 128_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/inference-net/schematron-v2-turbo.toml b/providers/kilo/models/inference-net/schematron-v2-turbo.toml
new file mode 100644
index 00000000000..d75abf2715a
--- /dev/null
+++ b/providers/kilo/models/inference-net/schematron-v2-turbo.toml
@@ -0,0 +1,22 @@
+name = "Inference.net: Schematron V2 Turbo"
+description = "Schematron V2 Turbo is a 3B-parameter HTML-to-JSON extraction model from Inference.net. It prioritizes throughput for high-volume extraction workloads. Extraction instructions must be supplied through a JSON schema in response_format rather..."
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.03
+output = 0.15
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144

From 2580a6f9cfee753dd08ecc460b6423c6469d4c9c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 03:29:04 +0000
Subject: [PATCH 1173/2012] chore(sync): update OpenRouter model catalog
 (#6942)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1b4ffd8a683..f35c0a05aa4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.849816
-output = 1.699632
-cache_read = 0.070818
+input = 0.839724
+output = 1.679448
+cache_read = 0.069977
 
 [limit]
 context = 1_048_576

From 2947c80eb1677f72d869b18cf0aa141bc897527e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 04:28:11 +0000
Subject: [PATCH 1174/2012] chore(sync): update Tinfoil model catalog (#6944)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/tinfoil/models/deepseek-v4-flash.toml b/providers/tinfoil/models/deepseek-v4-flash.toml
index c36b1fdc555..b2b050f3391 100644
--- a/providers/tinfoil/models/deepseek-v4-flash.toml
+++ b/providers/tinfoil/models/deepseek-v4-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.65
-output = 1.45
-cache_read = 0.13
+input = 0.3
+output = 0.7
+cache_read = 0.06
 
 [limit]
 context = 1_048_576

From 801652200106eb4612829f93d6ab220ec19a2166 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 04:28:21 +0000
Subject: [PATCH 1175/2012] chore(sync): update OpenRouter model catalog
 (#6943)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a10a9ce83c..867ad009e1f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06706
-output = 0.13412
-cache_read = 0.013412
+input = 0.06678
+output = 0.13356
+cache_read = 0.013356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f35c0a05aa4..501bc3d086f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.839724
-output = 1.679448
-cache_read = 0.069977
+input = 0.829458
+output = 1.658916
+cache_read = 0.069122
 
 [limit]
 context = 1_048_576

From 1714715c174f4d93e2030926f9d5007879834866 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 05:24:49 +0000
Subject: [PATCH 1176/2012] chore(sync): update OpenRouter model catalog
 (#6946)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 6 +++---
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index d06dd2cd8d1..8dce0b17a1e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.016
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 867ad009e1f..1a10a9ce83c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06678
-output = 0.13356
-cache_read = 0.013356
+input = 0.06706
+output = 0.13412
+cache_read = 0.013412
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 27d5c88cb9e..a8bf0f0a408 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 11d35e9f1e7..14b0cdd0d13 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8727
-output = 3.36
-cache_read = 0.1639
+input = 0.85
+output = 2.8985
+cache_read = 0.17
 
 [limit]
 context = 1_310_720

From a7ac5919fb1166f45aa7a8e0f41214ea0d2c07e6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 05:24:51 +0000
Subject: [PATCH 1177/2012] chore(sync): update Kilo model catalog (#6945)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml               | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 80dc12b4ce8..5b1c7e5857f 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.16
-cache_read = 0.013
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2829ed5b448..d603d7bc671 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8727
-output = 3.36
-cache_read = 0.1639
+input = 0.85
+output = 2.8985
+cache_read = 0.17
 
 [limit]
 context = 1_048_576

From 8879c9e5d72fd59069e01ba3382a4bc006ede20d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 05:25:05 +0000
Subject: [PATCH 1178/2012] chore(sync): update Eden AI model catalog (#6947)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366

From 2184465ef73ebd3553c48fb2f60c130330f14bb6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 06:36:31 +0000
Subject: [PATCH 1179/2012] chore(sync): update OpenRouter model catalog
 (#6948)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a10a9ce83c..867ad009e1f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06706
-output = 0.13412
-cache_read = 0.013412
+input = 0.06678
+output = 0.13356
+cache_read = 0.013356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 501bc3d086f..cc4bcd1c726 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.829458
-output = 1.658916
-cache_read = 0.069122
+input = 0.819366
+output = 1.638732
+cache_read = 0.068281
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 14b0cdd0d13..11d35e9f1e7 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.85
-output = 2.8985
-cache_read = 0.17
+input = 0.8727
+output = 3.36
+cache_read = 0.1639
 
 [limit]
 context = 1_310_720

From 306accd0016ce01a7fe3126d10af371820af4956 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 06:36:34 +0000
Subject: [PATCH 1180/2012] chore(sync): update Kilo model catalog (#6949)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index d603d7bc671..2829ed5b448 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.85
-output = 2.8985
-cache_read = 0.17
+input = 0.8727
+output = 3.36
+cache_read = 0.1639
 
 [limit]
 context = 1_048_576

From 373f7a90cc4c0494233d7c17767b284ac07de1ee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 07:25:12 +0000
Subject: [PATCH 1181/2012] chore(sync): update OpenRouter model catalog
 (#6950)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cc4bcd1c726..a21494f5a5a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.819366
-output = 1.638732
-cache_read = 0.068281
+input = 0.809274
+output = 1.618548
+cache_read = 0.06744
 
 [limit]
 context = 1_048_576

From 648e9b0997f711478d0fa79addde0a20803bfaae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 08:29:07 +0000
Subject: [PATCH 1182/2012] chore(sync): update Kilo model catalog (#6952)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml    | 2 +-
 providers/kilo/models/thinkingmachines/inkling.toml       | 3 ++-
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 providers/kilo/models/~moonshotai/kimi-latest.toml        | 6 +++---
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 75e607a6e02..3e83e7378b1 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/kilo/models/thinkingmachines/inkling.toml b/providers/kilo/models/thinkingmachines/inkling.toml
index 79cba59e119..a367cecf27a 100644
--- a/providers/kilo/models/thinkingmachines/inkling.toml
+++ b/providers/kilo/models/thinkingmachines/inkling.toml
@@ -12,4 +12,5 @@ output = 4.05
 cache_read = 0.16
 
 [limit]
-output = 32_768
+context = 524_288
+output = 471_859
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 5b1c7e5857f..ddced169212 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.08
-cache_read = 0.008
+input = 0.03
+output = 0.07
+cache_read = 0.005
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index b5d891762ad..aa14366b64c 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.302729
-output = 11.550195
-cache_read = 0.263169
+input = 2.125
+output = 11.9
+cache_read = 0.2465
 
 [limit]
 context = 1_048_576

From 5a909191a87d252763d74b14d3897d4136a2a5fd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 08:29:11 +0000
Subject: [PATCH 1183/2012] chore(sync): update OpenRouter model catalog
 (#6951)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/moonshotai/kimi-k2-thinking.toml    | 3 +--
 providers/openrouter/models/thinkingmachines/inkling.toml | 2 +-
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 5 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a21494f5a5a..263d2ca4a64 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.809274
-output = 1.618548
-cache_read = 0.06744
+input = 0.799182
+output = 1.598364
+cache_read = 0.066599
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 91ad22ed77f..3ccd5a4f3ef 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,7 +8,6 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
-cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 235_929
diff --git a/providers/openrouter/models/thinkingmachines/inkling.toml b/providers/openrouter/models/thinkingmachines/inkling.toml
index bbed1c7e0bc..1a64ecd26f3 100644
--- a/providers/openrouter/models/thinkingmachines/inkling.toml
+++ b/providers/openrouter/models/thinkingmachines/inkling.toml
@@ -12,4 +12,4 @@ output = 4.05
 cache_read = 0.17
 
 [limit]
-output = 32_768
+output = 471_859
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index a8bf0f0a408..8500749b841 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.08
-cache_read = 0.008
+input = 0.03
+output = 0.07
+cache_read = 0.005
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 6d2db7673f9..2e898adea4e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.302729
-output = 11.550195
-cache_read = 0.263169
+input = 2.125
+output = 11.9
+cache_read = 0.2465
 
 [limit]
 context = 1_048_576

From dbfc7a5ba5b46f3fad02ff799f0f3f9c159dab67 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 09:25:19 +0000
Subject: [PATCH 1184/2012] chore(sync): update OpenRouter model catalog
 (#6954)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml        | 6 +++---
 .../openrouter/models/~anthropic/claude-haiku-latest.toml  | 2 +-
 .../openrouter/models/~anthropic/claude-sonnet-latest.toml | 2 +-
 .../models/~deepseek/deepseek-v4-flash-latest.toml         | 2 +-
 .../openrouter/models/~google/gemini-flash-latest.toml     | 2 +-
 providers/openrouter/models/~google/gemini-pro-latest.toml | 2 +-
 providers/openrouter/models/~moonshotai/kimi-latest.toml   | 2 +-
 providers/openrouter/models/~openai/gpt-astra-latest.toml  | 2 +-
 providers/openrouter/models/~openai/gpt-luna-latest.toml   | 2 +-
 providers/openrouter/models/~openai/gpt-mini-latest.toml   | 7 +++++--
 providers/openrouter/models/~openai/gpt-sol-latest.toml    | 2 +-
 providers/openrouter/models/~openai/gpt-terra-latest.toml  | 2 +-
 13 files changed, 21 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 263d2ca4a64..a21494f5a5a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.799182
-output = 1.598364
-cache_read = 0.066599
+input = 0.809274
+output = 1.618548
+cache_read = 0.06744
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index c54d27b9d18..d5918c075a2 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~anthropic/claude-haiku-latest.toml b/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
index 6aae1a4db8d..b943d5f3841 100644
--- a/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
+++ b/providers/openrouter/models/~anthropic/claude-haiku-latest.toml
@@ -1,6 +1,6 @@
 # Toggle: reasoning.enabled = true|false
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Anthropic Claude Haiku Latest"
+name = "Claude Haiku Latest"
 description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
 family = "claude-haiku"
 release_date = "2026-04-27"
diff --git a/providers/openrouter/models/~anthropic/claude-sonnet-latest.toml b/providers/openrouter/models/~anthropic/claude-sonnet-latest.toml
index 18e2ed6e310..b121dec2b42 100644
--- a/providers/openrouter/models/~anthropic/claude-sonnet-latest.toml
+++ b/providers/openrouter/models/~anthropic/claude-sonnet-latest.toml
@@ -1,6 +1,6 @@
 # OpenRouter alias currently redirects to Claude Sonnet 5.
 # https://openrouter.ai/~anthropic/claude-sonnet-latest
-name = "Anthropic Claude Sonnet Latest"
+name = "Claude Sonnet Latest"
 description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
 family = "claude-sonnet"
 release_date = "2026-04-27"
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 8500749b841..1f5af5de2fa 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.07
-cache_read = 0.005
+cache_read = 0.003
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~google/gemini-flash-latest.toml b/providers/openrouter/models/~google/gemini-flash-latest.toml
index b664a0fc44a..c9245734fc4 100644
--- a/providers/openrouter/models/~google/gemini-flash-latest.toml
+++ b/providers/openrouter/models/~google/gemini-flash-latest.toml
@@ -1,6 +1,6 @@
 # OpenRouter alias currently redirects to Gemini 3.5 Flash.
 # https://openrouter.ai/~google/gemini-flash-latest
-name = "Google Gemini Flash Latest"
+name = "Gemini Flash Latest"
 description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
 family = "gemini-flash"
 release_date = "2026-04-27"
diff --git a/providers/openrouter/models/~google/gemini-pro-latest.toml b/providers/openrouter/models/~google/gemini-pro-latest.toml
index b25ad638e3f..ac436847d43 100644
--- a/providers/openrouter/models/~google/gemini-pro-latest.toml
+++ b/providers/openrouter/models/~google/gemini-pro-latest.toml
@@ -1,6 +1,6 @@
 # OpenRouter alias currently redirects to Gemini 3.1 Pro Preview.
 # https://openrouter.ai/~google/gemini-pro-latest
-name = "Google Gemini Pro Latest"
+name = "Gemini Pro Latest"
 description = "Advanced Gemini model for complex reasoning, coding, and multimodal analysis"
 family = "gemini-pro"
 release_date = "2026-04-27"
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 2e898adea4e..2c015355855 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -1,6 +1,6 @@
 # Toggle: reasoning.enabled = true|false
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "MoonshotAI Kimi Latest"
+name = "Kimi Latest"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 family = "kimi"
 release_date = "2026-04-27"
diff --git a/providers/openrouter/models/~openai/gpt-astra-latest.toml b/providers/openrouter/models/~openai/gpt-astra-latest.toml
index c5d810b7e39..9922e3acf6d 100644
--- a/providers/openrouter/models/~openai/gpt-astra-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-astra-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Astra Latest"
+name = "GPT Astra Latest"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt-astra"
 release_date = "2026-09-11"
diff --git a/providers/openrouter/models/~openai/gpt-luna-latest.toml b/providers/openrouter/models/~openai/gpt-luna-latest.toml
index 433ee7fb1fc..e349af31375 100644
--- a/providers/openrouter/models/~openai/gpt-luna-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-luna-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Luna Latest"
+name = "GPT Luna Latest"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt-luna"
 release_date = "2026-09-11"
diff --git a/providers/openrouter/models/~openai/gpt-mini-latest.toml b/providers/openrouter/models/~openai/gpt-mini-latest.toml
index 9b3481fbb5b..3106a5836a3 100644
--- a/providers/openrouter/models/~openai/gpt-mini-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-mini-latest.toml
@@ -1,5 +1,4 @@
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-name = "OpenAI GPT Mini Latest"
+name = "GPT Mini Latest"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
 family = "gpt-mini"
 release_date = "2026-04-27"
@@ -12,6 +11,10 @@ structured_output = true
 knowledge = "2025-08-31"
 open_weights = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
+
 [cost]
 input = 0.75
 output = 4.5
diff --git a/providers/openrouter/models/~openai/gpt-sol-latest.toml b/providers/openrouter/models/~openai/gpt-sol-latest.toml
index 028f2642f96..b2cafae187b 100644
--- a/providers/openrouter/models/~openai/gpt-sol-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-sol-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Sol Latest"
+name = "GPT Sol Latest"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt-sol"
 release_date = "2026-09-11"
diff --git a/providers/openrouter/models/~openai/gpt-terra-latest.toml b/providers/openrouter/models/~openai/gpt-terra-latest.toml
index 5d33c886c13..cf611fa8fc5 100644
--- a/providers/openrouter/models/~openai/gpt-terra-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-terra-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Terra Latest"
+name = "GPT Terra Latest"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
 family = "gpt-terra"
 release_date = "2026-09-11"

From 35d26af2a0df90146c1535a6b51caa8dc291543b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 09:25:31 +0000
Subject: [PATCH 1185/2012] chore(sync): update Kilo model catalog (#6955)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~anthropic/claude-haiku-latest.toml     | 2 +-
 providers/kilo/models/~anthropic/claude-sonnet-latest.toml    | 2 +-
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 4 ++--
 providers/kilo/models/~google/gemini-flash-latest.toml        | 2 +-
 providers/kilo/models/~google/gemini-pro-latest.toml          | 2 +-
 providers/kilo/models/~moonshotai/kimi-latest.toml            | 2 +-
 providers/kilo/models/~openai/gpt-astra-latest.toml           | 2 +-
 providers/kilo/models/~openai/gpt-luna-latest.toml            | 2 +-
 providers/kilo/models/~openai/gpt-mini-latest.toml            | 2 +-
 providers/kilo/models/~openai/gpt-sol-latest.toml             | 2 +-
 providers/kilo/models/~openai/gpt-terra-latest.toml           | 2 +-
 11 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/kilo/models/~anthropic/claude-haiku-latest.toml b/providers/kilo/models/~anthropic/claude-haiku-latest.toml
index 2af11ba5520..5f72f1f9aec 100644
--- a/providers/kilo/models/~anthropic/claude-haiku-latest.toml
+++ b/providers/kilo/models/~anthropic/claude-haiku-latest.toml
@@ -1,5 +1,5 @@
 # Tracks the current Claude Haiku alias target on Kilo (Claude Haiku 4.5).
-name = "Anthropic Claude Haiku Latest"
+name = "Anthropic: Claude Haiku Latest"
 description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
 family = "claude-haiku"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~anthropic/claude-sonnet-latest.toml b/providers/kilo/models/~anthropic/claude-sonnet-latest.toml
index e7ef48d0ce7..363ecffe942 100644
--- a/providers/kilo/models/~anthropic/claude-sonnet-latest.toml
+++ b/providers/kilo/models/~anthropic/claude-sonnet-latest.toml
@@ -1,5 +1,5 @@
 # Tracks the current Claude Sonnet alias target on Kilo (Claude Sonnet 4.6 by pricing).
-name = "Anthropic Claude Sonnet Latest"
+name = "Anthropic: Claude Sonnet Latest"
 description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
 family = "claude-sonnet"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index ddced169212..e8bd485a54c 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -1,4 +1,4 @@
-name = "DeepSeek V4 Flash Latest"
+name = "DeepSeek: DeepSeek V4 Flash Latest"
 description = "This model always redirects to the latest model in the DeepSeek V4 Flash family."
 family = "deepseek"
 release_date = "2026-08-01"
@@ -17,7 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.07
-cache_read = 0.005
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~google/gemini-flash-latest.toml b/providers/kilo/models/~google/gemini-flash-latest.toml
index 9e99c8689bf..2bc9f722462 100644
--- a/providers/kilo/models/~google/gemini-flash-latest.toml
+++ b/providers/kilo/models/~google/gemini-flash-latest.toml
@@ -1,5 +1,5 @@
 # Tracks the current Gemini Flash alias target (Gemini 3.5 Flash).
-name = "Google Gemini Flash Latest"
+name = "Google: Gemini Flash Latest"
 description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
 family = "gemini-flash"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~google/gemini-pro-latest.toml b/providers/kilo/models/~google/gemini-pro-latest.toml
index ff879c579ee..a9a18a7f48e 100644
--- a/providers/kilo/models/~google/gemini-pro-latest.toml
+++ b/providers/kilo/models/~google/gemini-pro-latest.toml
@@ -1,4 +1,4 @@
-name = "Google Gemini Pro Latest"
+name = "Google: Gemini Pro Latest"
 description = "Advanced Gemini model for complex reasoning, coding, and multimodal analysis"
 family = "gemini-pro"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index aa14366b64c..5f704ff4ddc 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -1,4 +1,4 @@
-name = "MoonshotAI Kimi Latest"
+name = "MoonshotAI: Kimi Latest"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 family = "kimi"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~openai/gpt-astra-latest.toml b/providers/kilo/models/~openai/gpt-astra-latest.toml
index cc33c625eda..805c6e510e2 100644
--- a/providers/kilo/models/~openai/gpt-astra-latest.toml
+++ b/providers/kilo/models/~openai/gpt-astra-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Astra Latest ($$$$)"
+name = "OpenAI: GPT Astra Latest ($$$$)"
 description = "This model always redirects to the latest model in the OpenAI GPT Astra family."
 family = "gpt-astra"
 release_date = "2026-09-11"
diff --git a/providers/kilo/models/~openai/gpt-luna-latest.toml b/providers/kilo/models/~openai/gpt-luna-latest.toml
index aa44302cd2e..6f3028a0af2 100644
--- a/providers/kilo/models/~openai/gpt-luna-latest.toml
+++ b/providers/kilo/models/~openai/gpt-luna-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Luna Latest"
+name = "OpenAI: GPT Luna Latest"
 description = "This model always redirects to the latest model in the OpenAI GPT Luna family."
 family = "gpt-luna"
 release_date = "2026-09-11"
diff --git a/providers/kilo/models/~openai/gpt-mini-latest.toml b/providers/kilo/models/~openai/gpt-mini-latest.toml
index 4454abcc892..afe30a4a3c1 100644
--- a/providers/kilo/models/~openai/gpt-mini-latest.toml
+++ b/providers/kilo/models/~openai/gpt-mini-latest.toml
@@ -1,5 +1,5 @@
 # Tracks the current GPT Mini alias target on Kilo (GPT-5.4 Mini).
-name = "OpenAI GPT Mini Latest"
+name = "OpenAI: GPT Mini Latest"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
 family = "gpt-mini"
 release_date = "2026-04-27"
diff --git a/providers/kilo/models/~openai/gpt-sol-latest.toml b/providers/kilo/models/~openai/gpt-sol-latest.toml
index 264602aafc2..1af280f2218 100644
--- a/providers/kilo/models/~openai/gpt-sol-latest.toml
+++ b/providers/kilo/models/~openai/gpt-sol-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Sol Latest"
+name = "OpenAI: GPT Sol Latest"
 description = "This model always redirects to the latest model in the OpenAI GPT Sol family."
 family = "gpt-sol"
 release_date = "2026-09-11"
diff --git a/providers/kilo/models/~openai/gpt-terra-latest.toml b/providers/kilo/models/~openai/gpt-terra-latest.toml
index 58f1f2010c2..b6b5ed44925 100644
--- a/providers/kilo/models/~openai/gpt-terra-latest.toml
+++ b/providers/kilo/models/~openai/gpt-terra-latest.toml
@@ -1,4 +1,4 @@
-name = "OpenAI GPT Terra Latest"
+name = "OpenAI: GPT Terra Latest"
 description = "This model always redirects to the latest model in the OpenAI GPT Terra family."
 family = "gpt-terra"
 release_date = "2026-09-11"

From 9b6715f92bc5a07bd145271b6e5af4d0c0e5632a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 10:23:42 +0000
Subject: [PATCH 1186/2012] chore(sync): update Kilo model catalog (#6957)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/minimax/minimax-m2.5.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/kilo/models/minimax/minimax-m2.5.toml b/providers/kilo/models/minimax/minimax-m2.5.toml
index 3f15c0bc21d..e89b49a7fbe 100644
--- a/providers/kilo/models/minimax/minimax-m2.5.toml
+++ b/providers/kilo/models/minimax/minimax-m2.5.toml
@@ -10,3 +10,7 @@ values = ["high"]
 input = 0.3
 output = 1.2
 cache_read = 0.03
+
+[limit]
+context = 200_000
+output = 128_000

From 682a6b531e1aa7eb0342172213f56738939cdd3a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 10:23:52 +0000
Subject: [PATCH 1187/2012] chore(sync): update OpenRouter model catalog
 (#6956)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 providers/openrouter/models/minimax/minimax-m2.5.toml    | 9 ++++++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a21494f5a5a..263d2ca4a64 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.809274
-output = 1.618548
-cache_read = 0.06744
+input = 0.799182
+output = 1.598364
+cache_read = 0.066599
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/minimax/minimax-m2.5.toml b/providers/openrouter/models/minimax/minimax-m2.5.toml
index 1edeb5ecdfc..ca75f8e4927 100644
--- a/providers/openrouter/models/minimax/minimax-m2.5.toml
+++ b/providers/openrouter/models/minimax/minimax-m2.5.toml
@@ -6,6 +6,9 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.03
+input = 0.27
+output = 1.08
+cache_read = 0.027
+
+[limit]
+output = 128_000

From 3cdf747fe40a6f1b08c9e7cd1065266cc4cf0a88 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 11:21:46 +0000
Subject: [PATCH 1188/2012] chore(sync): update LLM Gateway model catalog
 (#6959)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/kimi-k3.toml                    | 14 ++++++++++++++
 .../models/runpod/kimi-k3.toml                     | 14 ++++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/alibaba/kimi-k3.toml
 create mode 100644 providers/llmgateway-providers/models/runpod/kimi-k3.toml

diff --git a/providers/llmgateway-providers/models/alibaba/kimi-k3.toml b/providers/llmgateway-providers/models/alibaba/kimi-k3.toml
new file mode 100644
index 00000000000..4f21a719d33
--- /dev/null
+++ b/providers/llmgateway-providers/models/alibaba/kimi-k3.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (Alibaba Cloud)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+output = 1_048_576
diff --git a/providers/llmgateway-providers/models/runpod/kimi-k3.toml b/providers/llmgateway-providers/models/runpod/kimi-k3.toml
new file mode 100644
index 00000000000..02de6d91ec6
--- /dev/null
+++ b/providers/llmgateway-providers/models/runpod/kimi-k3.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (Runpod)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+output = 1_048_576

From e7a74d6dc7ce56d1a9dda94c18fcf185248d77b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 11:21:54 +0000
Subject: [PATCH 1189/2012] chore(sync): update OpenRouter model catalog
 (#6960)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 263d2ca4a64..18a8171f70e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.799182
-output = 1.598364
-cache_read = 0.066599
+input = 0.788916
+output = 1.577832
+cache_read = 0.065743
 
 [limit]
 context = 1_048_576

From 18ec22eff3458a9da06e0e679fe6ec75c4a02808 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 12:32:12 +0000
Subject: [PATCH 1190/2012] chore(sync): update LLM Gateway model catalog
 (#6962)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aws-mantle/gpt-6-astra.toml               | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/aws-mantle/gpt-6-astra.toml

diff --git a/providers/llmgateway-providers/models/aws-mantle/gpt-6-astra.toml b/providers/llmgateway-providers/models/aws-mantle/gpt-6-astra.toml
new file mode 100644
index 00000000000..a9a970a8f7e
--- /dev/null
+++ b/providers/llmgateway-providers/models/aws-mantle/gpt-6-astra.toml
@@ -0,0 +1,12 @@
+base_model = "openai/gpt-6-astra"
+name = "GPT-6 Astra (AWS Mantle)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5

From 5914f3403b91064b23443e848fd3503abeca73a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 12:32:19 +0000
Subject: [PATCH 1191/2012] chore(sync): update OpenRouter model catalog
 (#6963)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 18a8171f70e..e418673776b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.788916
-output = 1.577832
-cache_read = 0.065743
+input = 0.778824
+output = 1.557648
+cache_read = 0.064902
 
 [limit]
 context = 1_048_576

From dafb57e823914c5e76e498eade35b77fc72cb396 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 13:23:01 +0000
Subject: [PATCH 1192/2012] chore(sync): update OpenRouter model catalog
 (#6964)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 867ad009e1f..1d5faa4a4b5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06678
-output = 0.13356
-cache_read = 0.013356
+input = 0.0665
+output = 0.133
+cache_read = 0.0133
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e418673776b..60abfac3e72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.778824
-output = 1.557648
-cache_read = 0.064902
+input = 0.768732
+output = 1.537464
+cache_read = 0.064061
 
 [limit]
 context = 1_048_576

From cca149262a59c78f0ca35c6ba66d02e339029d38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 14:24:11 +0000
Subject: [PATCH 1193/2012] chore(sync): update Kilo model catalog (#6968)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000

From 48e83c31cfb19a8263ae038ed4bf57d4eb861b3e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 14:24:14 +0000
Subject: [PATCH 1194/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#6965)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5-turbo.toml  | 10 ++++++++
 providers/llmgateway/models/glm-5v-turbo.toml | 10 ++++++++
 providers/llmgateway/models/hy-mt2-plus.toml  | 23 +++++++++++++++++++
 providers/llmgateway/models/hy3.toml          |  8 +++----
 providers/llmgateway/models/hy4-preview.toml  | 13 +++++++++++
 .../models/kimi-k2.7-code-highspeed.toml      |  2 +-
 6 files changed, 61 insertions(+), 5 deletions(-)
 create mode 100644 providers/llmgateway/models/glm-5-turbo.toml
 create mode 100644 providers/llmgateway/models/glm-5v-turbo.toml
 create mode 100644 providers/llmgateway/models/hy-mt2-plus.toml
 create mode 100644 providers/llmgateway/models/hy4-preview.toml

diff --git a/providers/llmgateway/models/glm-5-turbo.toml b/providers/llmgateway/models/glm-5-turbo.toml
new file mode 100644
index 00000000000..db90b38197b
--- /dev/null
+++ b/providers/llmgateway/models/glm-5-turbo.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5-turbo"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.24
diff --git a/providers/llmgateway/models/glm-5v-turbo.toml b/providers/llmgateway/models/glm-5v-turbo.toml
new file mode 100644
index 00000000000..b8a2465703f
--- /dev/null
+++ b/providers/llmgateway/models/glm-5v-turbo.toml
@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5v-turbo"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.24
diff --git a/providers/llmgateway/models/hy-mt2-plus.toml b/providers/llmgateway/models/hy-mt2-plus.toml
new file mode 100644
index 00000000000..2191653753d
--- /dev/null
+++ b/providers/llmgateway/models/hy-mt2-plus.toml
@@ -0,0 +1,23 @@
+name = "Hy-MT2 Plus"
+description = "Tool-capable chat model for instruction following and agentic application workflows"
+family = "Hy"
+release_date = "2026-05-21"
+last_updated = "2026-05-21"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[cost]
+input = 0.074
+output = 0.295
+
+[limit]
+context = 8_192
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/llmgateway/models/hy3.toml b/providers/llmgateway/models/hy3.toml
index f4ddcb6f714..696f1077b2d 100644
--- a/providers/llmgateway/models/hy3.toml
+++ b/providers/llmgateway/models/hy3.toml
@@ -2,12 +2,12 @@ base_model = "tencent/hy3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/llmgateway/models/hy4-preview.toml b/providers/llmgateway/models/hy4-preview.toml
new file mode 100644
index 00000000000..a9c0cb59404
--- /dev/null
+++ b/providers/llmgateway/models/hy4-preview.toml
@@ -0,0 +1,13 @@
+base_model = "tencent/hy4-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
+
+[limit]
+context = 1_000_000
diff --git a/providers/llmgateway/models/kimi-k2.7-code-highspeed.toml b/providers/llmgateway/models/kimi-k2.7-code-highspeed.toml
index 09a0c8b14ed..6ec637888c2 100644
--- a/providers/llmgateway/models/kimi-k2.7-code-highspeed.toml
+++ b/providers/llmgateway/models/kimi-k2.7-code-highspeed.toml
@@ -8,7 +8,7 @@ field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.9

From b64be284225e0642cfcb0ab23dfd455a173d19d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 14:24:25 +0000
Subject: [PATCH 1195/2012] chore(sync): update OpenRouter model catalog
 (#6967)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 +++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1d5faa4a4b5..0d13f8802e8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0665
-output = 0.133
-cache_read = 0.0133
+input = 0.06622
+output = 0.13244
+cache_read = 0.013244
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c435d69ac6..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.018396
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 60abfac3e72..ec612bbcbdb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.768732
-output = 1.537464
-cache_read = 0.064061
+input = 0.75864
+output = 1.51728
+cache_read = 0.06322
 
 [limit]
 context = 1_048_576

From e3d67b3ff9d92671e6fd64d44f21aefa12f57f80 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 15:22:48 +0000
Subject: [PATCH 1196/2012] chore(sync): update OpenRouter model catalog
 (#6971)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0d13f8802e8..d438ecaa08e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06622
-output = 0.13244
-cache_read = 0.013244
+input = 0.06594
+output = 0.13188
+cache_read = 0.013188
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ec612bbcbdb..d6b763edc78 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.75864
-output = 1.51728
-cache_read = 0.06322
+input = 0.748374
+output = 1.496748
+cache_read = 0.062365
 
 [limit]
 context = 1_048_576

From ae140b7784a04c7a48f176137b201d8119e7ec33 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 15:22:52 +0000
Subject: [PATCH 1197/2012] chore(sync): update Eden AI model catalog (#6972)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424

From 0a8495d0d7a3ce7275d46993892ac8aac468463a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 16:25:40 +0000
Subject: [PATCH 1198/2012] chore(sync): update OpenRouter model catalog
 (#6973)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index d6b763edc78..1530b62b57b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.748374
-output = 1.496748
-cache_read = 0.062365
+input = 0.738282
+output = 1.476564
+cache_read = 0.061524
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 4c9aa9350948d6860022383dbce9ca0b553e6c4c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 17:21:47 +0000
Subject: [PATCH 1199/2012] chore(sync): update OpenRouter model catalog
 (#6974)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 7 ++++---
 providers/openrouter/models/moonshotai/kimi-k3.toml       | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml       | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 4 files changed, 13 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 1530b62b57b..95460b6f5f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.738282
-output = 1.476564
-cache_read = 0.061524
+input = 1.6
+output = 3.2
+cache_read = 0.135
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 22846abd378..3a0c7f47171 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.302729
-output = 11.550195
-cache_read = 0.263169
+input = 2.648138
+output = 13.282724
+cache_read = 0.302644
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index d5918c075a2..c54d27b9d18 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 2c015355855..44bf8744327 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.125
-output = 11.9
-cache_read = 0.2465
+input = 2.375
+output = 13.3
+cache_read = 0.2755
 
 [limit]
 context = 1_048_576

From b03ecbe4c939474fbad7aaec83ad36fe0e7a6b67 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 17:22:32 +0000
Subject: [PATCH 1200/2012] chore(sync): update Kilo model catalog (#6975)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml | 3 ++-
 providers/kilo/models/~moonshotai/kimi-latest.toml  | 6 +++---
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index 88500bd7dd8..e58e24281d4 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,4 +11,5 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 5f704ff4ddc..e0c53b97309 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.125
-output = 11.9
-cache_read = 0.2465
+input = 2.375
+output = 13.3
+cache_read = 0.2755
 
 [limit]
 context = 1_048_576

From ca8bc3fda222c2449af6bca5b4fc8fc55c408cec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 18:27:01 +0000
Subject: [PATCH 1201/2012] chore(sync): update Kilo model catalog (#6977)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 6fc67901197..881a90e0d15 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384

From e4ae2771dcc67d70370fd4c4a4d691718d063575 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 18:27:24 +0000
Subject: [PATCH 1202/2012] chore(sync): update OpenRouter model catalog
 (#6976)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml           | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d438ecaa08e..1913a52d832 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06594
-output = 0.13188
-cache_read = 0.013188
+input = 0.06566
+output = 0.13132
+cache_read = 0.013132
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 1352c1ee599..881a90e0d15 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.72
-output = 0.72
+input = 0.4
+output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384

From 17c659b0a7d6b0114a6fe046cdd1de289fc24f33 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 19:20:42 +0000
Subject: [PATCH 1203/2012] chore(sync): update DigitalOcean model catalog
 (#6979)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/glm-5.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/digitalocean/models/glm-5.3.toml b/providers/digitalocean/models/glm-5.3.toml
index 3325586991b..9ba7811a008 100644
--- a/providers/digitalocean/models/glm-5.3.toml
+++ b/providers/digitalocean/models/glm-5.3.toml
@@ -12,4 +12,4 @@ cache_read = 0.2
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 128_000

From 39e46fe84e884382d9e6070102f701642a92b9ec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 22:23:27 +0000
Subject: [PATCH 1204/2012] chore(sync): update OpenRouter model catalog
 (#6980)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k2-thinking.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 3ccd5a4f3ef..91ad22ed77f 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -8,6 +8,7 @@ field = "reasoning_details"
 [cost]
 input = 0.6
 output = 2.5
+cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352

From 732586172e2ed9100a447c4ced19539e180a8f70 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 22:23:33 +0000
Subject: [PATCH 1205/2012] chore(sync): update Kilo model catalog (#6981)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 3e83e7378b1..75e607a6e02 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 235_929
+output = 100_352

From 73050dbdfbbe7a508931de2eb895899ad1775ed7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 01:04:56 +0000
Subject: [PATCH 1206/2012] chore(sync): update OpenRouter model catalog
 (#6984)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 .../openrouter/models/google/gemma-4-26b-a4b-it.toml      | 8 ++++++--
 providers/openrouter/models/meta/muse-glimmer-30b.toml    | 4 ++--
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 7 +++----
 5 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..7c435d69ac6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 0.57816
+output = 1.73448
+cache_read = 0.018396
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index a27c6e610ef..7d230a2be6f 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -6,8 +6,12 @@ base_model = "google/gemma-4-26b-a4b-it"
 type = "toggle"
 
 [cost]
-input = 0.042
-output = 0.22
+input = 0.09
+output = 0.3
+cache_read = 0.05
+
+[limit]
+output = 235_929
 
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index 12578d9072c..3b46ee2987a 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -10,8 +10,8 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.3
-output = 1.1
+input = 0.35
+output = 1.5
 cache_read = 0.04
 
 [limit]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index d939b2691d5..bff4df6e006 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,10 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.092
+output = 3.432
+cache_read = 0.2028
 
 [limit]
 context = 1_310_720
-output = 943_718

From add0c5fd38e0671e6cc937553b3e1a54e3a1524f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 01:05:36 +0000
Subject: [PATCH 1207/2012] chore(sync): update Kilo model catalog (#6985)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 ++-
 providers/kilo/models/google/gemma-4-26b-a4b-it.toml     | 2 +-
 providers/kilo/models/z-ai/glm-5.3.toml                  | 1 -
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/google/gemma-4-26b-a4b-it.toml b/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
index efcf99dab3b..baa53082fe1 100644
--- a/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/kilo/models/google/gemma-4-26b-a4b-it.toml
@@ -10,7 +10,7 @@ input = 0.042
 output = 0.22
 
 [limit]
-context = 131_072
+output = 235_929
 
 [modalities]
 input = ["image", "text", "video"]
diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index 959b6c62561..e9d5bdad891 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -12,4 +12,3 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 943_718

From f67cf8516f19f9b7a64f704635584e8d3b5344f7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 02:34:33 +0000
Subject: [PATCH 1208/2012] chore(sync): update OpenRouter model catalog
 (#6987)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index c0c2c8b5cd7..2ce7e2e4c07 100644
--- a/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/openrouter/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -12,12 +12,12 @@ knowledge = "2025-06-30"
 open_weights = true
 
 [cost]
-input = 0.09
-output = 0.3
+input = 0.04815
+output = 0.19305
 
 [limit]
 context = 262_144
-output = 235_929
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 1f5af5de2fa..fd4d4c58220 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,11 +22,11 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.07
-cache_read = 0.003
+cache_read = 0.005
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From 71008c04bce244b8a3255a48a42a170f0d313e19 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 02:35:01 +0000
Subject: [PATCH 1209/2012] chore(sync): update Kilo model catalog (#6986)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml   | 4 ++--
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index 1da6f655c94..48baf441187 100644
--- a/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/kilo/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 262_144
-output = 235_929
+context = 128_000
+output = 32_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index e8bd485a54c..5ff5b94bf7a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -17,11 +17,11 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.07
-cache_read = 0.003
+cache_read = 0.005
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From fa90a5d07ef1613dad8e0ce350e7fe15e774e516 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:30:38 +0000
Subject: [PATCH 1210/2012] chore(sync): update Kilo model catalog (#6988)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index e0c53b97309..ffd58df078c 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.375
-output = 13.3
-cache_read = 0.2755
+input = 2.1
+output = 10.95
 
 [limit]
 context = 1_048_576

From 5ab29fce19855551cc7d10430f3a364454086ef2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:30:50 +0000
Subject: [PATCH 1211/2012] chore(sync): update OpenRouter model catalog
 (#6989)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 5 ++---
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1913a52d832..b94d9fc9091 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06566
-output = 0.13132
-cache_read = 0.013132
+input = 0.06538
+output = 0.13076
+cache_read = 0.013076
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 44bf8744327..9b28cb1e8f4 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.375
-output = 13.3
-cache_read = 0.2755
+input = 2.1
+output = 10.95
 
 [limit]
 context = 1_048_576

From 4de0ed89a60bb42b7ac9c33e1665a38a718e69f9 Mon Sep 17 00:00:00 2001
From: Zain Hasan <zain_has@hotmail.com>
Date: Sun, 13 Sep 2026 00:46:08 -0400
Subject: [PATCH 1212/2012] [together ai] add deepseek v4.1 flash (#6983)

* [together ai] add deepseek v4.1 flash

* Reduce cache_read cost in DeepSeek-V4.1-Flash
---
 .../models/deepseek-ai/DeepSeek-V4.1-Flash.toml     | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/togetherai/models/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/togetherai/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/togetherai/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..be629662429
--- /dev/null
+++ b/providers/togetherai/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,13 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576

From 6038dcfb3f14a6ac100e6ca0f586896c7258133a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 23:46:27 -0500
Subject: [PATCH 1213/2012] chore(sync): update LLM Gateway model catalog
 (#6966)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/tencent/deepseek-v4-flash.toml     | 18 ++++++++++++++
 .../models/tencent/deepseek-v4-pro.toml       | 18 ++++++++++++++
 .../models/tencent/glm-5-turbo.toml           | 15 ++++++++++++
 .../models/tencent/glm-5.1.toml               | 18 ++++++++++++++
 .../models/tencent/glm-5.2.toml               | 18 ++++++++++++++
 .../models/tencent/glm-5.toml                 | 19 +++++++++++++++
 .../models/tencent/glm-5v-turbo.toml          | 15 ++++++++++++
 .../models/tencent/hy-mt2-plus.toml           | 23 ++++++++++++++++++
 .../models/tencent/hy3.toml                   | 17 +++++++++++++
 .../models/tencent/hy4-preview.toml           | 16 +++++++++++++
 .../models/tencent/kimi-k2.6.toml             | 14 +++++++++++
 .../tencent/kimi-k2.7-code-highspeed.toml     | 14 +++++++++++
 .../models/tencent/kimi-k2.7-code.toml        | 14 +++++++++++
 .../models/tencent/kimi-k3.toml               | 14 +++++++++++
 .../models/tencent/mimo-v2.5-pro.toml         | 24 +++++++++++++++++++
 .../models/tencent/minimax-m2.7.toml          | 16 +++++++++++++
 .../models/tencent/minimax-m3.toml            | 16 +++++++++++++
 17 files changed, 289 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/tencent/deepseek-v4-flash.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/deepseek-v4-pro.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/glm-5-turbo.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/glm-5.1.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/glm-5.2.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/glm-5.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/glm-5v-turbo.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/hy-mt2-plus.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/hy3.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/hy4-preview.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/kimi-k2.6.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/kimi-k2.7-code-highspeed.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/kimi-k2.7-code.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/kimi-k3.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/mimo-v2.5-pro.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/minimax-m2.7.toml
 create mode 100644 providers/llmgateway-providers/models/tencent/minimax-m3.toml

diff --git a/providers/llmgateway-providers/models/tencent/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/tencent/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..697b620ad81
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/deepseek-v4-flash.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4-flash"
+name = "DeepSeek V4 Flash (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+output = 393_216
diff --git a/providers/llmgateway-providers/models/tencent/deepseek-v4-pro.toml b/providers/llmgateway-providers/models/tencent/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..146ff62b8e3
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/deepseek-v4-pro.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4-pro"
+name = "DeepSeek V4 Pro (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.00363
+
+[limit]
+output = 393_216
diff --git a/providers/llmgateway-providers/models/tencent/glm-5-turbo.toml b/providers/llmgateway-providers/models/tencent/glm-5-turbo.toml
new file mode 100644
index 00000000000..4a609a2477e
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/glm-5-turbo.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5-turbo"
+name = "GLM-5 Turbo (Tencent Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.24
+
+[limit]
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/glm-5.1.toml b/providers/llmgateway-providers/models/tencent/glm-5.1.toml
new file mode 100644
index 00000000000..f1e5a56aed7
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/glm-5.1.toml
@@ -0,0 +1,18 @@
+base_model = "zhipuai/glm-5.1"
+name = "GLM-5.1 (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/glm-5.2.toml b/providers/llmgateway-providers/models/tencent/glm-5.2.toml
new file mode 100644
index 00000000000..62e6bf18232
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/glm-5.2.toml
@@ -0,0 +1,18 @@
+base_model = "zhipuai/glm-5.2"
+name = "GLM-5.2 (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/glm-5.toml b/providers/llmgateway-providers/models/tencent/glm-5.toml
new file mode 100644
index 00000000000..11e53b3534b
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/glm-5.toml
@@ -0,0 +1,19 @@
+base_model = "zhipuai/glm-5"
+name = "GLM-5 (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1
+output = 3.2
+cache_read = 0.2
+
+[limit]
+context = 200_000
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/glm-5v-turbo.toml b/providers/llmgateway-providers/models/tencent/glm-5v-turbo.toml
new file mode 100644
index 00000000000..93ee5736129
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/glm-5v-turbo.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5v-turbo"
+name = "GLM-5V Turbo (Tencent Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.24
+
+[limit]
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/hy-mt2-plus.toml b/providers/llmgateway-providers/models/tencent/hy-mt2-plus.toml
new file mode 100644
index 00000000000..a17fce21909
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/hy-mt2-plus.toml
@@ -0,0 +1,23 @@
+name = "Hy-MT2 Plus (Tencent Cloud)"
+description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
+family = "Hy"
+release_date = "2026-05-21"
+last_updated = "2026-05-21"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = false
+open_weights = false
+
+[cost]
+input = 0.074
+output = 0.295
+
+[limit]
+context = 8_192
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/llmgateway-providers/models/tencent/hy3.toml b/providers/llmgateway-providers/models/tencent/hy3.toml
new file mode 100644
index 00000000000..1b895b0482a
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/hy3.toml
@@ -0,0 +1,17 @@
+base_model = "tencent/hy3"
+base_model_omit = ["limit.input"]
+name = "Hy3 (Tencent Cloud)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.132
+output = 0.528
+cache_read = 0.033
+
+[limit]
+context = 262_144
+output = 131_072
diff --git a/providers/llmgateway-providers/models/tencent/hy4-preview.toml b/providers/llmgateway-providers/models/tencent/hy4-preview.toml
new file mode 100644
index 00000000000..587eeffca84
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/hy4-preview.toml
@@ -0,0 +1,16 @@
+base_model = "tencent/hy4-preview"
+name = "Hy4 Preview (Tencent Cloud)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.834
+output = 2.501
+cache_read = 0.042
+
+[limit]
+context = 1_000_000
+output = 65_536
diff --git a/providers/llmgateway-providers/models/tencent/kimi-k2.6.toml b/providers/llmgateway-providers/models/tencent/kimi-k2.6.toml
new file mode 100644
index 00000000000..e8e971eb490
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/kimi-k2.6.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k2.6"
+name = "Kimi K2.6 (Tencent Cloud)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.858
+output = 3.566
+cache_read = 0.145
diff --git a/providers/llmgateway-providers/models/tencent/kimi-k2.7-code-highspeed.toml b/providers/llmgateway-providers/models/tencent/kimi-k2.7-code-highspeed.toml
new file mode 100644
index 00000000000..773cdd8b8c0
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/kimi-k2.7-code-highspeed.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k2.7-code-highspeed"
+name = "Kimi K2.7 Code Highspeed (Tencent Cloud)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.9
+output = 8
+cache_read = 0.38
diff --git a/providers/llmgateway-providers/models/tencent/kimi-k2.7-code.toml b/providers/llmgateway-providers/models/tencent/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..090a8c09335
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/kimi-k2.7-code.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k2.7-code"
+name = "Kimi K2.7 Code (Tencent Cloud)"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/llmgateway-providers/models/tencent/kimi-k3.toml b/providers/llmgateway-providers/models/tencent/kimi-k3.toml
new file mode 100644
index 00000000000..046d3f6d3ca
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/kimi-k3.toml
@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (Tencent Cloud)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+output = 1_048_576
diff --git a/providers/llmgateway-providers/models/tencent/mimo-v2.5-pro.toml b/providers/llmgateway-providers/models/tencent/mimo-v2.5-pro.toml
new file mode 100644
index 00000000000..745d3f8b533
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/mimo-v2.5-pro.toml
@@ -0,0 +1,24 @@
+base_model = "xiaomi/mimo-v2.5-pro"
+name = "MiMo V2.5 Pro (Tencent Cloud)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 2
+output = 6
+cache_read = 0.4
+
+[limit]
+context = 1_000_000
diff --git a/providers/llmgateway-providers/models/tencent/minimax-m2.7.toml b/providers/llmgateway-providers/models/tencent/minimax-m2.7.toml
new file mode 100644
index 00000000000..b156b040421
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/minimax-m2.7.toml
@@ -0,0 +1,16 @@
+base_model = "minimax/MiniMax-M2.7"
+name = "MiniMax M2.7 (Tencent Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+context = 200_000
+output = 128_000
diff --git a/providers/llmgateway-providers/models/tencent/minimax-m3.toml b/providers/llmgateway-providers/models/tencent/minimax-m3.toml
new file mode 100644
index 00000000000..f6574935452
--- /dev/null
+++ b/providers/llmgateway-providers/models/tencent/minimax-m3.toml
@@ -0,0 +1,16 @@
+base_model = "minimax/MiniMax-M3"
+name = "MiniMax M3 (Tencent Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+context = 1_000_000
+output = 131_072

From 7d460072c407cd0ff2483051b6fb802fe62d56dd Mon Sep 17 00:00:00 2001
From: NamanSinha <naman.shripura@gmail.com>
Date: Sun, 13 Sep 2026 10:17:03 +0530
Subject: [PATCH 1214/2012] Add deepseek-v4-flash and glm-5.3 to AgentRouter
 provider (#6958)

---
 providers/agentrouter/models/deepseek-v4-flash.toml | 6 ++++++
 providers/agentrouter/models/glm-5.3.toml           | 5 +++++
 2 files changed, 11 insertions(+)
 create mode 100644 providers/agentrouter/models/deepseek-v4-flash.toml
 create mode 100644 providers/agentrouter/models/glm-5.3.toml

diff --git a/providers/agentrouter/models/deepseek-v4-flash.toml b/providers/agentrouter/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..53933a438bb
--- /dev/null
+++ b/providers/agentrouter/models/deepseek-v4-flash.toml
@@ -0,0 +1,6 @@
+# Catalog and endpoint support: https://agentrouter.org/api/pricing
+# AgentRouter publishes relative token ratios, not USD prices, so cost is intentionally omitted.
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+base_model = "deepseek/deepseek-v4-flash"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
diff --git a/providers/agentrouter/models/glm-5.3.toml b/providers/agentrouter/models/glm-5.3.toml
new file mode 100644
index 00000000000..83ef4f53d00
--- /dev/null
+++ b/providers/agentrouter/models/glm-5.3.toml
@@ -0,0 +1,5 @@
+# Catalog and endpoint support: https://agentrouter.org/api/pricing
+# AgentRouter publishes relative token ratios, not USD prices, so cost is intentionally omitted.
+# GLM-5.3 always reasons (thinking cannot be disabled); effort levels low|high|max, default max.
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]

From a13a83a31433e503b418fa2ad9c6fa2147e8b70d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 23:47:18 -0500
Subject: [PATCH 1215/2012] chore(sync): update Requesty model catalog (#6953)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../requesty/models/deepseek-v4.1-flash.toml  |  6 ++--
 .../models/deepseek-v4.1-flash@eu.toml        | 18 +++++++++++
 .../models/thinkingcap-qwen3.6-27b.toml       | 30 -------------------
 .../models/thinkingcap-qwen3.6-27b@eu.toml    | 30 -------------------
 4 files changed, 21 insertions(+), 63 deletions(-)
 create mode 100644 providers/requesty/models/deepseek-v4.1-flash@eu.toml
 delete mode 100644 providers/requesty/models/thinkingcap-qwen3.6-27b.toml
 delete mode 100644 providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml

diff --git a/providers/requesty/models/deepseek-v4.1-flash.toml b/providers/requesty/models/deepseek-v4.1-flash.toml
index 83545096971..193eb7f54cb 100644
--- a/providers/requesty/models/deepseek-v4.1-flash.toml
+++ b/providers/requesty/models/deepseek-v4.1-flash.toml
@@ -8,9 +8,9 @@ values = ["none", "low", "medium", "high", "max"]
 type = "budget_tokens"
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.5
+output = 1.5
+cache_read = 0.05
 
 [limit]
 context = 1_048_576
diff --git a/providers/requesty/models/deepseek-v4.1-flash@eu.toml b/providers/requesty/models/deepseek-v4.1-flash@eu.toml
new file mode 100644
index 00000000000..ebcd8dc5960
--- /dev/null
+++ b/providers/requesty/models/deepseek-v4.1-flash@eu.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.5
+output = 1.5
+cache_read = 0.05
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
deleted file mode 100644
index 6b202e90585..00000000000
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "thinkingcap-qwen3.6-27b"
-description = "ThinkingCap-Qwen3.6-27B is a reasoning tuned model from BottlecapAI built on Qwen3.6 27B. It supports extended thinking with tool calling and a 256K context window. Served via Sference."
-family = "qwen3.6"
-release_date = "2026-07-13"
-last_updated = "2026-07-13"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-
-[cost]
-input = 0.4
-output = 2.6
-cache_read = 0.05
-
-[limit]
-context = 262_144
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml b/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
deleted file mode 100644
index d72b7f5786e..00000000000
--- a/providers/requesty/models/thinkingcap-qwen3.6-27b@eu.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "thinkingcap-qwen3.6-27b@eu"
-description = "ThinkingCap-Qwen3.6-27B is a reasoning tuned model from BottlecapAI built on Qwen3.6 27B. It supports extended thinking with tool calling and a 256K context window. Served via Sference."
-family = "qwen3.6"
-release_date = "2026-07-13"
-last_updated = "2026-07-13"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-
-[cost]
-input = 0.4
-output = 2.6
-cache_read = 0.05
-
-[limit]
-context = 262_144
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 8be4b14f22c75bece2503d8350e429e9a58c3dd2 Mon Sep 17 00:00:00 2001
From: YongYuH <larry830326@hotmail.com>
Date: Sun, 13 Sep 2026 12:47:57 +0800
Subject: [PATCH 1216/2012] feat(ollama-cloud): add deepseek-v4-pro:0813 model
 (#6941)

Co-authored-by: larry.huang <larry.huang@scmp.com>
---
 .../ollama-cloud/models/deepseek-v4-pro:0813.toml  | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/ollama-cloud/models/deepseek-v4-pro:0813.toml

diff --git a/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml b/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml
new file mode 100644
index 00000000000..9ae3ed43255
--- /dev/null
+++ b/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml
@@ -0,0 +1,14 @@
+# Toggle: native POST /api/chat and /api/generate use `think = true|false`.
+# Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
+# or `reasoning.effort` with high|max (DeepSeek V4 maps low/medium -> high).
+# Sources: https://docs.ollama.com/capabilities/thinking, https://docs.ollama.com/openai
+# API-verified context: https://ollama.com/api/show (deepseek4.context_length = 1048576)
+# Output: Ollama has no separate output cap; deepseek4 peers on this host
+# (deepseek-v4-flash:0731, deepseek-v4-pro) serve output = 1048576.
+base_model = "deepseek/deepseek-v4-pro-0813"
+
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
+
+[limit]
+context = 1048576
+output = 1048576

From 2baa648c2f4326cfd786870dc9f6017089dcb8b6 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Sun, 13 Sep 2026 11:48:30 +0700
Subject: [PATCH 1217/2012] add GLM-5.3-Flash to NVIDIA (#6924)

---
 providers/nvidia/models/z-ai/glm-5.3-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/nvidia/models/z-ai/glm-5.3-flash.toml

diff --git a/providers/nvidia/models/z-ai/glm-5.3-flash.toml b/providers/nvidia/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..4e846d402ad
--- /dev/null
+++ b/providers/nvidia/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,14 @@
+# Hosted NIM: model listed at GET https://integrate.api.nvidia.com/v1/models
+# (verified 2026-09-12). Thinking is always enabled; reasoning budget controlled
+# with top-level `reasoning_effort`: low|high|max, default max.
+# https://docs.nvidia.com/nim/vision-language-models/latest/get-started/advanced/get-started-glm-5-3-flash.html
+base_model = "zhipuai/glm-5.3-flash"
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0

From 46d4181b0ca45c7c2c29fc2fdf665fd6c2566bb4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 00:12:50 -0500
Subject: [PATCH 1218/2012] feat(sync): track Ollama Cloud models (#6990)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/index.ts               |  5 +-
 .../core/src/sync/providers/ollama-cloud.ts   | 72 +++++++++++++++++++
 packages/core/test/ollama-cloud.test.ts       | 66 +++++++++++++++++
 sync.md                                       |  3 +
 4 files changed, 145 insertions(+), 1 deletion(-)
 create mode 100644 packages/core/src/sync/providers/ollama-cloud.ts
 create mode 100644 packages/core/test/ollama-cloud.test.ts

diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 7f88aacb5bb..e6c30874235 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -28,6 +28,7 @@ import { llmgateway, llmgatewayProviders } from "./providers/llmgateway.js";
 import { mergeGateway } from "./providers/merge-gateway.js";
 import { meta } from "./providers/meta.js";
 import { nanoGpt } from "./providers/nano-gpt.js";
+import { ollamaCloud } from "./providers/ollama-cloud.js";
 import { openai } from "./providers/openai.js";
 import { ofox } from "./providers/ofox.js";
 import { openrouter } from "./providers/openrouter.js";
@@ -154,6 +155,7 @@ export const providers: {
   meta: SyncProvider<any>;
   "nano-gpt": SyncProvider<any>;
   ofox: SyncProvider<any>;
+  "ollama-cloud": SyncProvider<any>;
   openai: SyncProvider<any>;
   openrouter: SyncProvider<any>;
   ovhcloud: SyncProvider<any>;
@@ -189,6 +191,7 @@ export const providers: {
   meta,
   "nano-gpt": nanoGpt,
   ofox,
+  "ollama-cloud": ollamaCloud,
   openai,
   openrouter,
   ovhcloud,
@@ -219,7 +222,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "meta", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
diff --git a/packages/core/src/sync/providers/ollama-cloud.ts b/packages/core/src/sync/providers/ollama-cloud.ts
new file mode 100644
index 00000000000..5eb499b534f
--- /dev/null
+++ b/packages/core/src/sync/providers/ollama-cloud.ts
@@ -0,0 +1,72 @@
+import { z } from "zod";
+
+import { AuthoredModel } from "../../schema.js";
+import type { ExistingModel, SyncProvider, SyncedBaseModel, SyncedModel } from "../index.js";
+
+const API_ENDPOINT = "https://ollama.com/v1/models";
+
+export const OllamaCloudModel = z.object({
+  id: z.string().min(1),
+  object: z.literal("model"),
+  created: z.number().int().nonnegative(),
+  owned_by: z.string(),
+}).passthrough();
+
+const OllamaCloudResponse = z.object({
+  object: z.literal("list"),
+  data: z.array(OllamaCloudModel),
+}).passthrough();
+
+export type OllamaCloudModel = z.infer<typeof OllamaCloudModel>;
+
+export function parseOllamaCloudModels(raw: unknown) {
+  return OllamaCloudResponse.parse(raw).data;
+}
+
+function preserveAuthoredModel(id: string, authored: ExistingModel): SyncedModel {
+  if (authored.base_model !== undefined) return authored as SyncedBaseModel;
+
+  const parsed = AuthoredModel.safeParse({ id, ...authored });
+  if (!parsed.success) {
+    parsed.error.cause = { provider: "ollama-cloud", model: id };
+    throw parsed.error;
+  }
+  const { id: _id, ...model } = parsed.data;
+  return model;
+}
+
+export async function fetchOllamaCloudModels(fetcher: typeof fetch = fetch) {
+  const response = await fetcher(API_ENDPOINT);
+  if (!response.ok) {
+    throw new Error(`Ollama Cloud models request failed: ${response.status} ${response.statusText}`);
+  }
+  return response.json();
+}
+
+export const ollamaCloud = {
+  id: "ollama-cloud",
+  name: "Ollama Cloud",
+  modelsDir: "providers/ollama-cloud/models",
+  skipCreates: true,
+  trackMissingModels: true,
+  deleteMissing: false,
+  sourceID(model) {
+    return model.id;
+  },
+  skippedNotice(ids) {
+    if (ids.length === 0) return [];
+    return [
+      `${ids.length} Ollama Cloud models returned by the API are missing from the local catalog and require hand-authored metadata.`,
+      `Missing remote IDs: ${ids.map((id) => `\`${id}\``).join(", ")}`,
+    ];
+  },
+  async fetchModels() {
+    return fetchOllamaCloudModels();
+  },
+  parseModels: parseOllamaCloudModels,
+  translateModel(model, context) {
+    const authored = context.authored(model.id);
+    if (authored === undefined) return undefined;
+    return { id: model.id, model: preserveAuthoredModel(model.id, authored) };
+  },
+} satisfies SyncProvider<OllamaCloudModel>;
diff --git a/packages/core/test/ollama-cloud.test.ts b/packages/core/test/ollama-cloud.test.ts
new file mode 100644
index 00000000000..46a0a81057b
--- /dev/null
+++ b/packages/core/test/ollama-cloud.test.ts
@@ -0,0 +1,66 @@
+import { expect, test } from "bun:test";
+
+import {
+  fetchOllamaCloudModels,
+  ollamaCloud,
+  parseOllamaCloudModels,
+  type OllamaCloudModel,
+} from "../src/sync/providers/ollama-cloud.js";
+
+function model(overrides: Partial<OllamaCloudModel> = {}): OllamaCloudModel {
+  return {
+    id: "deepseek-v4-pro:0813",
+    object: "model",
+    created: 1_786_633_200,
+    owned_by: "ollama",
+    ...overrides,
+  };
+}
+
+test("parses the public Ollama Cloud model inventory", () => {
+  expect(parseOllamaCloudModels({
+    object: "list",
+    data: [model()],
+  })).toEqual([model()]);
+});
+
+test("fetches the public Ollama Cloud model endpoint without authentication", async () => {
+  let request: RequestInfo | URL | undefined;
+  const fetcher: typeof fetch = async (input) => {
+    request = input;
+    return Response.json({ object: "list", data: [model()] });
+  };
+
+  await expect(fetchOllamaCloudModels(fetcher)).resolves.toEqual({
+    object: "list",
+    data: [model()],
+  });
+  expect(String(request)).toBe("https://ollama.com/v1/models");
+});
+
+test("tracks remote-only Ollama Cloud models without creating or deleting TOMLs", () => {
+  expect(ollamaCloud.skipCreates).toBe(true);
+  expect(ollamaCloud.trackMissingModels).toBe(true);
+  expect(ollamaCloud.deleteMissing).toBe(false);
+  expect(ollamaCloud.sourceID(model())).toBe("deepseek-v4-pro:0813");
+});
+
+test("preserves existing Ollama Cloud provider overrides", () => {
+  const authored = {
+    base_model: "deepseek/deepseek-v4-pro-0813",
+    reasoning_options: [
+      { type: "toggle" as const },
+      { type: "effort" as const, values: ["high", "max"] },
+    ],
+    limit: { context: 1_048_576, output: 1_048_576 },
+  };
+
+  expect(ollamaCloud.translateModel(model(), {
+    existing: () => undefined,
+    authored: () => authored,
+  })).toEqual({ id: "deepseek-v4-pro:0813", model: authored });
+  expect(ollamaCloud.translateModel(model({ id: "new-model" }), {
+    existing: () => undefined,
+    authored: () => undefined,
+  })).toBeUndefined();
+});
diff --git a/sync.md b/sync.md
index 7595a924dc4..06ceb1b2a42 100644
--- a/sync.md
+++ b/sync.md
@@ -20,6 +20,7 @@ The grouped sync targets are available for local convenience, but CI syncs each
 - `bun models:sync kilo` syncs only Kilo.
 - `bun models:sync merge-gateway` syncs only Merge Gateway.
 - `bun models:sync openai` syncs only OpenAI catalog availability.
+- `bun models:sync ollama-cloud` syncs Ollama Cloud catalog availability.
 - `bun models:sync github-copilot` syncs only GitHub Copilot pricing.
 - `bun models:sync tinfoil` syncs only Tinfoil.
 - `bun models:sync aggregators --dry-run` prints changes without writing model files.
@@ -64,6 +65,8 @@ The first Actions run may open a batch of issues per provider, including remote
 
 Pioneer and Ofox track remote-only chat models as missing-model issues. Their APIs are not authoritative enough to create complete TOMLs directly, so the issue-fixer agent researches the missing canonical and provider-specific metadata before opening a PR.
 
+Ollama Cloud tracks every remote-only ID from its public `/v1/models` endpoint as a missing-model issue. Existing provider TOMLs and entries absent from the endpoint are preserved because the inventory does not provide enough metadata to author complete entries or determine removals safely.
+
 OpenAI also sets `trackMissingModels: false`: `/v1/models` is scoped to the automation account and mixes public models with legacy, internal experiment, dated snapshot, and non-catalog IDs without lifecycle metadata. Existing OpenAI TOMLs are still preserved by the availability sync.
 
 Google sets `trackMissingModels: false`: `/v1beta/models` does not expose lifecycle metadata and can retain shut-down models, superseded snapshots, moving aliases, and EAP IDs. Existing Google TOMLs are still updated from API-authoritative fields.

From ec65d64750711ee47fb36b472cac88840d299623 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 05:24:46 +0000
Subject: [PATCH 1219/2012] chore(sync): update OpenRouter model catalog
 (#6991)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-r1.toml     | 2 +-
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-r1.toml b/providers/openrouter/models/deepseek/deepseek-r1.toml
index 69fd9fdd280..08925e91e21 100644
--- a/providers/openrouter/models/deepseek/deepseek-r1.toml
+++ b/providers/openrouter/models/deepseek/deepseek-r1.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-r1"
-structured_output = true
+structured_output = false
 reasoning_options = []
 
 [cost]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b94d9fc9091..af2c01ab974 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06538
-output = 0.13076
-cache_read = 0.013076
+input = 0.0651
+output = 0.1302
+cache_read = 0.01302
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index fd4d4c58220..c1e3b255d01 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.07
-cache_read = 0.005
+input = 0.0352
+output = 0.1056
+cache_read = 0.00112
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From eeccc534ec375e79bc5e1ec1271d64af6d881d97 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 05:24:51 +0000
Subject: [PATCH 1220/2012] chore(sync): update Kilo model catalog (#6992)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-r1.toml           | 2 +-
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-r1.toml b/providers/kilo/models/deepseek/deepseek-r1.toml
index a4d870608da..0154d4c6e92 100644
--- a/providers/kilo/models/deepseek/deepseek-r1.toml
+++ b/providers/kilo/models/deepseek/deepseek-r1.toml
@@ -1,6 +1,6 @@
 base_model = "deepseek/deepseek-r1"
 description = "DeepSeek reasoning model for multi-step analysis, math, coding, and tools"
-structured_output = true
+structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 5ff5b94bf7a..67ee14fe333 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.07
-cache_read = 0.005
+input = 0.0352
+output = 0.1056
+cache_read = 0.00112
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 84021294d0e4415ca1c3594f65df3b1d88d6fc6b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 05:25:30 +0000
Subject: [PATCH 1221/2012] chore(sync): update Eden AI model catalog (#6993)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/openai/gpt-5.4-pro.toml    |  2 --
 providers/edenai/models/openai/gpt-5.5-pro.toml    |  2 --
 providers/edenai/models/openai/gpt-pro-latest.toml |  2 --
 .../edenai/models/qwen/deepseek-v4-flash-0731.toml |  4 ++--
 .../edenai/models/qwen/deepseek-v4-pro-0813.toml   |  4 ++--
 .../deepseek-ai/DeepSeek-V4.1-Flash.toml           | 14 ++++++++++++++
 6 files changed, 18 insertions(+), 10 deletions(-)
 create mode 100644 providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/edenai/models/openai/gpt-5.4-pro.toml b/providers/edenai/models/openai/gpt-5.4-pro.toml
index 18caf358b3c..5213b8ffc52 100644
--- a/providers/edenai/models/openai/gpt-5.4-pro.toml
+++ b/providers/edenai/models/openai/gpt-5.4-pro.toml
@@ -8,13 +8,11 @@ values = ["medium", "high", "xhigh"]
 [cost]
 input = 30
 output = 180
-cache_read = 3
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
 input = 60
 output = 270
-cache_read = 6
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/openai/gpt-5.5-pro.toml b/providers/edenai/models/openai/gpt-5.5-pro.toml
index eca4c05cfd4..664304915ee 100644
--- a/providers/edenai/models/openai/gpt-5.5-pro.toml
+++ b/providers/edenai/models/openai/gpt-5.5-pro.toml
@@ -7,10 +7,8 @@ values = ["medium", "high", "xhigh"]
 [cost]
 input = 30
 output = 180
-cache_read = 3
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
 input = 60
 output = 270
-cache_read = 6
diff --git a/providers/edenai/models/openai/gpt-pro-latest.toml b/providers/edenai/models/openai/gpt-pro-latest.toml
index 31f2a4df936..b52aca44d42 100644
--- a/providers/edenai/models/openai/gpt-pro-latest.toml
+++ b/providers/edenai/models/openai/gpt-pro-latest.toml
@@ -8,10 +8,8 @@ values = ["medium", "high", "xhigh"]
 [cost]
 input = 30
 output = 180
-cache_read = 3
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
 input = 60
 output = 270
-cache_read = 6
diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366
diff --git a/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..6de196ab790
--- /dev/null
+++ b/providers/edenai/models/together_ai/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576

From 36f890984779dd7ebfb48bcbfe7d43b6b407fc3c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 06:41:11 +0000
Subject: [PATCH 1222/2012] chore(sync): update OpenRouter model catalog
 (#6995)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index af2c01ab974..a9469f150d2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0651
-output = 0.1302
-cache_read = 0.01302
+input = 0.04984
+output = 0.09968
+cache_read = 0.009968
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 11d35e9f1e7..5274a327618 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8727
-output = 3.36
-cache_read = 0.1639
+input = 0.936
+output = 3.168
+cache_read = 0.1872
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From b9f3f62b1462dcb79fb73a91d437eeb498749247 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 06:41:16 +0000
Subject: [PATCH 1223/2012] chore(sync): update Kilo model catalog (#6996)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2829ed5b448..05cbd588a93 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8727
-output = 3.36
-cache_read = 0.1639
+input = 0.936
+output = 3.168
+cache_read = 0.1872
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From dd27428f46e45ca0a0394d93cf118db1150ee946 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 07:25:56 +0000
Subject: [PATCH 1224/2012] chore(sync): update OpenRouter model catalog
 (#6999)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k2-0905.toml    | 2 +-
 .../openrouter/models/moonshotai/kimi-k2-thinking.toml      | 2 +-
 providers/openrouter/models/moonshotai/kimi-k2.toml         | 2 +-
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a9469f150d2..887f8fb33c5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04984
-output = 0.09968
-cache_read = 0.009968
+input = 0.04956
+output = 0.09912
+cache_read = 0.009912
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-0905.toml b/providers/openrouter/models/moonshotai/kimi-k2-0905.toml
index af63305cb7e..a867b9f2383 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-0905.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-0905.toml
@@ -17,7 +17,7 @@ output = 2.5
 
 [limit]
 context = 262_144
-output = 100_352
+output = 98_304
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
index 91ad22ed77f..589eb077162 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2-thinking.toml
@@ -11,4 +11,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 98_304
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.toml b/providers/openrouter/models/moonshotai/kimi-k2.toml
index da05ef41c84..6bc9044c209 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.toml
@@ -17,7 +17,7 @@ output = 2.3
 
 [limit]
 context = 131_072
-output = 100_352
+output = 98_304
 
 [modalities]
 input = ["text"]

From 5e72307283aca41dbd853bb579e8f2c0ab2d7722 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 07:25:58 +0000
Subject: [PATCH 1225/2012] chore(sync): update Kilo model catalog (#6998)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k2-0905.toml     | 2 +-
 providers/kilo/models/moonshotai/kimi-k2-thinking.toml | 2 +-
 providers/kilo/models/moonshotai/kimi-k2.toml          | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k2-0905.toml b/providers/kilo/models/moonshotai/kimi-k2-0905.toml
index 43681235bd0..39cad5b17ea 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-0905.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-0905.toml
@@ -16,7 +16,7 @@ output = 2.5
 
 [limit]
 context = 262_144
-output = 100_352
+output = 98_304
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
index 75e607a6e02..c0ea113422d 100644
--- a/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2-thinking.toml
@@ -12,4 +12,4 @@ output = 2.5
 cache_read = 0.15
 
 [limit]
-output = 100_352
+output = 98_304
diff --git a/providers/kilo/models/moonshotai/kimi-k2.toml b/providers/kilo/models/moonshotai/kimi-k2.toml
index 86b749e2024..5c774ee8a6d 100644
--- a/providers/kilo/models/moonshotai/kimi-k2.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2.toml
@@ -16,7 +16,7 @@ output = 2.3
 
 [limit]
 context = 131_072
-output = 100_352
+output = 98_304
 
 [modalities]
 input = ["text"]

From 2532e39852174e855983d28335d6301652772437 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 08:30:34 +0000
Subject: [PATCH 1226/2012] chore(sync): update OpenRouter model catalog
 (#7000)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 887f8fb33c5..58e87cdc226 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04956
-output = 0.09912
-cache_read = 0.009912
+input = 0.04928
+output = 0.09856
+cache_read = 0.009856
 
 [limit]
 context = 1_048_576

From 391c025baeaeb98a50b365bf7fa596309b17c436 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 08:30:50 +0000
Subject: [PATCH 1227/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7001)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/deepseek-v4.1-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/llmgateway/models/deepseek-v4.1-flash.toml b/providers/llmgateway/models/deepseek-v4.1-flash.toml
index 9a73dde4fdf..db5dbbe5fbc 100644
--- a/providers/llmgateway/models/deepseek-v4.1-flash.toml
+++ b/providers/llmgateway/models/deepseek-v4.1-flash.toml
@@ -2,7 +2,7 @@ base_model = "deepseek/deepseek-v4.1-flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high", "max"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.15

From 7b8e68f94959e93d27f92ca0e8709344d3013e1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 08:30:53 +0000
Subject: [PATCH 1228/2012] chore(sync): update LLM Gateway model catalog
 (#7002)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/deepseek-v4.1-flash.toml    | 15 +++++++++++++++
 .../models/fireworks/deepseek-v4.1-flash.toml    | 15 +++++++++++++++
 .../models/novita/deepseek-v4.1-flash.toml       | 16 ++++++++++++++++
 .../models/together-ai/deepseek-v4.1-flash.toml  | 15 +++++++++++++++
 4 files changed, 61 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/deepinfra/deepseek-v4.1-flash.toml
 create mode 100644 providers/llmgateway-providers/models/fireworks/deepseek-v4.1-flash.toml
 create mode 100644 providers/llmgateway-providers/models/novita/deepseek-v4.1-flash.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/deepinfra/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/deepinfra/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..c5319df2e3b
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 0.6
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/llmgateway-providers/models/fireworks/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/fireworks/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..b7f489724a1
--- /dev/null
+++ b/providers/llmgateway-providers/models/fireworks/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Fireworks AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/llmgateway-providers/models/novita/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/novita/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..06e79ea19ce
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/deepseek-v4.1-flash.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (NovitaAI)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/llmgateway-providers/models/together-ai/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/together-ai/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..3d215b854a0
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216

From 4a0154da2410bca18e97be812fecb9e8f85466e8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 11:21:38 +0000
Subject: [PATCH 1229/2012] chore(sync): update OpenRouter model catalog
 (#7005)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 58e87cdc226..141d4b41ca6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04928
-output = 0.09856
-cache_read = 0.009856
+input = 0.049
+output = 0.098
+cache_read = 0.0098
 
 [limit]
 context = 1_048_576

From 2671cfe1743e05e420da45c143b0ec96ae61a8f1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 12:32:31 +0000
Subject: [PATCH 1230/2012] chore(sync): update OpenRouter model catalog
 (#7007)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 141d4b41ca6..58e87cdc226 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.049
-output = 0.098
-cache_read = 0.0098
+input = 0.04928
+output = 0.09856
+cache_read = 0.009856
 
 [limit]
 context = 1_048_576

From a2a673950c6e09af0dbcb4744116401fc1ee0048 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 13:22:50 +0000
Subject: [PATCH 1231/2012] chore(sync): update OpenRouter model catalog
 (#7009)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 58e87cdc226..141d4b41ca6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04928
-output = 0.09856
-cache_read = 0.009856
+input = 0.049
+output = 0.098
+cache_read = 0.0098
 
 [limit]
 context = 1_048_576

From 09de18239264652101540eda401d54643e3f5ed3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 14:23:34 +0000
Subject: [PATCH 1232/2012] chore(sync): update OpenRouter model catalog
 (#7011)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 141d4b41ca6..2309a32472b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.049
-output = 0.098
-cache_read = 0.0098
+input = 0.04872
+output = 0.09744
+cache_read = 0.009744
 
 [limit]
 context = 1_048_576

From cc5b38714d129333c03043830a0833cbd4a50351 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 15:23:09 +0000
Subject: [PATCH 1233/2012] chore(sync): update Eden AI model catalog (#7012)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 4 ++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..4c0fef2aa45 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.176
+output = 0.528
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..b3756adb444 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.5808
+output = 1.7424

From 015465698c286e6b3dd618d1e79484ca5e2690e3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 15:23:11 +0000
Subject: [PATCH 1234/2012] chore(sync): update Kilo model catalog (#7014)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000

From dfa3c8f02fb8a3e3ad80161f9b81bc25aeb723a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 15:23:14 +0000
Subject: [PATCH 1235/2012] chore(sync): update OpenRouter model catalog
 (#7013)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 +++----
 providers/openrouter/models/z-ai/glm-5.3-flash.toml        | 6 +++---
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 2309a32472b..814013f36dd 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04872
-output = 0.09744
-cache_read = 0.009744
+input = 0.04844
+output = 0.09688
+cache_read = 0.009688
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c435d69ac6..3cbd82a5b7c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.018396
+input = 0.57948
+output = 1.73844
+cache_read = 0.019316
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index c54d27b9d18..d5918c075a2 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720

From a6b6ace63541598bd649716f55a7726f03a4264c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 16:25:19 +0000
Subject: [PATCH 1236/2012] chore(sync): update OpenRouter model catalog
 (#7015)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 814013f36dd..b5c7d255938 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04844
-output = 0.09688
-cache_read = 0.009688
+input = 0.04816
+output = 0.09632
+cache_read = 0.009632
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 22aa0a314657189ef12e9d6124b8bb7b86d6bb0d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 17:21:30 +0000
Subject: [PATCH 1237/2012] chore(sync): update OpenRouter model catalog
 (#7016)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b5c7d255938..14ec7ef4a45 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04816
-output = 0.09632
-cache_read = 0.009632
+input = 0.04788
+output = 0.09576
+cache_read = 0.009576
 
 [limit]
 context = 1_048_576

From d9b18aba5889467389adbc165daeb0478caf2ba1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 18:27:44 +0000
Subject: [PATCH 1238/2012] chore(sync): update Kilo model catalog (#7019)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index ffd58df078c..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -17,6 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 2.1
 output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 521e4ee3a710a5dea672e234ceddcc89c2af5dd0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 18:27:47 +0000
Subject: [PATCH 1239/2012] chore(sync): update OpenRouter model catalog
 (#7018)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../models/inference-net/schematron-v2-small.toml           | 1 +
 .../models/inference-net/schematron-v2-turbo.toml           | 1 +
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 1 +
 4 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 14ec7ef4a45..cc0ba12eb64 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04788
-output = 0.09576
-cache_read = 0.009576
+input = 0.0476
+output = 0.0952
+cache_read = 0.00952
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inference-net/schematron-v2-small.toml b/providers/openrouter/models/inference-net/schematron-v2-small.toml
index 90109228329..8b5149f5d77 100644
--- a/providers/openrouter/models/inference-net/schematron-v2-small.toml
+++ b/providers/openrouter/models/inference-net/schematron-v2-small.toml
@@ -12,6 +12,7 @@ open_weights = true
 [cost]
 input = 0.05
 output = 0.23
+cache_read = 0.05
 
 [limit]
 context = 128_000
diff --git a/providers/openrouter/models/inference-net/schematron-v2-turbo.toml b/providers/openrouter/models/inference-net/schematron-v2-turbo.toml
index 1ecb2bb230d..e5c154e9b76 100644
--- a/providers/openrouter/models/inference-net/schematron-v2-turbo.toml
+++ b/providers/openrouter/models/inference-net/schematron-v2-turbo.toml
@@ -12,6 +12,7 @@ open_weights = true
 [cost]
 input = 0.03
 output = 0.15
+cache_read = 0.03
 
 [limit]
 context = 128_000
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 9b28cb1e8f4..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -22,6 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 2.1
 output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From b825dd86eb3646987b3c00873bb892548a351b78 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 19:21:28 +0000
Subject: [PATCH 1240/2012] chore(sync): update OpenRouter model catalog
 (#7020)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml         | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index cc0ba12eb64..c5c46f3604a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0476
-output = 0.0952
-cache_read = 0.00952
+input = 0.04732
+output = 0.09464
+cache_read = 0.009464
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index d5918c075a2..c54d27b9d18 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.5
+cache_read = 0.03
 
 [limit]
 context = 1_310_720

From c7c31737794cc02dfe3523371ecc490f2aa8a2ae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 19:21:33 +0000
Subject: [PATCH 1241/2012] chore(sync): update Kilo model catalog (#7021)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/inference-net/schematron-v2-small.toml | 1 +
 providers/kilo/models/inference-net/schematron-v2-turbo.toml | 1 +
 2 files changed, 2 insertions(+)

diff --git a/providers/kilo/models/inference-net/schematron-v2-small.toml b/providers/kilo/models/inference-net/schematron-v2-small.toml
index b3551182e5f..17606c49e47 100644
--- a/providers/kilo/models/inference-net/schematron-v2-small.toml
+++ b/providers/kilo/models/inference-net/schematron-v2-small.toml
@@ -12,6 +12,7 @@ open_weights = false
 [cost]
 input = 0.05
 output = 0.23
+cache_read = 0.05
 
 [limit]
 context = 128_000
diff --git a/providers/kilo/models/inference-net/schematron-v2-turbo.toml b/providers/kilo/models/inference-net/schematron-v2-turbo.toml
index d75abf2715a..a3ab0173806 100644
--- a/providers/kilo/models/inference-net/schematron-v2-turbo.toml
+++ b/providers/kilo/models/inference-net/schematron-v2-turbo.toml
@@ -12,6 +12,7 @@ open_weights = false
 [cost]
 input = 0.03
 output = 0.15
+cache_read = 0.03
 
 [limit]
 context = 128_000

From df9e6f1af50781d2328e218c3549dc7b98ff7e49 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 20:24:48 +0000
Subject: [PATCH 1242/2012] chore(sync): update Kilo model catalog (#7023)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.3-flash.toml     | 1 +
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 0104ea56ade..c464db465b8 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -12,6 +12,7 @@ cache_read = 0.03
 
 [limit]
 context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index e4ce4e5dfe6..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From eeea0c4d69920f23ddd7a17159bdc2790bf0755c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 20:24:50 +0000
Subject: [PATCH 1243/2012] chore(sync): update OpenRouter model catalog
 (#7022)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c5c46f3604a..664a084d4ad 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04732
-output = 0.09464
-cache_read = 0.009464
+input = 0.04704
+output = 0.09408
+cache_read = 0.009408
 
 [limit]
 context = 1_048_576

From 2285365092a2a557e5e6325fc878f7fde841a01f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 21:22:39 +0000
Subject: [PATCH 1244/2012] chore(sync): update Kilo model catalog (#7024)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.3-flash.toml     | 1 -
 providers/kilo/models/~z-ai/glm-flash-latest.toml | 2 +-
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index c464db465b8..0104ea56ade 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -12,7 +12,6 @@ cache_read = 0.03
 
 [limit]
 context = 1_048_576
-output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..e4ce4e5dfe6 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]

From de6e0789a8b5145a6891aaf585b33d3b8bc6db86 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 21:22:42 +0000
Subject: [PATCH 1245/2012] chore(sync): update OpenRouter model catalog
 (#7025)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 664a084d4ad..e4781a23b60 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04704
-output = 0.09408
-cache_read = 0.009408
+input = 0.04676
+output = 0.09352
+cache_read = 0.009352
 
 [limit]
 context = 1_048_576

From 92394d8974d716cf4292cbfc0aa1c6139399677b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 22:24:25 +0000
Subject: [PATCH 1246/2012] chore(sync): update OpenRouter model catalog
 (#7026)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index e4781a23b60..c95e402fad8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04676
-output = 0.09352
-cache_read = 0.009352
+input = 0.04648
+output = 0.09296
+cache_read = 0.009296
 
 [limit]
 context = 1_048_576

From 7d79f48ad49dcf3c36c6baf54c0778e1fade5bcb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 23:22:35 +0000
Subject: [PATCH 1247/2012] chore(sync): update LLM Gateway model catalog
 (#7027)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../consensusprotocol/deepseek-v4.1-flash.toml  | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..5cd9d0ac1c8
--- /dev/null
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
@@ -0,0 +1,17 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Consensus Protocol)"
+attachment = false
+reasoning = false
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.2
+output = 1
+cache_read = 0.005
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text"]

From a7a66db9196cf51dfbdd33d01ee8796eb80c65b0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 23:22:38 +0000
Subject: [PATCH 1248/2012] chore(sync): update OpenRouter model catalog
 (#7028)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c95e402fad8..8f39ed72a9d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04648
-output = 0.09296
-cache_read = 0.009296
+input = 0.0462
+output = 0.0924
+cache_read = 0.00924
 
 [limit]
 context = 1_048_576

From 1ebf0bd0fbb7ea1e533a5159898a1a63191ebda0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 01:05:04 +0000
Subject: [PATCH 1249/2012] chore(sync): update Kilo model catalog (#7029)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml | 4 ++--
 providers/kilo/models/qwen/qwen3-14b.toml                    | 4 ++--
 providers/kilo/models/z-ai/glm-5.2.toml                      | 3 +--
 providers/kilo/models/z-ai/glm-5.3.toml                      | 3 ++-
 4 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520
diff --git a/providers/kilo/models/qwen/qwen3-14b.toml b/providers/kilo/models/qwen/qwen3-14b.toml
index 67828967f41..2ce8bc46042 100644
--- a/providers/kilo/models/qwen/qwen3-14b.toml
+++ b/providers/kilo/models/qwen/qwen3-14b.toml
@@ -19,8 +19,8 @@ input = 0.2275
 output = 0.91
 
 [limit]
-context = 131_072
-output = 8_192
+context = 40_960
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index d1193203e34..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,5 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 202_752
-output = 182_476
+context = 1_048_576
diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index e9d5bdad891..70fec865ada 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -11,4 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_048_575
+output = 943_717

From c2d15660f434e97fd73483febd4928a70792e8e9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 01:05:07 +0000
Subject: [PATCH 1250/2012] chore(sync): update OpenRouter model catalog
 (#7030)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 providers/openrouter/models/qwen/qwen3-14b.toml           | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 7 +++----
 providers/openrouter/models/z-ai/glm-5.3.toml             | 7 ++++---
 9 files changed, 29 insertions(+), 29 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 8dce0b17a1e..d06dd2cd8d1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.08
-cache_read = 0.008
+input = 0.065
+output = 0.18
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 8f39ed72a9d..57d98266566 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0462
-output = 0.0924
-cache_read = 0.00924
+input = 0.089866
+output = 0.179732
+cache_read = 0.017973
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 3cbd82a5b7c..f4456edae5d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.019316
+input = 1.05468
+output = 3.16404
+cache_read = 0.035156
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520
diff --git a/providers/openrouter/models/qwen/qwen3-14b.toml b/providers/openrouter/models/qwen/qwen3-14b.toml
index b6ce24c73e0..b10030292e3 100644
--- a/providers/openrouter/models/qwen/qwen3-14b.toml
+++ b/providers/openrouter/models/qwen/qwen3-14b.toml
@@ -17,12 +17,12 @@ open_weights = true
 type = "toggle"
 
 [cost]
-input = 0.2275
-output = 0.91
+input = 0.12
+output = 0.24
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index f6e49b1cd7a..3adac002861 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.6
-output = 2
-cache_read = 0.15
+input = 0.6832
+output = 2.1472
+cache_read = 0.12688
 
 [limit]
 context = 1_048_576
-output = 182_476
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index bff4df6e006..d077461d393 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.092
-output = 3.432
-cache_read = 0.2028
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
 context = 1_310_720
+output = 943_717

From bf76d5dd9f5e9a0a338c1eb3280fc9f7c7c065d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 01:41:29 +0000
Subject: [PATCH 1251/2012] chore(sync): update OpenRouter model catalog
 (#7031)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index d06dd2cd8d1..df31c0bbb32 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.016
+input = 0.06
+output = 0.12
+cache_read = 0.012
 
 [limit]
 context = 1_310_720

From 06501753deb2f85a4913efbc8b0b7837108bdacc Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Sun, 13 Sep 2026 20:45:01 -0500
Subject: [PATCH 1252/2012] fix(google-vertex): route gpt-oss MaaS models
 through the OpenAI-compatible endpoint (#7032)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 providers/google-vertex/models/openai/gpt-oss-120b-maas.toml | 4 ++++
 providers/google-vertex/models/openai/gpt-oss-20b-maas.toml  | 4 ++++
 2 files changed, 8 insertions(+)

diff --git a/providers/google-vertex/models/openai/gpt-oss-120b-maas.toml b/providers/google-vertex/models/openai/gpt-oss-120b-maas.toml
index d442d9715bc..d44155f7b7e 100644
--- a/providers/google-vertex/models/openai/gpt-oss-120b-maas.toml
+++ b/providers/google-vertex/models/openai/gpt-oss-120b-maas.toml
@@ -21,3 +21,7 @@ output = 32_768
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
diff --git a/providers/google-vertex/models/openai/gpt-oss-20b-maas.toml b/providers/google-vertex/models/openai/gpt-oss-20b-maas.toml
index 83c4ea25071..a17cc9ba22c 100644
--- a/providers/google-vertex/models/openai/gpt-oss-20b-maas.toml
+++ b/providers/google-vertex/models/openai/gpt-oss-20b-maas.toml
@@ -24,3 +24,7 @@ output = 32_768
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"

From 55bd0a2a2ddb99fbda4ca57f0d9847a88fe24635 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 03:33:19 +0000
Subject: [PATCH 1253/2012] chore(sync): update OpenRouter model catalog
 (#7034)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 57d98266566..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.089866
-output = 0.179732
-cache_read = 0.017973
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index f4456edae5d..451e7c3012c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.05468
-output = 3.16404
-cache_read = 0.035156
+input = 0.9834
+output = 2.9502
+cache_read = 0.03278
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..03afb202432 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.275
+output = 12.74
+cache_read = 0.2639
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 5274a327618..f6e2f7079d5 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.936
-output = 3.168
-cache_read = 0.1872
+input = 0.92
+output = 3.1372
+cache_read = 0.184
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From ac8a9b55ddb51bf2f0334beb2665d364c2cbe3e6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 03:33:21 +0000
Subject: [PATCH 1254/2012] chore(sync): update Kilo model catalog (#7035)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 10 +++++-----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..5d4f2a560e7 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.275
+output = 12.74
+cache_read = 0.2639
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 05cbd588a93..ff325b7d21b 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.936
-output = 3.168
-cache_read = 0.1872
+input = 0.92
+output = 3.1372
+cache_read = 0.184
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 1ecebe789a05f4febe0613e0105d72be60679152 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 04:32:32 +0000
Subject: [PATCH 1255/2012] chore(sync): update OpenRouter model catalog
 (#7038)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 03afb202432..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.275
-output = 12.74
-cache_read = 0.2639
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From c598d97adbac40b21ef28beafbb466bb60377bcc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 04:32:39 +0000
Subject: [PATCH 1256/2012] chore(sync): update Kilo model catalog (#7039)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 5d4f2a560e7..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.275
-output = 12.74
-cache_read = 0.2639
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 4e1648552acf60df6cf1acc365a6d3653f0f4741 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 05:27:39 +0000
Subject: [PATCH 1257/2012] chore(sync): update Eden AI model catalog (#7041)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/qwen/deepseek-v4-flash-0731.toml       |  4 ++--
 .../edenai/models/qwen/deepseek-v4-pro-0813.toml  |  4 ++--
 .../edenai/models/qwen/deepseek-v4.1-flash.toml   | 10 ++++++++++
 .../tensorx/deepseek/deepseek-v4.1-flash.toml     | 15 +++++++++++++++
 4 files changed, 29 insertions(+), 4 deletions(-)
 create mode 100644 providers/edenai/models/qwen/deepseek-v4.1-flash.toml
 create mode 100644 providers/edenai/models/tensorx/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index 4c0fef2aa45..c736c35eae8 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index b3756adb444..1846852dc15 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
+input = 1.122
+output = 3.366
diff --git a/providers/edenai/models/qwen/deepseek-v4.1-flash.toml b/providers/edenai/models/qwen/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..5df514cc7cd
--- /dev/null
+++ b/providers/edenai/models/qwen/deepseek-v4.1-flash.toml
@@ -0,0 +1,10 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Alibaba)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
diff --git a/providers/edenai/models/tensorx/deepseek/deepseek-v4.1-flash.toml b/providers/edenai/models/tensorx/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..add18f0e36c
--- /dev/null
+++ b/providers/edenai/models/tensorx/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (TensorX)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.5
+output = 1.5
+cache_read = 0.125
+
+[limit]
+context = 1_048_576

From ff90fff47eb82c4764c218a653e0819723a0df7a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 06:51:20 +0000
Subject: [PATCH 1258/2012] chore(sync): update OpenRouter model catalog
 (#7044)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576

From eb807e9a68bc6f52cc755262a4cf9ff024abf056 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 09:34:51 +0000
Subject: [PATCH 1259/2012] chore(sync): update LLM Gateway model catalog
 (#7047)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/deepseek-v4.1-flash.toml       | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/alibaba/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/alibaba/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/alibaba/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..47bdbfcba40
--- /dev/null
+++ b/providers/llmgateway-providers/models/alibaba/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Alibaba Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.03
+
+[limit]
+output = 393_216

From a3bd3dc4a8d8b1964b8bf523fec800655d604188 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 09:34:55 +0000
Subject: [PATCH 1260/2012] chore(sync): update Deep Infra model catalog
 (#7048)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/Qwen/Qwen3.8-Flash.toml | 7 +++++++
 1 file changed, 7 insertions(+)
 create mode 100644 providers/deepinfra/models/Qwen/Qwen3.8-Flash.toml

diff --git a/providers/deepinfra/models/Qwen/Qwen3.8-Flash.toml b/providers/deepinfra/models/Qwen/Qwen3.8-Flash.toml
new file mode 100644
index 00000000000..914a0e258fb
--- /dev/null
+++ b/providers/deepinfra/models/Qwen/Qwen3.8-Flash.toml
@@ -0,0 +1,7 @@
+base_model = "alibaba/qwen3.8-flash"
+reasoning = false
+
+[cost]
+input = 0.113
+output = 0.382
+cache_read = 0.0141

From e60b55c081b3352bad7b84a229cfa388f7af98af Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 10:27:21 +0000
Subject: [PATCH 1261/2012] chore(sync): update OpenRouter model catalog
 (#7051)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576

From 71f1a88cbe956f923200222876b33ccdb4888d93 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 12:38:30 +0000
Subject: [PATCH 1262/2012] chore(sync): update OpenRouter model catalog
 (#7053)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nvidia/nemotron-3-super-120b-a12b.toml         | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
index 455636bfb0a..cc742eb161c 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -14,8 +14,8 @@ values = ["low", "medium"]
 type = "budget_tokens"
 
 [cost]
-input = 0.085
-output = 0.4
+input = 0.08
+output = 0.45
 
 [limit]
-output = 16_384
+output = 235_929
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index c1e3b255d01..12c99d28b38 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.0352
-output = 0.1056
-cache_read = 0.00112
+input = 0.04
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 393_216
 
 [modalities]
 input = ["text"]

From ea790283c30043e6b29aa5241362d4ab7bad54f2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 12:38:33 +0000
Subject: [PATCH 1263/2012] chore(sync): update Kilo model catalog (#7054)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/nvidia/nemotron-3-super-120b-a12b.toml    | 2 +-
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 8 ++++----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
index a591c13e651..f1b0bbb6bfd 100644
--- a/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-super-120b-a12b.toml
@@ -10,4 +10,4 @@ input = 0.08
 output = 0.45
 
 [limit]
-output = 16_384
+output = 235_929
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 67ee14fe333..b586ef3f991 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.0352
-output = 0.1056
-cache_read = 0.00112
+input = 0.04
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 393_216
 
 [modalities]
 input = ["text"]

From 9df518aae587f231d955a3f849306082c543aa4d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 12:38:38 +0000
Subject: [PATCH 1264/2012] chore(sync): update Eden AI model catalog (#7055)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/qwen/deepseek-v4-flash-0731.toml | 5 +++--
 providers/edenai/models/qwen/deepseek-v4-pro-0813.toml   | 5 +++--
 providers/edenai/models/qwen/deepseek-v4.1-flash.toml    | 5 +++--
 3 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
index c736c35eae8..8bebf930c58 100644
--- a/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-flash-0731.toml
@@ -6,5 +6,6 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.352
-output = 1.056
+input = 0.22
+output = 0.66
+cache_read = 0.022
diff --git a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
index 1846852dc15..f8e6a947c22 100644
--- a/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/qwen/deepseek-v4-pro-0813.toml
@@ -6,5 +6,6 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.122
-output = 3.366
+input = 0.66
+output = 1.98
+cache_read = 0.066
diff --git a/providers/edenai/models/qwen/deepseek-v4.1-flash.toml b/providers/edenai/models/qwen/deepseek-v4.1-flash.toml
index 5df514cc7cd..f6b4321f953 100644
--- a/providers/edenai/models/qwen/deepseek-v4.1-flash.toml
+++ b/providers/edenai/models/qwen/deepseek-v4.1-flash.toml
@@ -6,5 +6,6 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
+input = 0.15
+output = 0.6
+cache_read = 0.015

From 68f0813d4cfb7ba8d57b81faf5c14c89e239827c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 13:28:12 +0000
Subject: [PATCH 1265/2012] chore(sync): update Charm Hyper model catalog
 (#7059)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-flash.toml   |  2 +-
 providers/hyper/models/deepseek-v4-pro.toml     |  2 +-
 providers/hyper/models/deepseek-v4.1-flash.toml | 11 +++++------
 providers/hyper/models/gemma-4-26b-a4b-it.toml  |  6 +++---
 providers/hyper/models/glm-5.1.toml             |  2 +-
 providers/hyper/models/glm-5.toml               |  6 +++---
 providers/hyper/models/gpt-oss-120b.toml        |  6 +++---
 providers/hyper/models/kimi-k2.5.toml           |  6 +++---
 providers/hyper/models/minimax-m2.7.toml        |  6 +++---
 9 files changed, 23 insertions(+), 24 deletions(-)

diff --git a/providers/hyper/models/deepseek-v4-flash.toml b/providers/hyper/models/deepseek-v4-flash.toml
index 0f594684ce7..202006bffd6 100644
--- a/providers/hyper/models/deepseek-v4-flash.toml
+++ b/providers/hyper/models/deepseek-v4-flash.toml
@@ -9,4 +9,4 @@ values = ["high", "xhigh"]
 [cost]
 input = 0.2
 output = 0.4
-cache_read = 0.04
+cache_write = 0.04
diff --git a/providers/hyper/models/deepseek-v4-pro.toml b/providers/hyper/models/deepseek-v4-pro.toml
index 165e9d2ed48..b69442283b6 100644
--- a/providers/hyper/models/deepseek-v4-pro.toml
+++ b/providers/hyper/models/deepseek-v4-pro.toml
@@ -9,4 +9,4 @@ values = ["high", "xhigh"]
 [cost]
 input = 2.4
 output = 4.8
-cache_read = 0.2
+cache_write = 0.2
diff --git a/providers/hyper/models/deepseek-v4.1-flash.toml b/providers/hyper/models/deepseek-v4.1-flash.toml
index f98b45365b0..c67cd095316 100644
--- a/providers/hyper/models/deepseek-v4.1-flash.toml
+++ b/providers/hyper/models/deepseek-v4.1-flash.toml
@@ -2,13 +2,12 @@ base_model = "deepseek/deepseek-v4.1-flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "xhigh"]
+values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.03
+input = 0.32664
+output = 1.30656
+cache_read = 0.032664
 
 [limit]
-context = 1_048_576
-output = 26_214
+output = 13_107
diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 6c033a0804a..2ef2f0a55c3 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.102
-output = 0.356
-cache_read = 0.051
+input = 0.1
+output = 0.374
+cache_read = 0.05
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index 2440add2e43..e102da3b5ad 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -5,7 +5,7 @@ reasoning_options = []
 
 [cost]
 input = 1.318
-output = 4.308
+output = 4.268
 cache_read = 0.659
 
 [limit]
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 9882a80de94..e92beec4670 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.86
-output = 2.752
-cache_read = 0.43
+input = 0.94
+output = 3.008
+cache_read = 0.47
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index d1529db1b94..22c22acac3a 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.168
-output = 0.66
-cache_read = 0.084
+input = 0.178
+output = 0.68
+cache_read = 0.089
 
 [limit]
 context = 128_072
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index 59545971994..5a60cf903a0 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5584
-output = 2.935
-cache_read = 0.2792
+input = 0.5344
+output = 2.815
+cache_read = 0.2672
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 1489a8dde84..13b1f1bd0ed 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.396
-output = 1.464
-cache_read = 0.198
+input = 0.462
+output = 1.728
+cache_read = 0.231
 
 [limit]
 context = 262_100

From 73988a657423444f0a74c5d003319f654f1df09e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 13:28:15 +0000
Subject: [PATCH 1266/2012] chore(sync): update OpenRouter model catalog
 (#7058)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576

From 4bdc55d6be8ed440ef0707b9b57f349d904156c7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 14:29:37 +0000
Subject: [PATCH 1267/2012] chore(sync): update OpenRouter model catalog
 (#7063)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576

From 734e56a1f9c67be202870f6ef2486414ab3df205 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 14:29:42 +0000
Subject: [PATCH 1268/2012] chore(sync): update Charm Hyper model catalog
 (#7062)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-pro-0813.toml | 3 +--
 providers/hyper/models/kimi-k2.7-code.toml       | 2 +-
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/hyper/models/deepseek-v4-pro-0813.toml b/providers/hyper/models/deepseek-v4-pro-0813.toml
index 3ef1ae8a3a4..0dcacb83cab 100644
--- a/providers/hyper/models/deepseek-v4-pro-0813.toml
+++ b/providers/hyper/models/deepseek-v4-pro-0813.toml
@@ -4,7 +4,7 @@ last_updated = "2026-08-14"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high", "max"]
+values = ["low", "high", "max"]
 
 [cost]
 input = 1.437216
@@ -12,5 +12,4 @@ output = 4.311648
 cache_read = 0.047907
 
 [limit]
-context = 1_048_576
 output = 262_144
diff --git a/providers/hyper/models/kimi-k2.7-code.toml b/providers/hyper/models/kimi-k2.7-code.toml
index b94d5b3888c..ef24433c661 100644
--- a/providers/hyper/models/kimi-k2.7-code.toml
+++ b/providers/hyper/models/kimi-k2.7-code.toml
@@ -9,7 +9,7 @@ output = 4.3552
 cache_read = 0.206872
 
 [limit]
-context = 262_000
+context = 256_000
 output = 16_000
 
 [modalities]

From 8d8be02fe2132d9efe5643a18b773480ee69acf5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 14:29:45 +0000
Subject: [PATCH 1269/2012] chore(sync): update LLM Gateway model catalog
 (#7064)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/deepseek-v4.1-flash.toml           | 2 +-
 providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
index 5cd9d0ac1c8..d946756de4a 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ structured_output = false
 
 [cost]
 input = 0.2
-output = 1
+output = 0.6
 cache_read = 0.005
 
 [limit]
diff --git a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
index 3a65bed82fd..ef625c449bf 100644
--- a/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
+++ b/providers/llmgateway-providers/models/scx-ai-gp/glm-5.2.toml
@@ -10,6 +10,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.8
+input = 0.88
 output = 2.55
 cache_read = 0.16

From 451d51173687c3d87777990b1a5740b365efd947 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 15:27:50 +0000
Subject: [PATCH 1270/2012] chore(sync): update Charm Hyper model catalog
 (#7067)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4-flash.toml | 2 +-
 providers/hyper/models/deepseek-v4-pro.toml   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/hyper/models/deepseek-v4-flash.toml b/providers/hyper/models/deepseek-v4-flash.toml
index 202006bffd6..0f594684ce7 100644
--- a/providers/hyper/models/deepseek-v4-flash.toml
+++ b/providers/hyper/models/deepseek-v4-flash.toml
@@ -9,4 +9,4 @@ values = ["high", "xhigh"]
 [cost]
 input = 0.2
 output = 0.4
-cache_write = 0.04
+cache_read = 0.04
diff --git a/providers/hyper/models/deepseek-v4-pro.toml b/providers/hyper/models/deepseek-v4-pro.toml
index b69442283b6..165e9d2ed48 100644
--- a/providers/hyper/models/deepseek-v4-pro.toml
+++ b/providers/hyper/models/deepseek-v4-pro.toml
@@ -9,4 +9,4 @@ values = ["high", "xhigh"]
 [cost]
 input = 2.4
 output = 4.8
-cache_write = 0.2
+cache_read = 0.2

From 91a3354e9edf051f3c9cc7440b87dc4b52ca9326 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 15:27:52 +0000
Subject: [PATCH 1271/2012] chore(sync): update Eden AI model catalog (#7068)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index bd3c18bf6a6..6917caeb537 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34776
-output = 0.46368
+input = 0.34653
+output = 0.46204
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 6c331de7ca0..cb91c32767e 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.75348
-output = 0.75348
+input = 0.750815
+output = 0.750815
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 9f8263abe1f..06f34e160e0 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17388
-output = 0.75348
+input = 0.173265
+output = 0.750815
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index cedc6b730a8..4b4d3fe1c8b 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46368
-output = 0.92736
+input = 0.46204
+output = 0.92408
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index d2d829b7f23..e18f7060b40 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17388
-output = 0.69552
+input = 0.173265
+output = 0.69306
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index efaf5da1374..4b4a9a94743 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.04328
-output = 1.04328
+input = 1.03959
+output = 1.03959

From bd92c7d284c906e3489ce39b8679935ad74a279f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 15:27:57 +0000
Subject: [PATCH 1272/2012] chore(sync): update Kilo model catalog (#7066)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-4-maverick.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 5c1df37c375..fab1dd0fc03 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -11,12 +11,12 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.2
-output = 0.696
+input = 0.1875
+output = 0.6525
 
 [limit]
 context = 128_000
-output = 115_200
+output = 16_384
 
 [modalities]
 input = ["text", "image"]

From f6e312911cf53054f7d7360cf1795346b84861b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 15:28:00 +0000
Subject: [PATCH 1273/2012] chore(sync): update OpenRouter model catalog
 (#7069)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++--
 .../models/meta-llama/llama-4-maverick.toml   |  6 ++--
 .../~deepseek/deepseek-flash-latest.toml      | 33 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml | 33 +++++++++++++++++++
 5 files changed, 75 insertions(+), 9 deletions(-)
 create mode 100644 providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
 create mode 100644 providers/openrouter/models/~deepseek/deepseek-pro-latest.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 451e7c3012c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9834
-output = 2.9502
-cache_read = 0.03278
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index f4b36c39886..b9b42749c83 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -12,12 +12,12 @@ knowledge = "2024-08-31"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.696
+input = 0.1875
+output = 0.6525
 
 [limit]
 context = 1_048_576
-output = 115_200
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
new file mode 100644
index 00000000000..182582c86a5
--- /dev/null
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -0,0 +1,33 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "DeepSeek Flash Latest"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+family = "deepseek-flash"
+release_date = "2026-09-14"
+last_updated = "2026-09-14"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
new file mode 100644
index 00000000000..e0b3ef90d2e
--- /dev/null
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -0,0 +1,33 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "DeepSeek Pro Latest"
+description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
+family = "deepseek"
+release_date = "2026-09-14"
+last_updated = "2026-09-14"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
+
+[limit]
+context = 1_048_576
+output = 393_216
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 070ae0c93a1eb28da70b7279b0bccf71454e96be Mon Sep 17 00:00:00 2001
From: Angelo Verlain <37999241+vixalien@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:18:49 +0200
Subject: [PATCH 1274/2012] feat(amd): add DeepSeek-V4.1-Flash, Qwen3.8-27B,
 MiniCPM5-2B; drop retired MiniCPM5-1B (#7065)

* feat(amd): refresh the Radeon Cloud catalog

Add DeepSeek-V4.1-Flash, Qwen3.8-27B and MiniCPM5-2B, and drop the retired MiniCPM5-1B entry. Live-verified against https://developer.amd.com.cn/radeon/api/v1 on 2026-09-14: GET /v1/models plus per-model chat completions probes for cost, context, modalities, tools, reasoning_effort enums and output caps.

* fix(amd): address review findings

MiniCPM5-2B: reasoning_options = [] since reasoning_effort is inert on this host (no separated reasoning, reasoning_tokens 0). Qwen3.8-27B: interleaved = true for the choices[0].message.reasoning side channel. DeepSeek-V4.1-Flash and DeepSeek-V4-Flash: override limit.context to the host's 1048576 (lab base is 1_000_000).
---
 models/openbmb/minicpm5-2b.toml               | 26 +++++++++++++++
 providers/amd/models/DeepSeek-V4-Flash.toml   |  5 +++
 providers/amd/models/DeepSeek-V4.1-Flash.toml | 21 ++++++++++++
 providers/amd/models/MiniCPM5-1B.toml         | 17 ----------
 providers/amd/models/MiniCPM5-2B.toml         | 17 ++++++++++
 providers/amd/models/Qwen3.8-27B.toml         | 33 +++++++++++++++++++
 6 files changed, 102 insertions(+), 17 deletions(-)
 create mode 100644 models/openbmb/minicpm5-2b.toml
 create mode 100644 providers/amd/models/DeepSeek-V4.1-Flash.toml
 delete mode 100644 providers/amd/models/MiniCPM5-1B.toml
 create mode 100644 providers/amd/models/MiniCPM5-2B.toml
 create mode 100644 providers/amd/models/Qwen3.8-27B.toml

diff --git a/models/openbmb/minicpm5-2b.toml b/models/openbmb/minicpm5-2b.toml
new file mode 100644
index 00000000000..584fcc99009
--- /dev/null
+++ b/models/openbmb/minicpm5-2b.toml
@@ -0,0 +1,26 @@
+# Sources (accessed 2026-09-14):
+# https://huggingface.co/openbmb/MiniCPM5-2B (model card, config.json)
+# Hub createdAt 2026-09-06, lastModified 2026-09-12; open-weights drop of
+# MiniCPM5-2B with BF16 final release (RL + OPD post-training).
+name = "MiniCPM5-2B"
+description = "Dense 2B-class open-source model for on-device and resource-constrained use, with native long-context support, tool calling, and agentic tasks"
+release_date = "2026-09-06"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+license = "apache-2.0"
+
+[limit]
+context = 131_072
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/openbmb/MiniCPM5-2B"
diff --git a/providers/amd/models/DeepSeek-V4-Flash.toml b/providers/amd/models/DeepSeek-V4-Flash.toml
index 4e3953f9d77..a4953405c08 100644
--- a/providers/amd/models/DeepSeek-V4-Flash.toml
+++ b/providers/amd/models/DeepSeek-V4-Flash.toml
@@ -10,6 +10,11 @@ reasoning_options = [
   { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
 ]
 
+# GET /v1/models reports 1048576 context, higher than the lab entry's
+# 1_000_000, so [limit] overrides it (same host value as DeepSeek-V4.1-Flash).
+[limit]
+context = 1_048_576
+
 [cost]
 input = 0.14
 output = 0.28
diff --git a/providers/amd/models/DeepSeek-V4.1-Flash.toml b/providers/amd/models/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..1bb5632d3ba
--- /dev/null
+++ b/providers/amd/models/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,21 @@
+# Live probes against the AMD Radeon Cloud Model API (accessed 2026-09-14):
+# GET /v1/models reports 1048576 context (higher than the lab entry's
+# 1_000_000, hence the [limit] override), text+image input, tools, reasoning,
+# streaming, and the same per-token pricing as DeepSeek-V4-Flash.
+# reasoning_effort could not be measured on this pool: every attempt returned
+# 503 no_available_workers. The enum below is taken from the sibling
+# DeepSeek-V4-Flash deployment on the same host, which was verified live on
+# 2026-08 as none|minimal|low|medium|high|xhigh|max with no reasoning_content.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [
+  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_048_576
diff --git a/providers/amd/models/MiniCPM5-1B.toml b/providers/amd/models/MiniCPM5-1B.toml
deleted file mode 100644
index 09d14d672f9..00000000000
--- a/providers/amd/models/MiniCPM5-1B.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-# Pricing and reasoning controls verified against the AMD Token Factory API
-# (accessed 2026-09-03): GET /v1/models and live POST /v1/chat/completions.
-# reasoning_effort accepts low|medium|high on this host (minimal, xhigh, and
-# max return 400). none is accepted but does not disable thinking - the model
-# keeps emitting reasoning (as <think> text in the content stream), and no
-# reasoning_content is ever returned, so there is no toggle and no
-# interleaved channel on this host.
-base_model = "openbmb/minicpm5-1b"
-
-reasoning_options = [
-  { type = "effort", values = ["low", "medium", "high"] },
-]
-
-[cost]
-input = 0.124
-output = 0.7425
-cache_read = 0.124
\ No newline at end of file
diff --git a/providers/amd/models/MiniCPM5-2B.toml b/providers/amd/models/MiniCPM5-2B.toml
new file mode 100644
index 00000000000..db899103c54
--- /dev/null
+++ b/providers/amd/models/MiniCPM5-2B.toml
@@ -0,0 +1,17 @@
+# Live probes against the AMD Radeon Cloud Model API (accessed 2026-09-14):
+# GET /v1/models reports 131072 context, text-only input, tools, and
+# reasoning. reasoning_effort is accepted as low|medium|high (none / minimal /
+# xhigh / max return 422) but is inert on this host: no separated reasoning is
+# ever returned and reasoning_tokens stays 0, so the model answers inline in
+# content. No caller control over reasoning is published.
+# Tools and json_object both work; image input is refused with 400.
+# Pricing from GET /v1/models: input $0.124, output $0.7425, cache read
+# $0.124 per MTok.
+base_model = "openbmb/minicpm5-2b"
+
+reasoning_options = []
+
+[cost]
+input = 0.124
+output = 0.7425
+cache_read = 0.124
diff --git a/providers/amd/models/Qwen3.8-27B.toml b/providers/amd/models/Qwen3.8-27B.toml
new file mode 100644
index 00000000000..9e89ddf71bf
--- /dev/null
+++ b/providers/amd/models/Qwen3.8-27B.toml
@@ -0,0 +1,33 @@
+# Live probes against the AMD Radeon Cloud Model API (accessed 2026-09-14):
+# GET /v1/models reports 131072 context, text-only input (no vision on this
+# host), tools, reasoning, and free pricing (free: true).
+# reasoning_effort: the gateway validates the enum low|medium|high and rejects
+# none / minimal / xhigh / max with 422, but the routing backend only supports
+# xhigh (default) | medium | low and rejects "high" with 400. Usable
+# end-to-end values are therefore low and medium. With reasoning_effort
+# omitted the model thinks anyway (backend default tier).
+base_model = "alibaba/qwen3.8-27b"
+
+# AMD hosts a text-only deployment of this model: an image content part is
+# refused with 400 "does not support image input", so the multimodal base
+# entry (text + image + video) is narrowed here.
+attachment = false
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium"] },
+]
+
+# Thinking is returned on choices[0].message.reasoning, a non-content field.
+# The schema's interleaved field enum only covers reasoning_content /
+# reasoning_details, so it is set bare.
+interleaved = true
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 131_072
+
+[modalities]
+input = ["text"]

From 1c1df315c7b72cc6d8303834cb81c9ff31538831 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 11:19:06 -0500
Subject: [PATCH 1275/2012] fix: vercel muse-spark contributor models have
 empty reasoning_options, hiding /variants in OpenCode (#7061)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../vercel/models/meta/muse-spark-1.2-contributor.toml    | 8 +++++++-
 .../vercel/models/meta/muse-spark-1.3-contributor.toml    | 8 +++++++-
 2 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/providers/vercel/models/meta/muse-spark-1.2-contributor.toml b/providers/vercel/models/meta/muse-spark-1.2-contributor.toml
index d2a048e3964..e3d37635419 100644
--- a/providers/vercel/models/meta/muse-spark-1.2-contributor.toml
+++ b/providers/vercel/models/meta/muse-spark-1.2-contributor.toml
@@ -1,3 +1,9 @@
+# Effort: reasoning_effort / reasoning.effort = minimal|low|medium|high|xhigh
+# Empty [] was incorrect: Muse Spark exposes graded effort; OpenCode treats [] as no variants.
+# max is not available on Contributor / 1.2 tiers per Meta reasoning docs.
+# https://ai.developer.meta.com/docs/reasoning
+# https://dev.meta.ai/docs/reasoning
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
 name = "Muse Spark 1.2 Contributor"
 description = "Open Llama multimodal model for image understanding and text reasoning"
 family = "muse"
@@ -7,7 +13,7 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 0.1
diff --git a/providers/vercel/models/meta/muse-spark-1.3-contributor.toml b/providers/vercel/models/meta/muse-spark-1.3-contributor.toml
index 2decd48753b..4607838838e 100644
--- a/providers/vercel/models/meta/muse-spark-1.3-contributor.toml
+++ b/providers/vercel/models/meta/muse-spark-1.3-contributor.toml
@@ -1,3 +1,9 @@
+# Effort: reasoning_effort / reasoning.effort = minimal|low|medium|high|xhigh
+# Empty [] was incorrect: Muse Spark exposes graded effort; OpenCode treats [] as no variants.
+# max is Standard-tier muse-spark-1.3 only (not Contributor) per Meta reasoning docs.
+# https://ai.developer.meta.com/docs/reasoning
+# https://dev.meta.ai/docs/reasoning
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
 name = "Muse Spark 1.3 Contributor"
 description = "Open Llama multimodal model for image understanding and text reasoning"
 family = "muse"
@@ -7,7 +13,7 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 0.1

From 9dbd88a4f18218d7c0e7be7513fd9ea1430e96b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 16:27:20 +0000
Subject: [PATCH 1276/2012] chore(sync): update OpenRouter model catalog
 (#7071)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index f6e2f7079d5..d6c9825dd85 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.92
-output = 3.1372
-cache_read = 0.184
+input = 0.91
+output = 2.86
+cache_read = 0.169
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 34b9d951386173af5681e6ddc4eb7091fca46fea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 16:27:27 +0000
Subject: [PATCH 1277/2012] chore(sync): update Weights & Biases model catalog
 (#7072)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml    | 6 +++---
 .../nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B.toml       | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/wandb/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml b/providers/wandb/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml
index b90c5c08a3c..94af899ca19 100644
--- a/providers/wandb/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml
+++ b/providers/wandb/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml
@@ -7,9 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.75
-output = 2.75
-cache_read = 0.15
+input = 0.5
+output = 2.15
+cache_read = 0.1
 
 [limit]
 context = 262_144
diff --git a/providers/wandb/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B.toml b/providers/wandb/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B.toml
index 50a9ad49020..1237a638da9 100644
--- a/providers/wandb/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B.toml
+++ b/providers/wandb/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B.toml
@@ -6,6 +6,6 @@ description = "Nemotron 3.5 Lightning is an MoE model built for fast, reliable a
 type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.25
-cache_read = 0.05
+input = 0.07
+output = 0.2
+cache_read = 0.04

From 4945a762fa0805179e043fa68893be7f6899758f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 16:27:30 +0000
Subject: [PATCH 1278/2012] chore(sync): update Kilo model catalog (#7070)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  3 +-
 providers/kilo/models/tencent/hy3.toml        |  6 ++--
 .../~deepseek/deepseek-flash-latest.toml      | 28 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml | 28 +++++++++++++++++++
 providers/kilo/models/~z-ai/glm-latest.toml   |  8 +++---
 5 files changed, 65 insertions(+), 8 deletions(-)
 create mode 100644 providers/kilo/models/~deepseek/deepseek-flash-latest.toml
 create mode 100644 providers/kilo/models/~deepseek/deepseek-pro-latest.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
new file mode 100644
index 00000000000..51ee78461f9
--- /dev/null
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek: DeepSeek Flash Latest"
+description = "This model always redirects to the latest model in the DeepSeek Flash family."
+family = "deepseek-flash"
+release_date = "2026-09-14"
+last_updated = "2026-09-14"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
+
+[limit]
+context = 1_048_576
+output = 943_718
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
new file mode 100644
index 00000000000..bd4537b7471
--- /dev/null
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek: DeepSeek Pro Latest"
+description = "This model always redirects to the latest model in the DeepSeek Pro family."
+family = "deepseek"
+release_date = "2026-09-14"
+last_updated = "2026-09-14"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
+
+[limit]
+context = 1_000_000
+output = 393_216
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index ff325b7d21b..53a08317945 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.92
-output = 3.1372
-cache_read = 0.184
+input = 0.91
+output = 2.86
+cache_read = 0.169
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From b7ac42d630d94115484dbf07de2d9cb0bc5c1819 Mon Sep 17 00:00:00 2001
From: PixelDrifter <wo.pandong@foxmail.com>
Date: Tue, 15 Sep 2026 00:28:40 +0800
Subject: [PATCH 1279/2012] fix(kimi-for-coding): update metadata for K2.8
 Preview (#7036)

---
 models/moonshotai/kimi-k2.8-preview.toml       | 18 ++++++++++++++++++
 .../models/kimi-for-coding.toml                | 15 +++++++++++++--
 2 files changed, 31 insertions(+), 2 deletions(-)
 create mode 100644 models/moonshotai/kimi-k2.8-preview.toml

diff --git a/models/moonshotai/kimi-k2.8-preview.toml b/models/moonshotai/kimi-k2.8-preview.toml
new file mode 100644
index 00000000000..358b443e6d5
--- /dev/null
+++ b/models/moonshotai/kimi-k2.8-preview.toml
@@ -0,0 +1,18 @@
+# https://www.kimi.com/code/docs/en/kimi-code/models.html
+# https://www.kimi.com/code/docs/en/kimi-code/whats-new.html#k2-8-preview-september-11-2026
+name = "Kimi K2.8 Preview"
+description = "Kimi coding model with more efficient thinking and up to 1M context, available through Kimi Code"
+family = "kimi-k2"
+release_date = "2026-09-11"
+last_updated = "2026-09-11"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/kimi-for-coding/models/kimi-for-coding.toml b/providers/kimi-for-coding/models/kimi-for-coding.toml
index c6486dd0500..2dc81d0d56d 100644
--- a/providers/kimi-for-coding/models/kimi-for-coding.toml
+++ b/providers/kimi-for-coding/models/kimi-for-coding.toml
@@ -1,5 +1,16 @@
-base_model = "moonshotai/kimi-k2.7-code"
-reasoning_options = []
+# https://www.kimi.com/code/docs/en/third-party-tools/opencode.html
+# Toggle: thinking.type = "enabled" | "disabled" | "adaptive"
+# Effort: output_config.effort = "low" | "high" | "max" (default: "max")
+# Retain the existing output limit; the K2.8 announcement only specifies context.
+base_model = "moonshotai/kimi-k2.8-preview"
+name = "kimi-for-coding"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
 input = 0

From 9857c3a181e08aa8322ebcd1650e64690970aef8 Mon Sep 17 00:00:00 2001
From: netphils <100473503+netphils@users.noreply.github.com>
Date: Tue, 15 Sep 2026 00:29:54 +0800
Subject: [PATCH 1280/2012] feat(alibaba-token-plan-cn): add DeepSeek V4.1
 Flash (#7045)

---
 .../models/deepseek-v4.1-flash.toml           | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/alibaba-token-plan-cn/models/deepseek-v4.1-flash.toml

diff --git a/providers/alibaba-token-plan-cn/models/deepseek-v4.1-flash.toml b/providers/alibaba-token-plan-cn/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..7bb457f885e
--- /dev/null
+++ b/providers/alibaba-token-plan-cn/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,20 @@
+# Reasoning HTTP format (accessed 2026-09-14):
+# Toggle: enable_thinking = true|false. Chat Completions `reasoning_effort`
+# accepts low, high (default), and max; medium and xhigh map to high.
+# Reasoning streams on `reasoning_content`.
+# Model availability: Token Plan model list (China, Beijing region only).
+# Sources:
+# https://platform.qianwenai.com/docs/api-reference/chat/openai-chat
+# https://help.aliyun.com/zh/model-studio/token-plan-personal-overview
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 4cd76e0947a18a9ef466eccff6c60fd3731f5a9e Mon Sep 17 00:00:00 2001
From: JI4JUN <ji4jun0097@gmail.com>
Date: Tue, 15 Sep 2026 00:30:08 +0800
Subject: [PATCH 1281/2012] feat(302ai): add deepseek-flash
 (DeepSeek-V4.1-Flash) (#7033)

302.AI serves DeepSeek-V4.1-Flash as `deepseek-flash`. Cost from the live
catalog ($0.15/$0.60 per MTok, $0.003 cache-read). Reasoning controls verified
live: thinking.type toggle suppresses reasoning_content; effort is validated
(low|high|max, matching DeepSeek first-party mapping); reasoning_content side
channel confirmed.
---
 providers/302ai/models/deepseek-flash.toml | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/302ai/models/deepseek-flash.toml

diff --git a/providers/302ai/models/deepseek-flash.toml b/providers/302ai/models/deepseek-flash.toml
new file mode 100644
index 00000000000..0e607680e15
--- /dev/null
+++ b/providers/302ai/models/deepseek-flash.toml
@@ -0,0 +1,19 @@
+# DeepSeek-V4.1-Flash on 302.AI, served as `deepseek-flash` (catalog SKU created 2026-09-08).
+# Cost: 302.AI catalog, $0.150 / $0.600 per MTok with $0.003 cache-read, https://api.302.ai/v1/models (retrieved 2026-09-14).
+# Toggle and `reasoning_content` side channel verified live against POST https://api.302.ai/v1/chat/completions (2026-09-14):
+# `thinking.type = disabled` suppresses reasoning output; invalid `reasoning_effort` values are rejected (-10003).
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003

From 022f4c3e43b4ca1621bdca03a767ac487c9b5d9a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 11:31:01 -0500
Subject: [PATCH 1282/2012] chore(sync): update Cortecs model catalog (#7049)

* chore(sync): update Cortecs model catalog

* fix(cortecs): add DeepSeek V4.1 Flash reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/cortecs/models/codestral-2508.toml      |  6 +++---
 providers/cortecs/models/deepseek-v4.1-flash.toml | 13 +++++++++++++
 providers/cortecs/models/ministral-14b-2512.toml  |  4 ++--
 providers/cortecs/models/ministral-3b-2512.toml   |  6 +++---
 providers/cortecs/models/ministral-8b-2512.toml   |  4 ++--
 providers/cortecs/models/mistral-large-2512.toml  |  6 +++---
 providers/cortecs/models/mistral-medium-3.5.toml  |  6 +++---
 providers/cortecs/models/mistral-small-2603.toml  |  6 +++---
 providers/cortecs/models/voxtral-small-2507.toml  |  6 +++---
 9 files changed, 35 insertions(+), 22 deletions(-)
 create mode 100644 providers/cortecs/models/deepseek-v4.1-flash.toml

diff --git a/providers/cortecs/models/codestral-2508.toml b/providers/cortecs/models/codestral-2508.toml
index f9547b3dd4e..844a12bf1a9 100644
--- a/providers/cortecs/models/codestral-2508.toml
+++ b/providers/cortecs/models/codestral-2508.toml
@@ -12,9 +12,9 @@ knowledge = "2025-03"
 open_weights = true
 
 [cost]
-input = 0.334
-output = 1.003
-cache_read = 0.033
+input = 0.368
+output = 1.103
+cache_read = 0.037
 
 [limit]
 context = 256_000
diff --git a/providers/cortecs/models/deepseek-v4.1-flash.toml b/providers/cortecs/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..1a21e3f6a14
--- /dev/null
+++ b/providers/cortecs/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning_effort = low|medium|high
+# https://docs.cortecs.ai/routing/reasoning
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.5
+output = 1.499
+cache_read = 0.13
+
+[limit]
+context = 1_048_576
+output = 64_000
diff --git a/providers/cortecs/models/ministral-14b-2512.toml b/providers/cortecs/models/ministral-14b-2512.toml
index e57017e1d77..6b79c2b3c20 100644
--- a/providers/cortecs/models/ministral-14b-2512.toml
+++ b/providers/cortecs/models/ministral-14b-2512.toml
@@ -10,8 +10,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.223
-output = 0.223
+input = 0.24
+output = 0.24
 cache_read = 0.022
 
 [limit]
diff --git a/providers/cortecs/models/ministral-3b-2512.toml b/providers/cortecs/models/ministral-3b-2512.toml
index 4e686e35de2..036dc387677 100644
--- a/providers/cortecs/models/ministral-3b-2512.toml
+++ b/providers/cortecs/models/ministral-3b-2512.toml
@@ -10,9 +10,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.111
-output = 0.111
-cache_read = 0.011
+input = 0.123
+output = 0.123
+cache_read = 0.012
 
 [limit]
 context = 256_000
diff --git a/providers/cortecs/models/ministral-8b-2512.toml b/providers/cortecs/models/ministral-8b-2512.toml
index ddbd9624484..429f3dd2215 100644
--- a/providers/cortecs/models/ministral-8b-2512.toml
+++ b/providers/cortecs/models/ministral-8b-2512.toml
@@ -10,8 +10,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.167
-output = 0.167
+input = 0.179
+output = 0.179
 cache_read = 0.017
 
 [limit]
diff --git a/providers/cortecs/models/mistral-large-2512.toml b/providers/cortecs/models/mistral-large-2512.toml
index 52b4e581a25..cbca6a6506d 100644
--- a/providers/cortecs/models/mistral-large-2512.toml
+++ b/providers/cortecs/models/mistral-large-2512.toml
@@ -2,9 +2,9 @@ base_model = "mistral/mistral-large-2512"
 structured_output = true
 
 [cost]
-input = 0.557
-output = 1.671
-cache_read = 0.056
+input = 0.613
+output = 1.838
+cache_read = 0.061
 
 [limit]
 context = 256_000
diff --git a/providers/cortecs/models/mistral-medium-3.5.toml b/providers/cortecs/models/mistral-medium-3.5.toml
index e814aa7640a..7c463c40224 100644
--- a/providers/cortecs/models/mistral-medium-3.5.toml
+++ b/providers/cortecs/models/mistral-medium-3.5.toml
@@ -11,9 +11,9 @@ open_weights = false
 reasoning_options = []
 
 [cost]
-input = 1.393
-output = 7.13
-cache_read = 0.139
+input = 1.532
+output = 7.843
+cache_read = 0.154
 
 [limit]
 context = 256_000
diff --git a/providers/cortecs/models/mistral-small-2603.toml b/providers/cortecs/models/mistral-small-2603.toml
index 89e30734258..050ea77050d 100644
--- a/providers/cortecs/models/mistral-small-2603.toml
+++ b/providers/cortecs/models/mistral-small-2603.toml
@@ -3,9 +3,9 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 0.143
-output = 0.568
-cache_read = 0.014
+input = 0.156
+output = 0.625
+cache_read = 0.016
 
 [limit]
 context = 262_144
diff --git a/providers/cortecs/models/voxtral-small-2507.toml b/providers/cortecs/models/voxtral-small-2507.toml
index f8ce5732ff8..040a8cd7951 100644
--- a/providers/cortecs/models/voxtral-small-2507.toml
+++ b/providers/cortecs/models/voxtral-small-2507.toml
@@ -10,9 +10,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.111
-output = 0.334
-cache_read = 0.011
+input = 0.123
+output = 0.368
+cache_read = 0.012
 
 [limit]
 context = 32_000

From 35ac8a3ca66d8799dd56b3fed811dc1a956937f4 Mon Sep 17 00:00:00 2001
From: MickeyXin <49945543+fx247562340@users.noreply.github.com>
Date: Tue, 15 Sep 2026 00:35:08 +0800
Subject: [PATCH 1283/2012] feat(vancine): refresh DeepSeek models and GLM
 pricing (#6852)

Co-authored-by: fx247562340 <fx247562340@users.noreply.github.com>
---
 providers/vancine/models/deepseek-flash.toml  | 33 +++++++++++++++++++
 .../models/deepseek-v4-flash-vision-exp.toml  | 28 ----------------
 .../vancine/models/deepseek-v4-flash.toml     | 27 ---------------
 providers/vancine/models/deepseek-v4-pro.toml | 27 ---------------
 providers/vancine/models/glm-5.3-flash.toml   | 16 ++++++---
 5 files changed, 44 insertions(+), 87 deletions(-)
 create mode 100644 providers/vancine/models/deepseek-flash.toml
 delete mode 100644 providers/vancine/models/deepseek-v4-flash-vision-exp.toml
 delete mode 100644 providers/vancine/models/deepseek-v4-flash.toml
 delete mode 100644 providers/vancine/models/deepseek-v4-pro.toml

diff --git a/providers/vancine/models/deepseek-flash.toml b/providers/vancine/models/deepseek-flash.toml
new file mode 100644
index 00000000000..022e839c94b
--- /dev/null
+++ b/providers/vancine/models/deepseek-flash.toml
@@ -0,0 +1,33 @@
+# Sources (accessed 2026-09-11):
+# https://api-docs.deepseek.com/quick_start/pricing/
+# https://api-docs.deepseek.com/guides/thinking_mode/
+# models/deepseek/deepseek-v4.1-flash.toml
+# providers/deepseek/models/deepseek-flash.toml
+# Vancine serves DeepSeek V4.1 Flash as the platform model id deepseek-flash.
+# Reasoning HTTP format: thinking.type = "enabled" | "disabled";
+# reasoning_effort = "low" | "high" | "max". Flash maps low -> low
+# (unlike Pro). xhigh maps to high. Reasoning tokens are billed at the
+# output rate (no separate chain-of-thought price).
+# Interleaved field: reasoning_content.
+# Lab cost: input 0.15 / output 0.6 / reasoning 0.6 / cache_read 0.003
+# (USD/MTok, DeepSeek list price).
+# Vancine price basis: /api/pricing ModelRatio 0.12, CompletionRatio 4,
+# CacheRatio 0.02 converted to USD/1M tokens as input = model_ratio * 2 =
+# 0.24, output = input * completion_ratio = 0.96, cacheRead = input *
+# cache_ratio = 0.0048.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.24
+output = 0.96
+cache_read = 0.0048
diff --git a/providers/vancine/models/deepseek-v4-flash-vision-exp.toml b/providers/vancine/models/deepseek-v4-flash-vision-exp.toml
deleted file mode 100644
index 9a7df2b3b8c..00000000000
--- a/providers/vancine/models/deepseek-v4-flash-vision-exp.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-# Sources (accessed 2026-08-30):
-# https://api-docs.deepseek.com/quick_start/pricing/
-# https://api-docs.deepseek.com/guides/vision
-# models/deepseek/deepseek-v4-flash-vision-exp.toml
-# providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
-# providers/vancine/models/deepseek-v4-flash.toml
-# Toggle: thinking.type = "enabled" | "disabled".
-# Effort: reasoning_effort = "low" | "high" | "max". Priced the same as
-# DeepSeek V4 Flash; reasoning tokens are billed at the output rate.
-# Interleaved field: reasoning_content.
-# Lab cost: input 0.14 / output 0.28 / reasoning 0.28 / cache_read 0.0028.
-# Vancine public pricing (/api/pricing) is 0.22 / 0.66 / n/a / 0.007 (USD/MTok).
-base_model = "deepseek/deepseek-v4-flash-vision-exp"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
diff --git a/providers/vancine/models/deepseek-v4-flash.toml b/providers/vancine/models/deepseek-v4-flash.toml
deleted file mode 100644
index 29c63395f0b..00000000000
--- a/providers/vancine/models/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-# Sources (accessed 2026-08-02):
-# https://api-docs.deepseek.com/guides/thinking_mode/
-# https://api-docs.deepseek.com/quick_start/pricing/
-# Reasoning HTTP format: thinking.type = "enabled" | "disabled";
-# reasoning_effort = "low" | "high" | "max". Flash maps low->low;
-# xhigh maps to high. Reasoning tokens are billed at the output rate.
-# Interleaved field: reasoning_content.
-# Lab cost: input 0.14 / output 0.28 / reasoning 0.28 / cache_read 0.0028.
-# Vancine public pricing (/api/pricing) is 0.22 / 0.66 / n/a / 0.007 (USD/MTok).
-# Vancine ModelList exposes "deepseek-v4-flash", "deepseek-v4-flash-none",
-# "deepseek-v4-flash-max" (relay/channel/deepseek/constants.go).
-base_model = "deepseek/deepseek-v4-flash"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
diff --git a/providers/vancine/models/deepseek-v4-pro.toml b/providers/vancine/models/deepseek-v4-pro.toml
deleted file mode 100644
index d49d1b23083..00000000000
--- a/providers/vancine/models/deepseek-v4-pro.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-# Sources (accessed 2026-08-12):
-# https://api-docs.deepseek.com/api/create-chat-completion
-# https://api-docs.deepseek.com/guides/thinking_mode/
-# Reasoning HTTP format: thinking.type = "enabled" | "disabled";
-# reasoning_effort = "high" | "max" (Pro does not expose "low"). Reasoning
-# tokens are billed at the output rate.
-# Interleaved field: reasoning_content.
-# Lab cost: input 0.435 / output 0.87 / reasoning 0.87 / cache_read 0.003625.
-# Vancine public pricing (/api/pricing) is 0.66 / 1.98 / n/a / 0.022 (USD/MTok).
-# Vancine ModelList exposes "deepseek-v4-pro", "deepseek-v4-pro-none",
-# "deepseek-v4-pro-max" (relay/channel/deepseek/constants.go).
-base_model = "deepseek/deepseek-v4-pro"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
diff --git a/providers/vancine/models/glm-5.3-flash.toml b/providers/vancine/models/glm-5.3-flash.toml
index 2ee2b64c754..3e8d22c0be1 100644
--- a/providers/vancine/models/glm-5.3-flash.toml
+++ b/providers/vancine/models/glm-5.3-flash.toml
@@ -1,8 +1,14 @@
-# Sources (accessed 2026-08-26):
+# Sources (accessed 2026-09-11):
 # https://docs.bigmodel.cn/cn/guide/capabilities/thinking
 # https://docs.z.ai/guides/overview/pricing
 # Lab cost (no 50% promo): input 0.15 / output 0.50 / cache_read 0.03.
-# Vancine public pricing (/api/pricing) is 0.06 / 0.20 / 0.012 (USD/MTok).
+# The Vancine half-price promotion for GLM-5.3-Flash has ended; the prices
+# below are the regular production rates, not promo rates.
+# Vancine public pricing (/api/pricing, verified 2026-09-11) is
+# ModelRatio 0.06, CompletionRatio 3.333333333333, CacheRatio 0.2, which
+# converts to USD/MTok as input = model_ratio * 2 = 0.12,
+# output = input * completion_ratio = 0.40, cache_read = input * cache_ratio
+# = 0.024, cache_write 0 (no create_cache_ratio configured).
 # Thinking is always enabled, so no toggle.
 base_model = "zhipuai/glm-5.3-flash"
 
@@ -11,6 +17,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.20
-cache_read = 0.012
+input = 0.12
+output = 0.40
+cache_read = 0.024

From cebe352f8c1e3e91cadcfdef89fe7b4259dba7cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 11:37:34 -0500
Subject: [PATCH 1284/2012] chore(sync): update NanoGPT model catalog (#6902)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/TEE/deepseek-v4.1-flash.toml       | 15 ++++++++++
 providers/nano-gpt/models/TEE/qwen3-8b.toml   | 27 ------------------
 .../models/TheDrummer/Cydonia-24B-v4.1.toml   |  2 +-
 .../claude-haiku-4.5.toml}                    |  3 +-
 .../claude-haiku-4.5:thinking.toml}           |  3 +-
 .../claude-opus-4.1.toml}                     |  2 +-
 .../claude-opus-4.1:thinking.toml}            |  0
 .../claude-opus-4.1:thinking:1024.toml}       |  0
 .../claude-opus-4.1:thinking:32768.toml}      |  0
 .../claude-opus-4.1:thinking:8192.toml}       |  0
 .../claude-opus-4.5.toml}                     |  2 +-
 .../claude-opus-4.5:thinking.toml}            |  2 +-
 .../models/anthropic/claude-opus-4.toml       | 24 ++++++++++++++++
 .../anthropic/claude-opus-4:thinking.toml     | 25 +++++++++++++++++
 .../claude-opus-4:thinking:1024.toml          | 25 +++++++++++++++++
 .../claude-opus-4:thinking:32768.toml         | 25 +++++++++++++++++
 .../claude-opus-4:thinking:8192.toml          | 25 +++++++++++++++++
 .../claude-sonnet-4.5.toml}                   |  6 ++--
 .../claude-sonnet-4.5:thinking.toml}          |  5 ++--
 .../models/anthropic/claude-sonnet-4.toml     | 24 ++++++++++++++++
 .../anthropic/claude-sonnet-4:thinking.toml   | 25 +++++++++++++++++
 .../claude-sonnet-4:thinking:1024.toml        | 25 +++++++++++++++++
 .../claude-sonnet-4:thinking:32768.toml       | 25 +++++++++++++++++
 .../claude-sonnet-4:thinking:64000.toml       | 25 +++++++++++++++++
 .../claude-sonnet-4:thinking:8192.toml        | 25 +++++++++++++++++
 providers/nano-gpt/models/brave-pro.toml      | 22 ---------------
 providers/nano-gpt/models/brave-research.toml | 22 ---------------
 providers/nano-gpt/models/brave.toml          | 22 ---------------
 .../models/claude-opus-4-20250514.toml        | 12 --------
 .../models/claude-opus-4-thinking.toml        | 12 --------
 .../models/claude-opus-4-thinking:1024.toml   | 12 --------
 .../models/claude-opus-4-thinking:32768.toml  | 12 --------
 .../models/claude-opus-4-thinking:8192.toml   | 12 --------
 .../models/claude-sonnet-4-20250514.toml      | 12 --------
 .../models/claude-sonnet-4-thinking.toml      | 13 ---------
 .../models/claude-sonnet-4-thinking:1024.toml | 13 ---------
 .../claude-sonnet-4-thinking:32768.toml       | 13 ---------
 .../claude-sonnet-4-thinking:64000.toml       | 13 ---------
 .../models/claude-sonnet-4-thinking:8192.toml | 13 ---------
 .../nano-gpt/models/crofai/greg-2-super.toml  | 23 ---------------
 .../nano-gpt/models/crofai/greg-2-ultra.toml  | 23 ---------------
 .../models/deepseek/deepseek-v4.1-flash.toml  |  4 +--
 .../deepseek-v4.1-flash:thinking.toml         |  4 +--
 providers/nano-gpt/models/exa-answer.toml     | 22 ---------------
 .../inference-net/schematron-v2-small.toml    | 23 +++++++++++++++
 .../inference-net/schematron-v2-turbo.toml    | 23 +++++++++++++++
 .../meta-llama/llama-3.3-70b-instruct.toml    |  1 -
 .../minimax-m2.toml}                          |  0
 .../models/mistral-small-31-24b-instruct.toml | 24 ----------------
 ...all-2505.toml => devstral-small-2505.toml} |  2 +-
 .../mistral-medium-3.5.toml                   |  3 +-
 .../mistral-medium-3.5:thinking.toml          |  3 +-
 ...7.toml => mistral-nemo-instruct-2407.toml} |  4 +--
 .../mistral-small-3.1-24b-instruct.toml       | 24 ++++++++++++++++
 .../mistral-small-3.2-24b-instruct.toml}      |  6 ++--
 .../perplexity-academic-researcher.toml       | 24 ----------------
 providers/nano-gpt/models/qwen-3.6-plus.toml  | 25 -----------------
 .../qwen/Qwen3-235B-A22B-Thinking-2507.toml   | 25 -----------------
 .../nano-gpt/models/{ => qwen}/qwen-long.toml |  4 +--
 .../nano-gpt/models/{ => qwen}/qwen-max.toml  |  0
 .../nano-gpt/models/{ => qwen}/qwen-plus.toml |  9 +-----
 .../models/{ => qwen}/qwen-turbo.toml         |  0
 ...t.toml => qwen2.5-coder-32b-instruct.toml} |  1 -
 ...oml => qwen3-235b-a22b-instruct-2507.toml} |  2 --
 .../qwen3-30b-a3b-instruct-2507.toml          |  3 +-
 .../qwen/{Qwen3-8B.toml => qwen3-8b.toml}     |  3 +-
 .../qwen3-coder-30b-a3b-instruct.toml         |  0
 .../{ => qwen}/qwen3-max-2026-01-23.toml      |  3 +-
 ....toml => qwen3-next-80b-a3b-instruct.toml} |  2 --
 ....toml => qwen3-vl-235b-a22b-instruct.toml} |  1 -
 .../qwen3-vl-235b-a22b-thinking.toml          | 10 +------
 .../models/{ => qwen}/qwen3.5-0.8b.toml       |  0
 .../models/{ => qwen}/qwen3.5-122b-a10b.toml  |  1 -
 .../qwen3.5-122b-a10b:thinking.toml           |  7 +----
 .../models/{ => qwen}/qwen3.5-27b.toml        |  1 -
 .../{ => qwen}/qwen3.5-27b:thinking.toml      |  7 +----
 .../models/{ => qwen}/qwen3.5-2b.toml         |  0
 .../models/{ => qwen}/qwen3.5-35b-a3b.toml    |  1 -
 .../{ => qwen}/qwen3.5-35b-a3b:thinking.toml  |  7 +----
 ...g.toml => qwen3.5-397b-a17b:thinking.toml} |  7 +----
 .../models/{ => qwen}/qwen3.5-4b.toml         |  0
 .../models/{ => qwen}/qwen3.5-flash.toml      |  1 -
 .../{ => qwen}/qwen3.5-flash:thinking.toml    |  7 +----
 .../models/{ => qwen}/qwen3.5-omni-flash.toml |  7 +----
 .../models/{ => qwen}/qwen3.5-omni-plus.toml  |  7 +----
 ...inking.toml => qwen3.5-plus:thinking.toml} |  7 +----
 .../models/{alibaba => qwen}/qwen3.6-27b.toml |  1 -
 .../qwen3.6-27b:thinking.toml                 |  7 +----
 ...n3.6-35B-A3B.toml => qwen3.6-35b-a3b.toml} |  1 -
 ...ing.toml => qwen3.6-35b-a3b:thinking.toml} |  1 -
 .../{alibaba => qwen}/qwen3.6-flash.toml      |  0
 .../{ => qwen}/qwen3.6-max-preview.toml       |  0
 .../nano-gpt/models/qwen/qwen3.6-plus.toml    | 14 ++++++++++
 .../models/{ => qwen}/qwen3.7-flash.toml      |  0
 .../{ => qwen}/qwen3.7-flash:thinking.toml    |  0
 .../models/{ => qwen}/qwen3.7-max.toml        |  1 -
 .../{ => qwen}/qwen3.7-max:thinking.toml      |  1 -
 .../models/{ => qwen}/qwen3.7-plus.toml       |  1 -
 .../{ => qwen}/qwen3.7-plus:thinking.toml     |  1 -
 .../models/qwen/qwen3.8-2.4t-a95b.toml        | 18 ------------
 .../models/qwen/qwen3.8-27b-obliterated.toml  |  2 +-
 .../qwen3.8-27b-obliterated:thinking.toml     |  2 +-
 .../models/{ => qwen}/qwen3.8-27b.toml        |  0
 .../{ => qwen}/qwen3.8-27b:thinking.toml      |  0
 .../{alibaba => qwen}/qwen3.8-flash.toml      |  0
 .../{alibaba => qwen}/qwen3.8-max-0902.toml   |  0
 .../models/{ => qwen}/qwen3.8-max.toml        |  0
 .../{ => qwen}/qwen3.8-max:thinking.toml      |  0
 .../models/qwen25-vl-72b-instruct.toml        | 23 ---------------
 .../nano-gpt/models/sonar-deep-research.toml  | 13 ---------
 providers/nano-gpt/models/sonar-pro.toml      | 16 -----------
 .../nano-gpt/models/sonar-reasoning-pro.toml  | 17 -----------
 providers/nano-gpt/models/sonar.toml          | 13 ---------
 .../models/xiaomi/mimo-v2.5-pro-crof.toml     | 28 -------------------
 .../xiaomi/mimo-v2.5-pro-crof:thinking.toml   | 28 -------------------
 .../models/z-ai/glm-5.3-flash-uncensored.toml |  6 ++--
 116 files changed, 419 insertions(+), 679 deletions(-)
 create mode 100644 providers/nano-gpt/models/TEE/deepseek-v4.1-flash.toml
 delete mode 100644 providers/nano-gpt/models/TEE/qwen3-8b.toml
 rename providers/nano-gpt/models/{claude-haiku-4-5-20251001.toml => anthropic/claude-haiku-4.5.toml} (62%)
 rename providers/nano-gpt/models/{claude-haiku-4-5-20251001-thinking.toml => anthropic/claude-haiku-4.5:thinking.toml} (65%)
 rename providers/nano-gpt/models/{claude-opus-4-1-20250805.toml => anthropic/claude-opus-4.1.toml} (73%)
 rename providers/nano-gpt/models/{claude-opus-4-1-thinking.toml => anthropic/claude-opus-4.1:thinking.toml} (100%)
 rename providers/nano-gpt/models/{claude-opus-4-1-thinking:1024.toml => anthropic/claude-opus-4.1:thinking:1024.toml} (100%)
 rename providers/nano-gpt/models/{claude-opus-4-1-thinking:32768.toml => anthropic/claude-opus-4.1:thinking:32768.toml} (100%)
 rename providers/nano-gpt/models/{claude-opus-4-1-thinking:8192.toml => anthropic/claude-opus-4.1:thinking:8192.toml} (100%)
 rename providers/nano-gpt/models/{claude-opus-4-5-20251101.toml => anthropic/claude-opus-4.5.toml} (80%)
 rename providers/nano-gpt/models/{claude-opus-4-5-20251101:thinking.toml => anthropic/claude-opus-4.5:thinking.toml} (81%)
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-4.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-4:thinking.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-4:thinking:1024.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-4:thinking:32768.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-4:thinking:8192.toml
 rename providers/nano-gpt/models/{claude-sonnet-4-5-20250929.toml => anthropic/claude-sonnet-4.5.toml} (65%)
 rename providers/nano-gpt/models/{claude-sonnet-4-5-20250929-thinking.toml => anthropic/claude-sonnet-4.5:thinking.toml} (69%)
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:1024.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:32768.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:64000.toml
 create mode 100644 providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:8192.toml
 delete mode 100644 providers/nano-gpt/models/brave-pro.toml
 delete mode 100644 providers/nano-gpt/models/brave-research.toml
 delete mode 100644 providers/nano-gpt/models/brave.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-20250514.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-thinking.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-thinking:1024.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-thinking:32768.toml
 delete mode 100644 providers/nano-gpt/models/claude-opus-4-thinking:8192.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-20250514.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-thinking.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-thinking:1024.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-thinking:32768.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-thinking:64000.toml
 delete mode 100644 providers/nano-gpt/models/claude-sonnet-4-thinking:8192.toml
 delete mode 100644 providers/nano-gpt/models/crofai/greg-2-super.toml
 delete mode 100644 providers/nano-gpt/models/crofai/greg-2-ultra.toml
 delete mode 100644 providers/nano-gpt/models/exa-answer.toml
 create mode 100644 providers/nano-gpt/models/inference-net/schematron-v2-small.toml
 create mode 100644 providers/nano-gpt/models/inference-net/schematron-v2-turbo.toml
 rename providers/nano-gpt/models/{MiniMax-M2.toml => minimax/minimax-m2.toml} (100%)
 delete mode 100644 providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
 rename providers/nano-gpt/models/mistralai/{Devstral-Small-2505.toml => devstral-small-2505.toml} (74%)
 rename providers/nano-gpt/models/{mistral => mistralai}/mistral-medium-3.5.toml (61%)
 rename providers/nano-gpt/models/{mistral => mistralai}/mistral-medium-3.5:thinking.toml (74%)
 rename providers/nano-gpt/models/mistralai/{Mistral-Nemo-Instruct-2407.toml => mistral-nemo-instruct-2407.toml} (73%)
 create mode 100644 providers/nano-gpt/models/mistralai/mistral-small-3.1-24b-instruct.toml
 rename providers/nano-gpt/models/{chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml => mistralai/mistral-small-3.2-24b-instruct.toml} (50%)
 delete mode 100644 providers/nano-gpt/models/perplexity-academic-researcher.toml
 delete mode 100644 providers/nano-gpt/models/qwen-3.6-plus.toml
 delete mode 100644 providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
 rename providers/nano-gpt/models/{ => qwen}/qwen-long.toml (69%)
 rename providers/nano-gpt/models/{ => qwen}/qwen-max.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen-plus.toml (62%)
 rename providers/nano-gpt/models/{ => qwen}/qwen-turbo.toml (100%)
 rename providers/nano-gpt/models/qwen/{Qwen2.5-Coder-32B-Instruct.toml => qwen2.5-coder-32b-instruct.toml} (89%)
 rename providers/nano-gpt/models/qwen/{Qwen3-235B-A22B-Instruct-2507.toml => qwen3-235b-a22b-instruct-2507.toml} (80%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3-30b-a3b-instruct-2507.toml (51%)
 rename providers/nano-gpt/models/qwen/{Qwen3-8B.toml => qwen3-8b.toml} (62%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3-coder-30b-a3b-instruct.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3-max-2026-01-23.toml (66%)
 rename providers/nano-gpt/models/qwen/{Qwen3-Next-80B-A3B-Instruct.toml => qwen3-next-80b-a3b-instruct.toml} (82%)
 rename providers/nano-gpt/models/qwen/{Qwen3-VL-235B-A22B-Instruct.toml => qwen3-vl-235b-a22b-instruct.toml} (86%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3-vl-235b-a22b-thinking.toml (56%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-0.8b.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-122b-a10b.toml (90%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-122b-a10b:thinking.toml (71%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-27b.toml (89%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-27b:thinking.toml (70%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-2b.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-35b-a3b.toml (90%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-35b-a3b:thinking.toml (71%)
 rename providers/nano-gpt/models/qwen/{qwen3.5-397b-a17b-thinking.toml => qwen3.5-397b-a17b:thinking.toml} (71%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-4b.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-flash.toml (86%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-flash:thinking.toml (66%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-omni-flash.toml (51%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.5-omni-plus.toml (51%)
 rename providers/nano-gpt/models/qwen/{qwen3.5-plus-thinking.toml => qwen3.5-plus:thinking.toml} (68%)
 rename providers/nano-gpt/models/{alibaba => qwen}/qwen3.6-27b.toml (89%)
 rename providers/nano-gpt/models/{alibaba => qwen}/qwen3.6-27b:thinking.toml (70%)
 rename providers/nano-gpt/models/qwen/{Qwen3.6-35B-A3B.toml => qwen3.6-35b-a3b.toml} (90%)
 rename providers/nano-gpt/models/qwen/{Qwen3.6-35B-A3B:thinking.toml => qwen3.6-35b-a3b:thinking.toml} (90%)
 rename providers/nano-gpt/models/{alibaba => qwen}/qwen3.6-flash.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.6-max-preview.toml (100%)
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.6-plus.toml
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-flash.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-flash:thinking.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-max.toml (87%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-max:thinking.toml (89%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-plus.toml (89%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.7-plus:thinking.toml (90%)
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.8-2.4t-a95b.toml
 rename providers/nano-gpt/models/{ => qwen}/qwen3.8-27b.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.8-27b:thinking.toml (100%)
 rename providers/nano-gpt/models/{alibaba => qwen}/qwen3.8-flash.toml (100%)
 rename providers/nano-gpt/models/{alibaba => qwen}/qwen3.8-max-0902.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.8-max.toml (100%)
 rename providers/nano-gpt/models/{ => qwen}/qwen3.8-max:thinking.toml (100%)
 delete mode 100644 providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
 delete mode 100644 providers/nano-gpt/models/sonar-deep-research.toml
 delete mode 100644 providers/nano-gpt/models/sonar-pro.toml
 delete mode 100644 providers/nano-gpt/models/sonar-reasoning-pro.toml
 delete mode 100644 providers/nano-gpt/models/sonar.toml
 delete mode 100644 providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof.toml
 delete mode 100644 providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof:thinking.toml

diff --git a/providers/nano-gpt/models/TEE/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/TEE/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..1ebcfb3ebac
--- /dev/null
+++ b/providers/nano-gpt/models/TEE/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash TEE"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[cost]
+input = 0.65
+output = 1.45
+cache_read = 0.13
+
+[limit]
+context = 1_048_576
+input = 1_048_576
diff --git a/providers/nano-gpt/models/TEE/qwen3-8b.toml b/providers/nano-gpt/models/TEE/qwen3-8b.toml
deleted file mode 100644
index cf35e5e6ed6..00000000000
--- a/providers/nano-gpt/models/TEE/qwen3-8b.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "Qwen3 8B TEE"
-description = "Qwen3 8B is an open-weight dense language model from Alibaba's Qwen team for efficient dialogue, reasoning, mathematics, coding, and tool use. Running inside a TEE (Trusted Execution Environment), with provider attestation support."
-release_date = "2026-09-04"
-last_updated = "2026-09-04"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 0.11
-output = 0.45
-cache_read = 0.11
-
-[limit]
-context = 40_960
-input = 40_960
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
index 356a664e624..a6cd77b0f78 100644
--- a/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
+++ b/providers/nano-gpt/models/TheDrummer/Cydonia-24B-v4.1.toml
@@ -5,7 +5,7 @@ release_date = "2025-08-19"
 last_updated = "2025-08-19"
 attachment = false
 reasoning = false
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = true
 
diff --git a/providers/nano-gpt/models/claude-haiku-4-5-20251001.toml b/providers/nano-gpt/models/anthropic/claude-haiku-4.5.toml
similarity index 62%
rename from providers/nano-gpt/models/claude-haiku-4-5-20251001.toml
rename to providers/nano-gpt/models/anthropic/claude-haiku-4.5.toml
index 3babc2fb8d9..af25254c598 100644
--- a/providers/nano-gpt/models/claude-haiku-4-5-20251001.toml
+++ b/providers/nano-gpt/models/anthropic/claude-haiku-4.5.toml
@@ -1,4 +1,5 @@
-base_model = "anthropic/claude-haiku-4-5-20251001"
+base_model = "anthropic/claude-haiku-4-5"
+name = "Claude Haiku 4.5"
 reasoning = false
 structured_output = true
 
diff --git a/providers/nano-gpt/models/claude-haiku-4-5-20251001-thinking.toml b/providers/nano-gpt/models/anthropic/claude-haiku-4.5:thinking.toml
similarity index 65%
rename from providers/nano-gpt/models/claude-haiku-4-5-20251001-thinking.toml
rename to providers/nano-gpt/models/anthropic/claude-haiku-4.5:thinking.toml
index 4947ff3afe5..85e2558189d 100644
--- a/providers/nano-gpt/models/claude-haiku-4-5-20251001-thinking.toml
+++ b/providers/nano-gpt/models/anthropic/claude-haiku-4.5:thinking.toml
@@ -1,5 +1,4 @@
-# Not included in subscription
-base_model = "anthropic/claude-haiku-4-5-20251001"
+base_model = "anthropic/claude-haiku-4-5"
 name = "Claude Haiku 4.5 Thinking"
 structured_output = true
 reasoning_options = []
diff --git a/providers/nano-gpt/models/claude-opus-4-1-20250805.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.1.toml
similarity index 73%
rename from providers/nano-gpt/models/claude-opus-4-1-20250805.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.1.toml
index bebe8a38d9e..a84153f3d46 100644
--- a/providers/nano-gpt/models/claude-opus-4-1-20250805.toml
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4.1.toml
@@ -1,4 +1,4 @@
-base_model = "anthropic/claude-opus-4-1-20250805"
+base_model = "anthropic/claude-opus-4-1"
 name = "Claude 4.1 Opus"
 reasoning = false
 structured_output = true
diff --git a/providers/nano-gpt/models/claude-opus-4-1-thinking.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/claude-opus-4-1-thinking.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking.toml
diff --git a/providers/nano-gpt/models/claude-opus-4-1-thinking:1024.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:1024.toml
similarity index 100%
rename from providers/nano-gpt/models/claude-opus-4-1-thinking:1024.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:1024.toml
diff --git a/providers/nano-gpt/models/claude-opus-4-1-thinking:32768.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:32768.toml
similarity index 100%
rename from providers/nano-gpt/models/claude-opus-4-1-thinking:32768.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:32768.toml
diff --git a/providers/nano-gpt/models/claude-opus-4-1-thinking:8192.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:8192.toml
similarity index 100%
rename from providers/nano-gpt/models/claude-opus-4-1-thinking:8192.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.1:thinking:8192.toml
diff --git a/providers/nano-gpt/models/claude-opus-4-5-20251101.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.5.toml
similarity index 80%
rename from providers/nano-gpt/models/claude-opus-4-5-20251101.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.5.toml
index 6e38409f610..fcf82fc8830 100644
--- a/providers/nano-gpt/models/claude-opus-4-5-20251101.toml
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4.5.toml
@@ -1,4 +1,4 @@
-base_model = "anthropic/claude-opus-4-5-20251101"
+base_model = "anthropic/claude-opus-4-5"
 name = "Claude 4.5 Opus"
 structured_output = true
 
diff --git a/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.5:thinking.toml
similarity index 81%
rename from providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml
rename to providers/nano-gpt/models/anthropic/claude-opus-4.5:thinking.toml
index 0cfb293c2f4..4e055168a35 100644
--- a/providers/nano-gpt/models/claude-opus-4-5-20251101:thinking.toml
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4.5:thinking.toml
@@ -1,4 +1,4 @@
-base_model = "anthropic/claude-opus-4-5-20251101"
+base_model = "anthropic/claude-opus-4-5"
 name = "Claude 4.5 Opus Thinking"
 structured_output = true
 
diff --git a/providers/nano-gpt/models/anthropic/claude-opus-4.toml b/providers/nano-gpt/models/anthropic/claude-opus-4.toml
new file mode 100644
index 00000000000..f9b4abae349
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4.toml
@@ -0,0 +1,24 @@
+name = "Claude 4 Opus"
+description = "Claude 4 Opus by Anthropic. The premium version of the new Claude models. A new generation model with improved capabilities, especially on programming and development."
+family = "claude-opus"
+release_date = "2025-08-05"
+last_updated = "2025-08-05"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 15
+output = 75
+cache_read = 1.5
+
+[limit]
+context = 200_000
+input = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-opus-4:thinking.toml b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking.toml
new file mode 100644
index 00000000000..df5fccc1cf6
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Opus Thinking"
+description = "Anthropic's Claude 4 Opus with the ability to show its thinking process step by step."
+family = "claude-opus"
+release_date = "2025-07-15"
+last_updated = "2025-07-15"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 15
+output = 75
+cache_read = 1.5
+
+[limit]
+context = 200_000
+input = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:1024.toml b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:1024.toml
new file mode 100644
index 00000000000..dca21bd93e9
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:1024.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Opus Thinking (1K)"
+description = "Claude 4 Opus with minimal thinking budget (1,024 tokens)."
+family = "claude-opus"
+release_date = "2025-08-05"
+last_updated = "2025-08-05"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 15
+output = 75
+cache_read = 1.5
+
+[limit]
+context = 200_000
+input = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:32768.toml b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:32768.toml
new file mode 100644
index 00000000000..b6e9ef29370
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:32768.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Opus Thinking (32K)"
+description = "Claude 4 Opus with extended thinking budget (32,768 tokens)."
+family = "claude-opus"
+release_date = "2025-08-05"
+last_updated = "2025-08-05"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 15
+output = 75
+cache_read = 1.5
+
+[limit]
+context = 200_000
+input = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:8192.toml b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:8192.toml
new file mode 100644
index 00000000000..2d086d5e1b0
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-4:thinking:8192.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Opus Thinking (8K)"
+description = "Claude 4 Opus with reduced thinking budget (8,192 tokens)."
+family = "claude-opus"
+release_date = "2025-08-05"
+last_updated = "2025-08-05"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 15
+output = 75
+cache_read = 1.5
+
+[limit]
+context = 200_000
+input = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/claude-sonnet-4-5-20250929.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4.5.toml
similarity index 65%
rename from providers/nano-gpt/models/claude-sonnet-4-5-20250929.toml
rename to providers/nano-gpt/models/anthropic/claude-sonnet-4.5.toml
index 8d090e4c6c7..82e28e53689 100644
--- a/providers/nano-gpt/models/claude-sonnet-4-5-20250929.toml
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4.5.toml
@@ -1,4 +1,5 @@
-base_model = "anthropic/claude-sonnet-4-5-20250929"
+base_model = "anthropic/claude-sonnet-4-5"
+name = "Claude Sonnet 4.5"
 structured_output = true
 
 [[reasoning_options]]
@@ -11,5 +12,4 @@ output = 15
 cache_read = 0.3
 
 [limit]
-context = 1_000_000
-input = 1_000_000
+input = 200_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-5-20250929-thinking.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4.5:thinking.toml
similarity index 69%
rename from providers/nano-gpt/models/claude-sonnet-4-5-20250929-thinking.toml
rename to providers/nano-gpt/models/anthropic/claude-sonnet-4.5:thinking.toml
index 9852bd1bef8..8edebd11a2a 100644
--- a/providers/nano-gpt/models/claude-sonnet-4-5-20250929-thinking.toml
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4.5:thinking.toml
@@ -1,4 +1,4 @@
-base_model = "anthropic/claude-sonnet-4-5-20250929"
+base_model = "anthropic/claude-sonnet-4-5"
 name = "Claude Sonnet 4.5 Thinking"
 structured_output = true
 
@@ -12,5 +12,4 @@ output = 15
 cache_read = 0.3
 
 [limit]
-context = 1_000_000
-input = 1_000_000
+input = 200_000
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4.toml
new file mode 100644
index 00000000000..f79dbf29d5f
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4.toml
@@ -0,0 +1,24 @@
+name = "Claude 4 Sonnet"
+description = "Claude 4 Sonnet by Anthropic. A new generation model with improved capabilities, especially on programming and development. NOTE: Inputs > 200k tokens are charged at 2x input, 1.5x output rate."
+family = "claude-sonnet"
+release_date = "2025-09-29"
+last_updated = "2025-09-29"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 200_000
+input = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking.toml
new file mode 100644
index 00000000000..56793324fca
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Sonnet Thinking"
+description = "Anthropic's Claude 4 Sonnet with the ability to show its thinking process step by step."
+family = "claude-sonnet"
+release_date = "2025-02-24"
+last_updated = "2025-02-24"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:1024.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:1024.toml
new file mode 100644
index 00000000000..084f12aa861
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:1024.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Sonnet Thinking (1K)"
+description = "Claude 4 Sonnet with minimal thinking budget (1,024 tokens)."
+family = "claude-sonnet"
+release_date = "2024-01-01"
+last_updated = "2024-01-01"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:32768.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:32768.toml
new file mode 100644
index 00000000000..2a63f438c9b
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:32768.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Sonnet Thinking (32K)"
+description = "Claude 4 Sonnet with extended thinking budget (32,768 tokens)."
+family = "claude-sonnet"
+release_date = "2024-01-01"
+last_updated = "2024-01-01"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:64000.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:64000.toml
new file mode 100644
index 00000000000..12270e79c75
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:64000.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Sonnet Thinking (64K)"
+description = "Claude 4 Sonnet with maximum thinking budget (64,000 tokens)."
+family = "claude-sonnet"
+release_date = "2024-01-01"
+last_updated = "2024-01-01"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:8192.toml b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:8192.toml
new file mode 100644
index 00000000000..c48ed8b14c9
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-sonnet-4:thinking:8192.toml
@@ -0,0 +1,25 @@
+name = "Claude 4 Sonnet Thinking (8K)"
+description = "Claude 4 Sonnet with reduced thinking budget (8,192 tokens)."
+family = "claude-sonnet"
+release_date = "2024-01-01"
+last_updated = "2024-01-01"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/brave-pro.toml b/providers/nano-gpt/models/brave-pro.toml
deleted file mode 100644
index 33249d9e5c3..00000000000
--- a/providers/nano-gpt/models/brave-pro.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Brave (Pro)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2026-02-10"
-last_updated = "2024-01-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 5
-output = 5
-
-[limit]
-context = 8_192
-input = 8_192
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/brave-research.toml b/providers/nano-gpt/models/brave-research.toml
deleted file mode 100644
index 6a66a940b5f..00000000000
--- a/providers/nano-gpt/models/brave-research.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Brave (Research)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2026-02-10"
-last_updated = "2024-01-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 5
-output = 5
-
-[limit]
-context = 16_384
-input = 16_384
-output = 16_384
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/brave.toml b/providers/nano-gpt/models/brave.toml
deleted file mode 100644
index c7a51f0e1a3..00000000000
--- a/providers/nano-gpt/models/brave.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Brave (Answers)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2026-02-13"
-last_updated = "2024-01-01"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 5
-output = 5
-
-[limit]
-context = 8_192
-input = 8_192
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/claude-opus-4-20250514.toml b/providers/nano-gpt/models/claude-opus-4-20250514.toml
deleted file mode 100644
index 133aae4b888..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-20250514.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-20250514"
-name = "Claude 4 Opus"
-reasoning = false
-structured_output = true
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-opus-4-thinking.toml b/providers/nano-gpt/models/claude-opus-4-thinking.toml
deleted file mode 100644
index 078795b2baf..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-thinking.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-name = "Claude 4 Opus Thinking"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-opus-4-thinking:1024.toml b/providers/nano-gpt/models/claude-opus-4-thinking:1024.toml
deleted file mode 100644
index 6d0334ab627..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-thinking:1024.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-name = "Claude 4 Opus Thinking (1K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-opus-4-thinking:32768.toml b/providers/nano-gpt/models/claude-opus-4-thinking:32768.toml
deleted file mode 100644
index ef6450e1339..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-thinking:32768.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-name = "Claude 4 Opus Thinking (32K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-opus-4-thinking:8192.toml b/providers/nano-gpt/models/claude-opus-4-thinking:8192.toml
deleted file mode 100644
index a4eb96b1e3b..00000000000
--- a/providers/nano-gpt/models/claude-opus-4-thinking:8192.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-opus-4-0"
-name = "Claude 4 Opus Thinking (8K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-20250514.toml b/providers/nano-gpt/models/claude-sonnet-4-20250514.toml
deleted file mode 100644
index d9eae66a888..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-20250514.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-20250514"
-name = "Claude 4 Sonnet"
-reasoning = false
-structured_output = true
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-input = 200_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-thinking.toml b/providers/nano-gpt/models/claude-sonnet-4-thinking.toml
deleted file mode 100644
index 81b94bd81ff..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-thinking.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-name = "Claude 4 Sonnet Thinking"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 1_000_000
-input = 1_000_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-thinking:1024.toml b/providers/nano-gpt/models/claude-sonnet-4-thinking:1024.toml
deleted file mode 100644
index eb6625a53e5..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-thinking:1024.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-name = "Claude 4 Sonnet Thinking (1K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 1_000_000
-input = 1_000_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-thinking:32768.toml b/providers/nano-gpt/models/claude-sonnet-4-thinking:32768.toml
deleted file mode 100644
index af58b34bcd6..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-thinking:32768.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-name = "Claude 4 Sonnet Thinking (32K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 1_000_000
-input = 1_000_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-thinking:64000.toml b/providers/nano-gpt/models/claude-sonnet-4-thinking:64000.toml
deleted file mode 100644
index 21468363647..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-thinking:64000.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-name = "Claude 4 Sonnet Thinking (64K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 1_000_000
-input = 1_000_000
diff --git a/providers/nano-gpt/models/claude-sonnet-4-thinking:8192.toml b/providers/nano-gpt/models/claude-sonnet-4-thinking:8192.toml
deleted file mode 100644
index a311bdb9cc2..00000000000
--- a/providers/nano-gpt/models/claude-sonnet-4-thinking:8192.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-0"
-name = "Claude 4 Sonnet Thinking (8K)"
-structured_output = true
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 1_000_000
-input = 1_000_000
diff --git a/providers/nano-gpt/models/crofai/greg-2-super.toml b/providers/nano-gpt/models/crofai/greg-2-super.toml
deleted file mode 100644
index f472a7969fe..00000000000
--- a/providers/nano-gpt/models/crofai/greg-2-super.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Greg 2 Super"
-description = "Greg 2 Super is CrofAI's balanced Greg 2 model for strong UI design, frontend iteration, coding, writing, and everyday agent tasks at a lower cost than Ultra."
-release_date = "2026-06-19"
-last_updated = "2026-06-19"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 1.5
-output = 5
-cache_read = 0.25
-
-[limit]
-context = 229_376
-input = 229_376
-output = 229_376
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/crofai/greg-2-ultra.toml b/providers/nano-gpt/models/crofai/greg-2-ultra.toml
deleted file mode 100644
index 5138d1fea2b..00000000000
--- a/providers/nano-gpt/models/crofai/greg-2-ultra.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Greg 2 Ultra"
-description = "Greg 2 Ultra is CrofAI's most capable Greg 2 model, tuned for premium UI design, agentic coding, creative writing, and higher-end general reasoning tasks."
-release_date = "2026-06-19"
-last_updated = "2026-06-19"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 3
-output = 10
-cache_read = 0.5
-
-[limit]
-context = 229_376
-input = 229_376
-output = 229_376
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 9c2641248c0..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,8 +5,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
+input = 0.1
+output = 0.4
 cache_read = 0.003
 
 [limit]
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 2e7355d5588..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
+input = 0.1
+output = 0.4
 cache_read = 0.003
 
 [limit]
diff --git a/providers/nano-gpt/models/exa-answer.toml b/providers/nano-gpt/models/exa-answer.toml
deleted file mode 100644
index 59dc04c78b0..00000000000
--- a/providers/nano-gpt/models/exa-answer.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Exa (Answer)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2025-12-23"
-last_updated = "2025-06-04"
-attachment = false
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.5
-output = 2.5
-
-[limit]
-context = 4_096
-input = 4_096
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/inference-net/schematron-v2-small.toml b/providers/nano-gpt/models/inference-net/schematron-v2-small.toml
new file mode 100644
index 00000000000..92bbd290059
--- /dev/null
+++ b/providers/nano-gpt/models/inference-net/schematron-v2-small.toml
@@ -0,0 +1,23 @@
+name = "Schematron V2 Small"
+description = "Inference.net's 3B-parameter HTML-to-JSON extraction model, focused on accuracy for complex schemas and long web pages. It turns HTML into typed, structured data for web scraping and product catalog ingestion, with a 128K-token context window. Supply HTML in the user message and extraction instructions in a JSON schema via response_format; it does not follow ordinary chat or system prompts."
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+tool_call = false
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.05
+output = 0.23
+cache_read = 0.025
+
+[limit]
+context = 128_000
+input = 128_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/inference-net/schematron-v2-turbo.toml b/providers/nano-gpt/models/inference-net/schematron-v2-turbo.toml
new file mode 100644
index 00000000000..8c35b1a6bb7
--- /dev/null
+++ b/providers/nano-gpt/models/inference-net/schematron-v2-turbo.toml
@@ -0,0 +1,23 @@
+name = "Schematron V2 Turbo"
+description = "Inference.net's 3B-parameter HTML-to-JSON extraction model, optimized for throughput and low cost on high-volume workloads. It turns HTML into typed, structured data for web scraping and product catalog ingestion, with a 128K-token context window. Supply HTML in the user message and extraction instructions in a JSON schema via response_format; it does not follow ordinary chat or system prompts."
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = false
+tool_call = false
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.03
+output = 0.15
+cache_read = 0.015
+
+[limit]
+context = 128_000
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/meta-llama/llama-3.3-70b-instruct.toml b/providers/nano-gpt/models/meta-llama/llama-3.3-70b-instruct.toml
index 9efae912148..eb46f8f5020 100644
--- a/providers/nano-gpt/models/meta-llama/llama-3.3-70b-instruct.toml
+++ b/providers/nano-gpt/models/meta-llama/llama-3.3-70b-instruct.toml
@@ -1,6 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70b Instruct"
-attachment = false
 structured_output = true
 
 [cost]
diff --git a/providers/nano-gpt/models/MiniMax-M2.toml b/providers/nano-gpt/models/minimax/minimax-m2.toml
similarity index 100%
rename from providers/nano-gpt/models/MiniMax-M2.toml
rename to providers/nano-gpt/models/minimax/minimax-m2.toml
diff --git a/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml b/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
deleted file mode 100644
index e7f724c55a3..00000000000
--- a/providers/nano-gpt/models/mistral-small-31-24b-instruct.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Mistral Small 31 24b Instruct"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "mistral-small"
-release_date = "2025-04-15"
-last_updated = "2025-04-15"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.1
-output = 0.3
-cache_read = 0.05
-
-[limit]
-context = 128_000
-input = 128_000
-output = 102_400
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/mistralai/Devstral-Small-2505.toml b/providers/nano-gpt/models/mistralai/devstral-small-2505.toml
similarity index 74%
rename from providers/nano-gpt/models/mistralai/Devstral-Small-2505.toml
rename to providers/nano-gpt/models/mistralai/devstral-small-2505.toml
index 04d6ee20dc8..a8cfcbf3f42 100644
--- a/providers/nano-gpt/models/mistralai/Devstral-Small-2505.toml
+++ b/providers/nano-gpt/models/mistralai/devstral-small-2505.toml
@@ -1,5 +1,5 @@
 name = "Mistral Devstral Small 2505"
-description = "Mistral coding agent model for repository tasks and software engineering workflows"
+description = "OpenHands+Devstral is 100% local 100% open, and is SOTA for the category on SWE-Bench Verified: 46.8% accuracy."
 family = "devstral"
 release_date = "2025-08-02"
 last_updated = "2025-08-02"
diff --git a/providers/nano-gpt/models/mistral/mistral-medium-3.5.toml b/providers/nano-gpt/models/mistralai/mistral-medium-3.5.toml
similarity index 61%
rename from providers/nano-gpt/models/mistral/mistral-medium-3.5.toml
rename to providers/nano-gpt/models/mistralai/mistral-medium-3.5.toml
index 0c11d88d490..f5857d73d12 100644
--- a/providers/nano-gpt/models/mistral/mistral-medium-3.5.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-medium-3.5.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "Mistral Medium 3.5"
-description = "Mistral model for multilingual chat, reasoning, and tool-assisted workflows"
+description = "Mistral Medium 3.5 is a 128B dense open-weights flagship model for instruction-following, reasoning, coding, long-horizon agentic work, tool use, structured output, and multimodal prompts. It supports a 256k context window and configurable reasoning effort."
 family = "mistral-medium"
 release_date = "2026-04-29"
 last_updated = "2026-04-29"
diff --git a/providers/nano-gpt/models/mistral/mistral-medium-3.5:thinking.toml b/providers/nano-gpt/models/mistralai/mistral-medium-3.5:thinking.toml
similarity index 74%
rename from providers/nano-gpt/models/mistral/mistral-medium-3.5:thinking.toml
rename to providers/nano-gpt/models/mistralai/mistral-medium-3.5:thinking.toml
index 1c5240e619a..7aefb13aa98 100644
--- a/providers/nano-gpt/models/mistral/mistral-medium-3.5:thinking.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-medium-3.5:thinking.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "Mistral Medium 3.5 Thinking"
-description = "Mistral model for multilingual chat, reasoning, and tool-assisted workflows"
+description = "Mistral Medium 3.5 with reasoning enabled by default (reasoning_effort=high), for complex coding, agentic, and multi-step reasoning prompts."
 family = "mistral-medium"
 release_date = "2026-04-30"
 last_updated = "2026-04-30"
diff --git a/providers/nano-gpt/models/mistralai/Mistral-Nemo-Instruct-2407.toml b/providers/nano-gpt/models/mistralai/mistral-nemo-instruct-2407.toml
similarity index 73%
rename from providers/nano-gpt/models/mistralai/Mistral-Nemo-Instruct-2407.toml
rename to providers/nano-gpt/models/mistralai/mistral-nemo-instruct-2407.toml
index 34cd455d40c..5ab5f3549ce 100644
--- a/providers/nano-gpt/models/mistralai/Mistral-Nemo-Instruct-2407.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-nemo-instruct-2407.toml
@@ -1,8 +1,8 @@
 name = "Mistral Nemo"
-description = "Mistral model for multilingual chat, reasoning, and tool-assisted workflows"
+description = "12B parameter model with multilingual support."
 family = "mistral-nemo"
 release_date = "2024-01-01"
-last_updated = "2024-07-18"
+last_updated = "2024-01-01"
 attachment = false
 reasoning = false
 tool_call = false
diff --git a/providers/nano-gpt/models/mistralai/mistral-small-3.1-24b-instruct.toml b/providers/nano-gpt/models/mistralai/mistral-small-3.1-24b-instruct.toml
new file mode 100644
index 00000000000..ed6ee8f4ca7
--- /dev/null
+++ b/providers/nano-gpt/models/mistralai/mistral-small-3.1-24b-instruct.toml
@@ -0,0 +1,24 @@
+name = "Mistral Small 3.1 24B (2503)"
+description = "Building upon Mistral Small 3 (2501), Mistral Small 3.1 (2503) adds state-of-the-art vision understanding and enhances long context capabilities up to 128k tokens without compromising text performance. With 24 billion parameters, this model achieves top-tier capabilities in both text and vision tasks."
+family = "mistral-small"
+release_date = "2025-04-15"
+last_updated = "2025-04-15"
+attachment = true
+reasoning = false
+tool_call = false
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.1
+output = 0.3
+cache_read = 0.05
+
+[limit]
+context = 128_000
+input = 128_000
+output = 102_400
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml b/providers/nano-gpt/models/mistralai/mistral-small-3.2-24b-instruct.toml
similarity index 50%
rename from providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml
rename to providers/nano-gpt/models/mistralai/mistral-small-3.2-24b-instruct.toml
index b4ab0f2f7a0..294870992cf 100644
--- a/providers/nano-gpt/models/chutesai/Mistral-Small-3.2-24B-Instruct-2506.toml
+++ b/providers/nano-gpt/models/mistralai/mistral-small-3.2-24b-instruct.toml
@@ -1,6 +1,6 @@
-name = "Mistral Small 3.2 24b Instruct"
-description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "chutesai"
+name = "Mistral Small 3.2 24B (2506)"
+description = "The latest iteration of Mistral Small, version 3.2 (2506) brings enhanced performance and capabilities. With 24 billion parameters, this model delivers state-of-the-art results across text generation tasks with improved efficiency."
+family = "mistral-small"
 release_date = "2025-04-15"
 last_updated = "2025-04-15"
 attachment = false
diff --git a/providers/nano-gpt/models/perplexity-academic-researcher.toml b/providers/nano-gpt/models/perplexity-academic-researcher.toml
deleted file mode 100644
index 3cb12eae9f3..00000000000
--- a/providers/nano-gpt/models/perplexity-academic-researcher.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Perplexity Academic Researcher"
-description = "Sonar Reasoning Pro with Perplexity's academic search mode. Prioritizes scholarly and peer-reviewed sources from academic repositories and returns cited research synthesis."
-release_date = "2026-07-10"
-last_updated = "2026-07-10"
-attachment = false
-reasoning = true
-tool_call = false
-structured_output = false
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 2
-output = 8
-cache_read = 1
-
-[limit]
-context = 128_000
-input = 128_000
-output = 115_200
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen-3.6-plus.toml b/providers/nano-gpt/models/qwen-3.6-plus.toml
deleted file mode 100644
index 76d18669d5a..00000000000
--- a/providers/nano-gpt/models/qwen-3.6-plus.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen 3.6 Plus"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "qwen3.6"
-release_date = "2026-04-02"
-last_updated = "2026-04-02"
-attachment = true
-reasoning = false
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.325
-output = 1.95
-cache_read = 0.0325
-cache_write = 0.40625
-
-[limit]
-context = 991_808
-input = 991_808
-output = 65_536
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml b/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
deleted file mode 100644
index e486bde9b43..00000000000
--- a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Thinking-2507.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-# Included in subscription
-name = "Qwen 3 235b A22B 2507 Thinking"
-description = "Qwen reasoning model for deliberate problem solving, math, and coding"
-family = "qwen"
-release_date = "2025-09-11"
-last_updated = "2025-09-11"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.3
-output = 0.5
-cache_read = 0.15
-
-[limit]
-context = 131_072
-input = 131_072
-output = 117_964
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen-long.toml b/providers/nano-gpt/models/qwen/qwen-long.toml
similarity index 69%
rename from providers/nano-gpt/models/qwen-long.toml
rename to providers/nano-gpt/models/qwen/qwen-long.toml
index 2d5577bad56..dc36c556c44 100644
--- a/providers/nano-gpt/models/qwen-long.toml
+++ b/providers/nano-gpt/models/qwen/qwen-long.toml
@@ -1,8 +1,8 @@
 name = "Qwen Long 10M"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
+description = "Alibaba's huge context window model. Takes in up to 10 million tokens, which is equivalent to dozens of books."
 family = "qwen"
 release_date = "2024-08-01"
-last_updated = "2025-01-25"
+last_updated = "2024-08-01"
 attachment = true
 reasoning = false
 tool_call = false
diff --git a/providers/nano-gpt/models/qwen-max.toml b/providers/nano-gpt/models/qwen/qwen-max.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen-max.toml
rename to providers/nano-gpt/models/qwen/qwen-max.toml
diff --git a/providers/nano-gpt/models/qwen-plus.toml b/providers/nano-gpt/models/qwen/qwen-plus.toml
similarity index 62%
rename from providers/nano-gpt/models/qwen-plus.toml
rename to providers/nano-gpt/models/qwen/qwen-plus.toml
index 533821e91df..89f28ecb6f6 100644
--- a/providers/nano-gpt/models/qwen-plus.toml
+++ b/providers/nano-gpt/models/qwen/qwen-plus.toml
@@ -1,14 +1,7 @@
 base_model = "alibaba/qwen-plus"
 tool_call = false
 structured_output = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.3995
diff --git a/providers/nano-gpt/models/qwen-turbo.toml b/providers/nano-gpt/models/qwen/qwen-turbo.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen-turbo.toml
rename to providers/nano-gpt/models/qwen/qwen-turbo.toml
diff --git a/providers/nano-gpt/models/qwen/Qwen2.5-Coder-32B-Instruct.toml b/providers/nano-gpt/models/qwen/qwen2.5-coder-32b-instruct.toml
similarity index 89%
rename from providers/nano-gpt/models/qwen/Qwen2.5-Coder-32B-Instruct.toml
rename to providers/nano-gpt/models/qwen/qwen2.5-coder-32b-instruct.toml
index d62bbdf8079..221615dee2a 100644
--- a/providers/nano-gpt/models/qwen/Qwen2.5-Coder-32B-Instruct.toml
+++ b/providers/nano-gpt/models/qwen/qwen2.5-coder-32b-instruct.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen2.5-coder-32b-instruct"
 name = "Qwen 2.5 Coder 32b"
 tool_call = false
diff --git a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml b/providers/nano-gpt/models/qwen/qwen3-235b-a22b-instruct-2507.toml
similarity index 80%
rename from providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml
rename to providers/nano-gpt/models/qwen/qwen3-235b-a22b-instruct-2507.toml
index ba1df7fad61..da2f8245205 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-235B-A22B-Instruct-2507.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-235b-a22b-instruct-2507.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3-235b-a22b-instruct-2507"
 name = "Qwen 3 235b A22B 2507"
 structured_output = true
@@ -10,4 +9,3 @@ cache_read = 0.065
 
 [limit]
 input = 262_144
-output = 235_929
diff --git a/providers/nano-gpt/models/qwen3-30b-a3b-instruct-2507.toml b/providers/nano-gpt/models/qwen/qwen3-30b-a3b-instruct-2507.toml
similarity index 51%
rename from providers/nano-gpt/models/qwen3-30b-a3b-instruct-2507.toml
rename to providers/nano-gpt/models/qwen/qwen3-30b-a3b-instruct-2507.toml
index a6f5f2c3866..879ab1c3b88 100644
--- a/providers/nano-gpt/models/qwen3-30b-a3b-instruct-2507.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-30b-a3b-instruct-2507.toml
@@ -1,5 +1,6 @@
 name = "Qwen3 30B A3B Instruct 2507"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
+description = "Qwen3-30B-A3B-Instruct-2507 is a 30.5B-parameter mixture-of-experts language model from Qwen, with 3.3B active parameters per inference. Significant improvements in general capabilities, including instruction following, logical reasoning, text comprehension, mathematics, science, coding and tool usage."
+family = "qwen"
 release_date = "2025-02-20"
 last_updated = "2025-02-20"
 attachment = false
diff --git a/providers/nano-gpt/models/qwen/Qwen3-8B.toml b/providers/nano-gpt/models/qwen/qwen3-8b.toml
similarity index 62%
rename from providers/nano-gpt/models/qwen/Qwen3-8B.toml
rename to providers/nano-gpt/models/qwen/qwen3-8b.toml
index bc8aaa1ddb2..5dd994f1ea3 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-8B.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-8b.toml
@@ -1,6 +1,5 @@
-# Included in subscription
 name = "Qwen 3 8B"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
+description = "Qwen 3 8B is a 8B model. Supports switching between thinking and non thinking: trigger thinking with /think and /no_think anywhere in a prompt or system message to toggle chain-of-thought reasoning."
 family = "qwen"
 release_date = "2024-01-01"
 last_updated = "2024-01-01"
diff --git a/providers/nano-gpt/models/qwen3-coder-30b-a3b-instruct.toml b/providers/nano-gpt/models/qwen/qwen3-coder-30b-a3b-instruct.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3-coder-30b-a3b-instruct.toml
rename to providers/nano-gpt/models/qwen/qwen3-coder-30b-a3b-instruct.toml
diff --git a/providers/nano-gpt/models/qwen3-max-2026-01-23.toml b/providers/nano-gpt/models/qwen/qwen3-max-2026-01-23.toml
similarity index 66%
rename from providers/nano-gpt/models/qwen3-max-2026-01-23.toml
rename to providers/nano-gpt/models/qwen/qwen3-max-2026-01-23.toml
index 514bb7a5472..8ae8fcfc81f 100644
--- a/providers/nano-gpt/models/qwen3-max-2026-01-23.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-max-2026-01-23.toml
@@ -1,5 +1,6 @@
 name = "Qwen3 Max 2026-01-23"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
+description = "Qwen3 Max is Alibaba's flagship Qwen 3 reasoning model with native tool use (web search, web extractor, code interpreter) and a 256K context window."
+family = "qwen"
 release_date = "2026-01-26"
 last_updated = "2026-01-26"
 attachment = false
diff --git a/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml b/providers/nano-gpt/models/qwen/qwen3-next-80b-a3b-instruct.toml
similarity index 82%
rename from providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml
rename to providers/nano-gpt/models/qwen/qwen3-next-80b-a3b-instruct.toml
index 86078c83dfc..deab693555e 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-Next-80B-A3B-Instruct.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-next-80b-a3b-instruct.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3-next-80b-a3b-instruct"
 name = "Qwen3 Next 80B A3B (Instruct)"
 structured_output = true
@@ -11,4 +10,3 @@ cache_read = 0.075
 [limit]
 context = 262_144
 input = 262_144
-output = 235_929
diff --git a/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
similarity index 86%
rename from providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml
rename to providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
index cfdc9a4897d..a648f070f2e 100644
--- a/providers/nano-gpt/models/qwen/Qwen3-VL-235B-A22B-Instruct.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3-vl-235b-a22b-instruct"
 tool_call = false
 structured_output = false
diff --git a/providers/nano-gpt/models/qwen3-vl-235b-a22b-thinking.toml b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-thinking.toml
similarity index 56%
rename from providers/nano-gpt/models/qwen3-vl-235b-a22b-thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-thinking.toml
index e93934fc273..7ded44f273c 100644
--- a/providers/nano-gpt/models/qwen3-vl-235b-a22b-thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-thinking.toml
@@ -1,17 +1,9 @@
 base_model = "alibaba/qwen3-vl-235b-a22b-thinking"
 tool_call = false
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.5
 output = 6
 cache_read = 0.25
-
-[limit]
-context = 32_768
-input = 32_768
diff --git a/providers/nano-gpt/models/qwen3.5-0.8b.toml b/providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.5-0.8b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml
diff --git a/providers/nano-gpt/models/qwen3.5-122b-a10b.toml b/providers/nano-gpt/models/qwen/qwen3.5-122b-a10b.toml
similarity index 90%
rename from providers/nano-gpt/models/qwen3.5-122b-a10b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-122b-a10b.toml
index 307ecd0c8e1..9d1311c4fc4 100644
--- a/providers/nano-gpt/models/qwen3.5-122b-a10b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
 attachment = false
diff --git a/providers/nano-gpt/models/qwen3.5-122b-a10b:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-122b-a10b:thinking.toml
similarity index 71%
rename from providers/nano-gpt/models/qwen3.5-122b-a10b:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-122b-a10b:thinking.toml
index 83acdd20d9b..be6d38169e7 100644
--- a/providers/nano-gpt/models/qwen3.5-122b-a10b:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-122b-a10b:thinking.toml
@@ -1,12 +1,7 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B Thinking"
 attachment = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.437
diff --git a/providers/nano-gpt/models/qwen3.5-27b.toml b/providers/nano-gpt/models/qwen/qwen3.5-27b.toml
similarity index 89%
rename from providers/nano-gpt/models/qwen3.5-27b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-27b.toml
index 2ee18716021..32d037d663e 100644
--- a/providers/nano-gpt/models/qwen3.5-27b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-27b.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-27b"
 reasoning = false
 structured_output = false
diff --git a/providers/nano-gpt/models/qwen3.5-27b:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-27b:thinking.toml
similarity index 70%
rename from providers/nano-gpt/models/qwen3.5-27b:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-27b:thinking.toml
index 978f3266618..80fa741e926 100644
--- a/providers/nano-gpt/models/qwen3.5-27b:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-27b:thinking.toml
@@ -1,12 +1,7 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-27b"
 name = "Qwen3.5 27B Thinking"
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.27
diff --git a/providers/nano-gpt/models/qwen3.5-2b.toml b/providers/nano-gpt/models/qwen/qwen3.5-2b.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.5-2b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-2b.toml
diff --git a/providers/nano-gpt/models/qwen3.5-35b-a3b.toml b/providers/nano-gpt/models/qwen/qwen3.5-35b-a3b.toml
similarity index 90%
rename from providers/nano-gpt/models/qwen3.5-35b-a3b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-35b-a3b.toml
index 96c810764f0..c24e51666d7 100644
--- a/providers/nano-gpt/models/qwen3.5-35b-a3b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-35b-a3b.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
 reasoning = false
diff --git a/providers/nano-gpt/models/qwen3.5-35b-a3b:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-35b-a3b:thinking.toml
similarity index 71%
rename from providers/nano-gpt/models/qwen3.5-35b-a3b:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-35b-a3b:thinking.toml
index cca1e2de29b..293b615da15 100644
--- a/providers/nano-gpt/models/qwen3.5-35b-a3b:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-35b-a3b:thinking.toml
@@ -1,12 +1,7 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B Thinking"
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.225
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-397b-a17b-thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-397b-a17b:thinking.toml
similarity index 71%
rename from providers/nano-gpt/models/qwen/qwen3.5-397b-a17b-thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-397b-a17b:thinking.toml
index fe9f90244de..871fa4d8523 100644
--- a/providers/nano-gpt/models/qwen/qwen3.5-397b-a17b-thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-397b-a17b:thinking.toml
@@ -1,12 +1,7 @@
-# Included in subscription
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B Thinking"
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.6
diff --git a/providers/nano-gpt/models/qwen3.5-4b.toml b/providers/nano-gpt/models/qwen/qwen3.5-4b.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.5-4b.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-4b.toml
diff --git a/providers/nano-gpt/models/qwen3.5-flash.toml b/providers/nano-gpt/models/qwen/qwen3.5-flash.toml
similarity index 86%
rename from providers/nano-gpt/models/qwen3.5-flash.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-flash.toml
index 90792fb38dc..1602bd5551c 100644
--- a/providers/nano-gpt/models/qwen3.5-flash.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-flash.toml
@@ -1,4 +1,3 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.5-flash"
 reasoning = false
 tool_call = false
diff --git a/providers/nano-gpt/models/qwen3.5-flash:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-flash:thinking.toml
similarity index 66%
rename from providers/nano-gpt/models/qwen3.5-flash:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-flash:thinking.toml
index 5f23dc9cec2..11bfed64c49 100644
--- a/providers/nano-gpt/models/qwen3.5-flash:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-flash:thinking.toml
@@ -1,13 +1,8 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.5-flash"
 name = "Qwen3.5 Flash Thinking"
 tool_call = false
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.1
diff --git a/providers/nano-gpt/models/qwen3.5-omni-flash.toml b/providers/nano-gpt/models/qwen/qwen3.5-omni-flash.toml
similarity index 51%
rename from providers/nano-gpt/models/qwen3.5-omni-flash.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-omni-flash.toml
index 0184330290c..ee5c50a1c6b 100644
--- a/providers/nano-gpt/models/qwen3.5-omni-flash.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-omni-flash.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "Qwen3.5 Omni Flash"
-description = "Omni-modal model for text, vision, audio, and multimodal agent tasks"
+description = "Qwen3.5 Omni Flash is Qwen's fast multimodal model. We verified live support for text prompts, images, audio files, and direct video URLs on Alibaba's chat-completions-compatible API. Alibaba describes Flash as a fully evolved version of Qwen3 Omni with audio input support across 60+ languages."
 family = "qwen3.5"
 release_date = "2026-03-30"
 last_updated = "2026-03-30"
@@ -10,10 +9,6 @@ tool_call = false
 structured_output = false
 open_weights = false
 
-[cost]
-input = 0
-output = 0
-
 [limit]
 context = 49_152
 input = 49_152
diff --git a/providers/nano-gpt/models/qwen3.5-omni-plus.toml b/providers/nano-gpt/models/qwen/qwen3.5-omni-plus.toml
similarity index 51%
rename from providers/nano-gpt/models/qwen3.5-omni-plus.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-omni-plus.toml
index 379e191256a..2de40cc5893 100644
--- a/providers/nano-gpt/models/qwen3.5-omni-plus.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-omni-plus.toml
@@ -1,6 +1,5 @@
-# Not included in subscription
 name = "Qwen3.5 Omni Plus"
-description = "Omni-modal model for text, vision, audio, and multimodal agent tasks"
+description = "Qwen3.5 Omni Plus is Qwen's stronger general multimodal model. We verified live support for text prompts, images, audio files, and direct video URLs on Alibaba's chat-completions-compatible API. Alibaba describes Plus as a comprehensive evolution of Qwen3 Omni with support for over 10 hours of audio input."
 family = "qwen3.5"
 release_date = "2026-03-30"
 last_updated = "2026-03-30"
@@ -10,10 +9,6 @@ tool_call = false
 structured_output = false
 open_weights = false
 
-[cost]
-input = 0
-output = 0
-
 [limit]
 context = 983_616
 input = 983_616
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-plus-thinking.toml b/providers/nano-gpt/models/qwen/qwen3.5-plus:thinking.toml
similarity index 68%
rename from providers/nano-gpt/models/qwen/qwen3.5-plus-thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.5-plus:thinking.toml
index c62dcbeaaa2..e65e53b70fd 100644
--- a/providers/nano-gpt/models/qwen/qwen3.5-plus-thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.5-plus:thinking.toml
@@ -1,14 +1,9 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.5-plus"
 name = "Qwen3.5 Plus Thinking"
 attachment = true
 tool_call = false
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 81_920
+reasoning_options = []
 
 [cost]
 input = 0.4
diff --git a/providers/nano-gpt/models/alibaba/qwen3.6-27b.toml b/providers/nano-gpt/models/qwen/qwen3.6-27b.toml
similarity index 89%
rename from providers/nano-gpt/models/alibaba/qwen3.6-27b.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-27b.toml
index 2af25f0ba71..660bbf5d035 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.6-27b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-27b.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.6-27b"
 reasoning = false
 structured_output = false
diff --git a/providers/nano-gpt/models/alibaba/qwen3.6-27b:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-27b:thinking.toml
similarity index 70%
rename from providers/nano-gpt/models/alibaba/qwen3.6-27b:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-27b:thinking.toml
index 9cafb1eba73..b8de29ec3db 100644
--- a/providers/nano-gpt/models/alibaba/qwen3.6-27b:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-27b:thinking.toml
@@ -1,12 +1,7 @@
-# Included in subscription
 base_model = "alibaba/qwen3.6-27b"
 name = "Qwen3.6 27B Thinking"
 structured_output = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 131_072
+reasoning_options = []
 
 [cost]
 input = 0.203
diff --git a/providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b.toml
similarity index 90%
rename from providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-35b-a3b.toml
index 815ba570fca..8f97165cf08 100644
--- a/providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B A3B"
 reasoning = false
diff --git a/providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b:thinking.toml
similarity index 90%
rename from providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-35b-a3b:thinking.toml
index faa635782f9..6305b4c7e97 100644
--- a/providers/nano-gpt/models/qwen/Qwen3.6-35B-A3B:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.6-35b-a3b:thinking.toml
@@ -1,4 +1,3 @@
-# Included in subscription
 base_model = "alibaba/qwen3.6-35b-a3b"
 name = "Qwen3.6 35B A3B Thinking"
 structured_output = false
diff --git a/providers/nano-gpt/models/alibaba/qwen3.6-flash.toml b/providers/nano-gpt/models/qwen/qwen3.6-flash.toml
similarity index 100%
rename from providers/nano-gpt/models/alibaba/qwen3.6-flash.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-flash.toml
diff --git a/providers/nano-gpt/models/qwen3.6-max-preview.toml b/providers/nano-gpt/models/qwen/qwen3.6-max-preview.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.6-max-preview.toml
rename to providers/nano-gpt/models/qwen/qwen3.6-max-preview.toml
diff --git a/providers/nano-gpt/models/qwen/qwen3.6-plus.toml b/providers/nano-gpt/models/qwen/qwen3.6-plus.toml
new file mode 100644
index 00000000000..32e5100a867
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.6-plus.toml
@@ -0,0 +1,14 @@
+base_model = "alibaba/qwen3.6-plus"
+name = "Qwen 3.6 Plus"
+reasoning = false
+structured_output = false
+
+[cost]
+input = 0.325
+output = 1.95
+cache_read = 0.0325
+cache_write = 0.40625
+
+[limit]
+context = 991_808
+input = 991_808
diff --git a/providers/nano-gpt/models/qwen3.7-flash.toml b/providers/nano-gpt/models/qwen/qwen3.7-flash.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.7-flash.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-flash.toml
diff --git a/providers/nano-gpt/models/qwen3.7-flash:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.7-flash:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.7-flash:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-flash:thinking.toml
diff --git a/providers/nano-gpt/models/qwen3.7-max.toml b/providers/nano-gpt/models/qwen/qwen3.7-max.toml
similarity index 87%
rename from providers/nano-gpt/models/qwen3.7-max.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-max.toml
index 8296afa1795..a7bb9d17077 100644
--- a/providers/nano-gpt/models/qwen3.7-max.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.7-max.toml
@@ -1,4 +1,3 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.7-max"
 structured_output = false
 
diff --git a/providers/nano-gpt/models/qwen3.7-max:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.7-max:thinking.toml
similarity index 89%
rename from providers/nano-gpt/models/qwen3.7-max:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-max:thinking.toml
index 5c4948444ef..e83e629bf83 100644
--- a/providers/nano-gpt/models/qwen3.7-max:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.7-max:thinking.toml
@@ -1,4 +1,3 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.7-max"
 name = "Qwen3.7 Max Thinking"
 structured_output = false
diff --git a/providers/nano-gpt/models/qwen3.7-plus.toml b/providers/nano-gpt/models/qwen/qwen3.7-plus.toml
similarity index 89%
rename from providers/nano-gpt/models/qwen3.7-plus.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-plus.toml
index 47994898e2c..f4a154e614c 100644
--- a/providers/nano-gpt/models/qwen3.7-plus.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.7-plus.toml
@@ -1,4 +1,3 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.7-plus"
 structured_output = false
 
diff --git a/providers/nano-gpt/models/qwen3.7-plus:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.7-plus:thinking.toml
similarity index 90%
rename from providers/nano-gpt/models/qwen3.7-plus:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.7-plus:thinking.toml
index ac846d4646c..11e249fc621 100644
--- a/providers/nano-gpt/models/qwen3.7-plus:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.7-plus:thinking.toml
@@ -1,4 +1,3 @@
-# Not included in subscription
 base_model = "alibaba/qwen3.7-plus"
 name = "Qwen3.7 Plus Thinking"
 structured_output = false
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/nano-gpt/models/qwen/qwen3.8-2.4t-a95b.toml
deleted file mode 100644
index 3a44fadfe61..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.8-2.4t-a95b.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "alibaba/qwen3.8-2.4t-a95b"
-name = "Qwen3.8 2.4T A95B (Max)"
-attachment = true
-reasoning = false
-
-[cost]
-input = 2
-output = 6
-cache_read = 0.25
-cache_write = 2.5
-
-[limit]
-context = 991_000
-input = 991_000
-output = 65_536
-
-[modalities]
-input = ["text", "image", "video", "pdf"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
index 214e4ccf3b3..4a735ca88fa 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
@@ -13,7 +13,7 @@ reasoning_options = []
 [cost]
 input = 0.25
 output = 1.5
-cache_read = 0.125
+cache_read = 0.2
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
index 29772075d7d..00bcf527067 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
@@ -13,7 +13,7 @@ reasoning_options = []
 [cost]
 input = 0.25
 output = 1.5
-cache_read = 0.125
+cache_read = 0.2
 
 [limit]
 context = 262_144
diff --git a/providers/nano-gpt/models/qwen3.8-27b.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.8-27b.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-27b.toml
diff --git a/providers/nano-gpt/models/qwen3.8-27b:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.8-27b:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-27b:thinking.toml
diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-flash.toml b/providers/nano-gpt/models/qwen/qwen3.8-flash.toml
similarity index 100%
rename from providers/nano-gpt/models/alibaba/qwen3.8-flash.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-flash.toml
diff --git a/providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml b/providers/nano-gpt/models/qwen/qwen3.8-max-0902.toml
similarity index 100%
rename from providers/nano-gpt/models/alibaba/qwen3.8-max-0902.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-max-0902.toml
diff --git a/providers/nano-gpt/models/qwen3.8-max.toml b/providers/nano-gpt/models/qwen/qwen3.8-max.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.8-max.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-max.toml
diff --git a/providers/nano-gpt/models/qwen3.8-max:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-max:thinking.toml
similarity index 100%
rename from providers/nano-gpt/models/qwen3.8-max:thinking.toml
rename to providers/nano-gpt/models/qwen/qwen3.8-max:thinking.toml
diff --git a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml b/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
deleted file mode 100644
index 4815829fbbe..00000000000
--- a/providers/nano-gpt/models/qwen25-vl-72b-instruct.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen25 VL 72b"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2025-05-10"
-last_updated = "2025-05-10"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.69989
-output = 0.69989
-cache_read = 0.349945
-
-[limit]
-context = 32_000
-input = 32_000
-output = 115_200
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/sonar-deep-research.toml b/providers/nano-gpt/models/sonar-deep-research.toml
deleted file mode 100644
index a14277752d8..00000000000
--- a/providers/nano-gpt/models/sonar-deep-research.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "perplexity/sonar-deep-research"
-name = "Perplexity Deep Research"
-reasoning = false
-structured_output = false
-
-[cost]
-input = 3.4
-output = 13.6
-cache_read = 1.7
-
-[limit]
-input = 128_000
-output = 115_200
diff --git a/providers/nano-gpt/models/sonar-pro.toml b/providers/nano-gpt/models/sonar-pro.toml
deleted file mode 100644
index 3268f10405e..00000000000
--- a/providers/nano-gpt/models/sonar-pro.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "perplexity/sonar-pro"
-name = "Perplexity Pro"
-attachment = false
-structured_output = false
-
-[cost]
-input = 3
-output = 15
-cache_read = 1.5
-
-[limit]
-input = 200_000
-output = 128_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/sonar-reasoning-pro.toml b/providers/nano-gpt/models/sonar-reasoning-pro.toml
deleted file mode 100644
index 99d46d6010b..00000000000
--- a/providers/nano-gpt/models/sonar-reasoning-pro.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "perplexity/sonar-reasoning-pro"
-name = "Perplexity Reasoning Pro"
-attachment = false
-structured_output = false
-reasoning_options = []
-
-[cost]
-input = 2
-output = 8
-cache_read = 1
-
-[limit]
-input = 128_000
-output = 115_200
-
-[modalities]
-input = ["text"]
diff --git a/providers/nano-gpt/models/sonar.toml b/providers/nano-gpt/models/sonar.toml
deleted file mode 100644
index 015df323ea3..00000000000
--- a/providers/nano-gpt/models/sonar.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "perplexity/sonar"
-name = "Perplexity Simple"
-structured_output = false
-
-[cost]
-input = 1
-output = 1
-cache_read = 0.5
-
-[limit]
-context = 127_072
-input = 127_072
-output = 114_364
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof.toml
deleted file mode 100644
index 82bd0f094c0..00000000000
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "MiMo V2.5 Pro (Crof)"
-description = "MiMo V2.5 Pro is Xiaomi's long-context flagship general model for coding and agentic orchestration. This separately served variant is intended for users concerned about censorship on the regular Xiaomi MiMo V2.5 Pro, and it is included in the NanoGPT subscription."
-family = "mimo-v2.5-pro"
-release_date = "2026-07-23"
-last_updated = "2026-07-23"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 0.4
-output = 0.8
-cache_read = 0.003
-
-[limit]
-context = 1_000_000
-input = 1_000_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof:thinking.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof:thinking.toml
deleted file mode 100644
index dd75923c267..00000000000
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.5-pro-crof:thinking.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "MiMo V2.5 Pro Thinking (Crof)"
-description = "MiMo V2.5 Pro with Xiaomi thinking enabled for coding, long-context reasoning, and agentic orchestration. This separately served thinking variant is intended for users concerned about censorship on the regular Xiaomi MiMo V2.5 Pro, and it is included in the NanoGPT subscription."
-family = "mimo-v2.5-pro"
-release_date = "2026-07-23"
-last_updated = "2026-07-23"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 0.4
-output = 0.8
-cache_read = 0.003
-
-[limit]
-context = 1_000_000
-input = 1_000_000
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 2d0019c9056..8136288a6f3 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -14,9 +14,9 @@ type = "effort"
 values = ["low", "high"]
 
 [cost]
-input = 0.35
-output = 1.4
-cache_read = 0.175
+input = 0.2
+output = 0.8
+cache_read = 0.1
 
 [limit]
 context = 1_048_576

From dbd755ee89e34bbc6729829e3244f0293774b3fd Mon Sep 17 00:00:00 2001
From: Michael Chin <mchin188@yahoo.com>
Date: Mon, 14 Sep 2026 09:38:18 -0700
Subject: [PATCH 1285/2012] Remove PDF input from Bedrock GPT-6 and GPT-5.6
 models (#6821)

---
 .../amazon-bedrock/models/global.openai.gpt-5.6-luna.toml      | 3 +++
 providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml | 3 +++
 .../amazon-bedrock/models/global.openai.gpt-5.6-terra.toml     | 3 +++
 providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml | 3 +++
 providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml    | 3 +++
 providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml     | 3 +++
 providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml   | 3 +++
 providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml     | 3 +++
 8 files changed, 24 insertions(+)

diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
index f1928817aaa..8e4d5dccc9b 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml
@@ -20,3 +20,6 @@ input = 0.40
 output = 1.80
 cache_read = 0.04
 cache_write = 0.50
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
index 4b757f188e2..d42a97688d3 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml
@@ -21,3 +21,6 @@ input = 8.00
 output = 30.00
 cache_read = 0.80
 cache_write = 10.00
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
index 3c9cd73b7d1..28ec6d1b3b8 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml
@@ -21,3 +21,6 @@ input = 4.00
 output = 18.00
 cache_read = 0.40
 cache_write = 5.00
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
index 0e751b4ebd6..4300dbcaa4f 100644
--- a/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-astra.toml
@@ -21,3 +21,6 @@ input = 20.00
 output = 75.00
 cache_read = 2.00
 cache_write = 25.00
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
index e0c96587bb0..36247664fc1 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-luna.toml
@@ -22,3 +22,6 @@ input = 0.44
 output = 1.98
 cache_read = 0.044
 cache_write = 0.55
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
index 326e35c16da..31090274654 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-sol.toml
@@ -22,3 +22,6 @@ input = 8.80
 output = 33.00
 cache_read = 0.88
 cache_write = 11.00
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
index cff259b2548..2d48790f4a6 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-5.6-terra.toml
@@ -22,3 +22,6 @@ input = 4.40
 output = 19.80
 cache_read = 0.44
 cache_write = 5.50
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
index 272d4a5ed09..19ccdd3731a 100644
--- a/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-astra.toml
@@ -21,3 +21,6 @@ input = 22.00
 output = 82.50
 cache_read = 2.20
 cache_write = 27.50
+
+[modalities]
+input = ["text", "image"]

From 9eabc50f866d1f2acf10b3c454ee940899b3f47a Mon Sep 17 00:00:00 2001
From: Fenil Modi <fenilmodi4820@gmail.com>
Date: Mon, 14 Sep 2026 22:08:56 +0530
Subject: [PATCH 1286/2012] fix(aiand): logo mark, cache pricing, and reasoning
 options (#6839)

* fix(aiand): use square A-mark logo for catalog display

The prior wordmark viewBox (~3:1) letterboxed into the 1rem provider slot and looked blank. Extract the brand A-mark into a 0 0 24 24 currentColor SVG so it renders like other providers.

* fix(aiand): sync cache_read pricing and reasoning_options

Align all ai& model entries with live GET /v1/models and effort probes: add missing cache_read rates, narrow reasoning_effort to host-accepted values, and document reasoning_content interleaved like other providers.
---
 providers/aiand/logo.svg                      |  7 +++---
 .../models/deepseek-ai/deepseek-v4-flash.toml | 18 ++++++++-------
 .../models/deepseek-ai/deepseek-v4-pro.toml   | 17 ++++++++------
 .../aiand/models/google/gemma-4-31b-it.toml   | 21 +++++++++---------
 .../models/moonshotai/kimi-k2.7-code.toml     | 19 +++++++++-------
 .../aiand/models/moonshotai/kimi-k3.toml      | 22 ++++++++-----------
 .../models/motif-technologies/motif-3.toml    | 17 ++++++++++++--
 .../aiand/models/openai/gpt-oss-120b.toml     | 14 +++++++-----
 providers/aiand/models/qwen/qwen3.6-27b.toml  | 21 +++++++++---------
 providers/aiand/models/qwen/qwen3.8-27b.toml  | 17 +++++++-------
 providers/aiand/models/zai-org/glm-5.2.toml   | 17 ++++++++------
 providers/aiand/models/zai-org/glm-5.3.toml   | 18 +++++++--------
 12 files changed, 115 insertions(+), 93 deletions(-)

diff --git a/providers/aiand/logo.svg b/providers/aiand/logo.svg
index 0a0726cda82..e36f046a233 100644
--- a/providers/aiand/logo.svg
+++ b/providers/aiand/logo.svg
@@ -1,5 +1,4 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 471 1280 430" width="24" height="24" fill="currentColor">
-  <g transform="translate(0 430)">
-    <path d="m1036.05 174.114 102.7-59.232 27.43 45.579c-140.14 75.45-215.743 92.363-192.311 166.512 35.811 65.478 124.001 31.082 146.991 2.463l-70.6-67.493 42.37-23.43c16.5 16.189 56.74 54.391 62.04 59.247 3.54-1.734 16.73-9.813 20-31.494h58.22c-14.86 33.303-28.14 42.07-41.99 63.027l78.56 80.071h-68.68l-41.87-41.008c-102.57 84.123-227.969 43.61-244.725-27.506-20.512-87.072 40.084-129.654 72.947-140.061l-60.566-59.407 40.229-30.788z"/><path d="m1036.63 168.825 103.63-59.764 31.92 53.045-3.97 2.133c-35.1 18.898-66.18 34.144-92.56 47.473-26.46 13.367-48.07 24.753-64.54 36.034-16.493 11.29-27.372 22.169-32.878 34.311-5.393 11.892-5.894 25.599-.394 43.228 16.822 30.395 45.632 37.773 73.752 33.954 14.16-1.922 27.99-6.69 39.67-12.719 9.99-5.159 18.2-11.133 23.65-16.829l-71.83-68.678 50.27-27.802 2.28 2.239c14.96 14.668 49.44 47.446 59.63 56.991 4.51-3.073 12.75-10.777 15.16-26.813l.55-3.653h68.53l-2.69 6.039c-7.59 17.006-14.81 27.813-21.89 37.013-6.49 8.43-12.38 14.952-18.5 23.766l83.27 84.862h-80.66l-40.4-39.564c-51.46 40.852-108.42 51.515-155.04 42.947-47.298-8.693-84.683-37.436-93.581-75.205-10.543-44.75-.266-78.449 17.17-102.295 15.196-20.783 35.684-33.899 52.071-40.471l-59.221-58.087L967.044 105zm-103.529-27.043 62.132 60.942-6.805 2.155c-15.704 4.973-38.249 17.742-54.321 39.723-15.95 21.815-25.715 52.943-15.745 95.263 7.857 33.348 41.551 60.42 86.778 68.732 45.01 8.271 100.72-2.284 151.05-43.558l2.97-2.441 43.37 42.475h56.71l-73.83-75.252 1.91-2.894c7.03-10.63 14.15-18.506 20.8-27.136 5.86-7.615 11.81-16.351 18.04-29.234h-47.91c-4.39 20.767-17.52 29.015-21.69 31.055l-2.62 1.289-2.16-1.976c-5.12-4.687-42.19-39.874-59.87-57.113l-34.46 19.062 69.22 66.18-2.47 3.07c-6.24 7.773-16.62 15.627-29.01 22.022-12.44 6.423-27.2 11.528-42.44 13.597-30.56 4.149-63.629-3.94-82.646-38.711l-.2-.367-.127-.398c-6.084-19.255-5.892-35.353.639-49.755 6.458-14.241 18.845-26.21 35.844-37.848 17.01-11.647 39.12-23.271 65.52-36.612 25.47-12.872 55.05-27.396 88.4-45.244l-22.93-38.105-101.78 58.7-68.924-63.215zM859.42 416.999h-71.842V118.77h71.842z"/><ellipse cx="823.5" cy="41" rx="45.5" ry="41"/><path d="M562.718 418q-37.308 0-61.977-12.41-24.672-12.409-36.705-33.091C452.003 351.817 452 343.743 452 327.59q0-28.363 14.442-49.044 15.044-20.683 43.925-31.909 28.885-11.82 71.004-11.819h76.422q0-23.636-6.619-39-6.62-15.954-20.459-23.636-13.841-7.682-35.502-7.682-24.673 0-42.121 11.227-17.453 10.636-21.663 33.091h-71.006q3.613-31.91 21.663-54.363 18.053-23.046 47.537-35.455 30.087-13 65.59-13 42.723 0 72.807 14.773 30.087 14.181 45.734 41.363Q730 188.728 730 226.546V410.91h-61.377l-7.22-47.864q-6.018 11.819-15.646 21.863-9.024 10.046-21.061 17.727-12.035 7.091-27.679 11.227-15.044 4.138-34.299 4.137m16.246-56.136q17.453 0 30.689-5.91 13.84-6.5 23.468-17.727 10.229-11.817 15.646-25.999 5.415-14.183 7.22-30.137v-1.182h-66.793q-21.06 0-34.899 5.317-13.84 4.73-19.859 14.182-6.015 9.456-6.016 21.865 0 12.409 6.016 21.272 6.02 8.865 17.452 13.592t27.076 4.727M347.361 144.73c11.758-9.086 28.639-.54 28.639 14.498v240.577c0 10.049-7.991 18.195-17.852 18.195H347.19a4.16 4.16 0 0 1-3.016-1.303l-69.965-73.545c-5.775-6.07-15.036-6.935-21.798-2.034l-104.986 76.09a4.15 4.15 0 0 1-2.434.792H46.885c-17.106-.002-24.416-22.157-10.786-32.693z"/><path d="M348.028 5.818c9.296-7.073 22.68.092 21.943 12.195l-4.377 71.729c-.254 4.158-2.288 7.974-5.534 10.441L21.965 357.192C12.975 364.026 0 357.569 0 345.873V277.68c0-4.473 2.072-8.662 5.557-11.315z"/>
-  </g>
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M20.119 1.616C20.590 1.257 21.269 1.620 21.232 2.234L21.010 5.873C20.997 6.084 20.894 6.278 20.729 6.403L3.577 19.441C3.121 19.788 2.462 19.461 2.462 18.867L2.462 15.408C2.462 15.181 2.568 14.968 2.744 14.834Z"/>
+  <path d="M20.085 8.663C20.681 8.202 21.538 8.635 21.538 9.398L21.538 21.603C21.538 22.113 21.132 22.526 20.632 22.526L20.076 22.526A0.211 0.211 0 0 1 19.923 22.460L16.374 18.729C16.081 18.421 15.611 18.377 15.268 18.626L9.942 22.486A0.211 0.211 0 0 1 9.818 22.526L4.841 22.526C3.973 22.526 3.602 21.402 4.294 20.868Z"/>
 </svg>
diff --git a/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml b/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
index 97421518a57..9b8bea586bd 100644
--- a/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
+++ b/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
@@ -1,19 +1,21 @@
-# Source: https://docs.aiand.com/models/catalog/ (USD list prices; accessed
-# 2026-07-24). GET /v1/models reports context_window = 1048576 (the catalog
-# rounds this to "1M" and the shared base model to 1_000_000), so the exact
-# figure is overridden here.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted ("none" produces no reasoning content, "high" does);
-# invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = none|high|max
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high|max → 200; minimal|low|medium|xhigh → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "deepseek/deepseek-v4-flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["none", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.15
 output = 0.25
+cache_read = 0.08
 
 [limit]
 context = 1_048_576
diff --git a/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml b/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
index 8a8cff1112e..88498c5c29a 100644
--- a/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
+++ b/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
@@ -1,18 +1,21 @@
-# Source: https://docs.aiand.com/models/catalog/ (USD list prices; accessed
-# 2026-07-24). GET /v1/models reports context_window = 1048576 (the catalog
-# rounds this to "1M" and the shared base model to 1_000_000), so the exact
-# figure is overridden here.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted; invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = none|high|max
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high|max → 200; low → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "deepseek/deepseek-v4-pro"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["none", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 1
 output = 2.5
+cache_read = 0.25
 
 [limit]
 context = 1_048_576
diff --git a/providers/aiand/models/google/gemma-4-31b-it.toml b/providers/aiand/models/google/gemma-4-31b-it.toml
index b006191cefe..312e02bf19b 100644
--- a/providers/aiand/models/google/gemma-4-31b-it.toml
+++ b/providers/aiand/models/google/gemma-4-31b-it.toml
@@ -1,21 +1,22 @@
-# Source: https://docs.aiand.com/models/catalog/ (USD list prices; accessed
-# 2026-07-24) and live probes against api.aiand.com on the same date.
-# Verified by direct test: image (data URL), video (video_url), and PDF
-# (Files API purpose="document", referenced by file_id) inputs all accepted.
-# ai& converts PDFs to per-page images at upload, so "pdf" is a provider
-# input modality here even though the base model itself is not PDF-native.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted ("none" produces no reasoning content, "high" does);
-# invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = none|high
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high → 200; minimal|low|xhigh → 400.
+# Modalities: image, video, and PDF accepted on this host (2026-07-24 probes).
+# Reasoning side channel: message.reasoning_content
 base_model = "google/gemma-4-31b-it"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["none", "high"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.2
 output = 0.5
+cache_read = 0.05
 
 [modalities]
 input = ["text", "image", "video", "pdf"]
diff --git a/providers/aiand/models/moonshotai/kimi-k2.7-code.toml b/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
index 2e553c932ff..a384aa028b6 100644
--- a/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
@@ -1,19 +1,22 @@
-# Source: https://docs.aiand.com/models/catalog/ (USD list prices; accessed
-# 2026-07-24) and live probes against api.aiand.com on the same date.
-# Verified by direct test: image and PDF (Files API, file_id) inputs are
-# accepted; video is rejected ("does not support video input"), so modalities
-# are overridden from the shared base model's text+image+video.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted; invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = high (always-on; none/low rejected)
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: high → 200; none|low → 400.
+# Modalities: image + PDF accepted; video rejected (2026-07-24 probes).
+# Reasoning side channel: message.reasoning_content
 base_model = "moonshotai/kimi-k2.7-code"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["high"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.75
 output = 3.5
+cache_read = 0.2
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/aiand/models/moonshotai/kimi-k3.toml b/providers/aiand/models/moonshotai/kimi-k3.toml
index 4e7d75f2c83..b28cbd68903 100644
--- a/providers/aiand/models/moonshotai/kimi-k3.toml
+++ b/providers/aiand/models/moonshotai/kimi-k3.toml
@@ -1,22 +1,18 @@
-# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
-# 2026-07-28). Pricing is aiand-specific: $3.00 input, $0.50 cache_read,
-# $12.50 output (differs from Moonshot official rates).
-# Modalities: text + image + pdf accepted per GET /v1/models; video rejected
-# ("does not support video input") — overridden from shared base
-# text+image+video. PDF kept: sibling aiand Moonshot entries (kimi-k2.6,
-# kimi-k2.7-code) include pdf after catalog/probe evidence; aiand treats
-# PDF as a provider-level Files API modality.
-# reasoning_effort: K3 backend accepts only low/high/max per official docs
-# (https://platform.kimi.ai/docs/guide/kimi-k3-quickstart). The aiand gateway
-# schema lists none/minimal/low/medium/high/xhigh/max but K3 rejects
-# none/minimal/medium/xhigh with 400. aiand does not expose a separate off
-# toggle; K3 always reasons.
+# Effort: reasoning_effort = low|high|max (always-on; no none)
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# K3 accepts only low|high|max (lab docs + host 400 on none/minimal/medium/xhigh).
+# Modalities: text + image + PDF; video rejected on this host.
+# Reasoning side channel: message.reasoning_content
 base_model = "moonshotai/kimi-k3"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 3
 output = 12.5
diff --git a/providers/aiand/models/motif-technologies/motif-3.toml b/providers/aiand/models/motif-technologies/motif-3.toml
index b2a69ab0fc4..d0dcf873cc7 100644
--- a/providers/aiand/models/motif-technologies/motif-3.toml
+++ b/providers/aiand/models/motif-technologies/motif-3.toml
@@ -1,18 +1,31 @@
+# Effort: reasoning_effort = none|high
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high → 200; low → 400.
+# First-party Motif host entry (no shared lab base_model in catalog yet).
+# Reasoning side channel: message.reasoning_content
 name = "Motif 3"
 description = "Motif 3 is a large-scale, decoder-only Mixture-of-Experts (MoE) language model with 314 billion total parameters and 13.2 billion parameters activated per token."
 release_date = "2026-08-12"
-last_updated = "2026-08-12"
+last_updated = "2026-09-11"
 attachment = false
 reasoning = true
 temperature = false
 tool_call = false
 structured_output = false
 open_weights = false
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.5
 output = 2
+cache_read = 0.2
 
 [limit]
 context = 262_144
diff --git a/providers/aiand/models/openai/gpt-oss-120b.toml b/providers/aiand/models/openai/gpt-oss-120b.toml
index d7bfcb30588..7083c30e20d 100644
--- a/providers/aiand/models/openai/gpt-oss-120b.toml
+++ b/providers/aiand/models/openai/gpt-oss-120b.toml
@@ -1,14 +1,18 @@
-# Source: https://docs.aiand.com/models/catalog/ (accessed 2026-07-18)
-# reasoning_effort verified by live probe on 2026-07-24: the backend rejects
-# "none", "minimal", and "xhigh" with 400 ("Supported values are: high,
-# medium, low"), so the array is narrowed from the gateway-wide documented
-# set. Invalid values also rejected with 400 (negative control).
+# Effort: reasoning_effort = low|medium|high
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-07-24: low|medium|high → 200; none|minimal|xhigh → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "openai/gpt-oss-120b"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 0.15
 output = 0.60
+cache_read = 0.08
diff --git a/providers/aiand/models/qwen/qwen3.6-27b.toml b/providers/aiand/models/qwen/qwen3.6-27b.toml
index 4859ce77a0b..1f874fc1bc4 100644
--- a/providers/aiand/models/qwen/qwen3.6-27b.toml
+++ b/providers/aiand/models/qwen/qwen3.6-27b.toml
@@ -1,22 +1,21 @@
-# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
-# 2026-07-28). The catalog Quick-picks section still lists this model as free,
-# but the live endpoint reports input = $0.32 / output = $3.20 per 1M tokens,
-# so pricing is taken from the API.
-# ai&'s deployment rejects image input ("does not support image input") and
-# the catalog lists no vision/video/audio capability, so modalities are
-# overridden from the shared base model's text+image+video+audio to text-only,
-# and attachment is set to false to match.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted; invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = none|high
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high → 200; minimal|low|xhigh → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "alibaba/qwen3.6-27b"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["none", "high"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.32
 output = 3.2
+cache_read = 0.2
 
 [modalities]
 input = ["text", "image", "video", "pdf"]
diff --git a/providers/aiand/models/qwen/qwen3.8-27b.toml b/providers/aiand/models/qwen/qwen3.8-27b.toml
index 30b2de2637a..487eb3e6f98 100644
--- a/providers/aiand/models/qwen/qwen3.8-27b.toml
+++ b/providers/aiand/models/qwen/qwen3.8-27b.toml
@@ -1,18 +1,17 @@
-# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
-# 2026-08-29). GET /v1/models reports context_window = 262144, input = $0.40,
-# output = $3.00, cached_input = $0.20 per 1M tokens.
-# Modalities: text, image, video, and document (PDF) supported per catalog
-# capabilities (vision|video|document).
-# reasoning_effort verified by live probe on 2026-08-29: accepts none, low,
-# medium, xhigh (200); rejects minimal, high, max with 400 ("Supported
-# values: none, low, medium, xhigh"). effort=none returns no
-# reasoning_content (off).
+# Effort: reasoning_effort = none|low|medium|xhigh
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-08-29: none|low|medium|xhigh → 200; minimal|high|max → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "alibaba/qwen3.8-27b"
 
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "medium", "xhigh"]
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 0.4
 output = 3
diff --git a/providers/aiand/models/zai-org/glm-5.2.toml b/providers/aiand/models/zai-org/glm-5.2.toml
index 2464bc134c0..ed3b498d41c 100644
--- a/providers/aiand/models/zai-org/glm-5.2.toml
+++ b/providers/aiand/models/zai-org/glm-5.2.toml
@@ -1,18 +1,21 @@
-# Source: https://docs.aiand.com/models/catalog/ (USD list prices; accessed
-# 2026-07-24). GET /v1/models reports context_window = 1048576 (the catalog
-# rounds this to "1M" and the shared base model to 1_000_000), so the exact
-# figure is overridden here.
-# reasoning_effort verified by live probe on 2026-07-24: all six documented
-# values accepted; invalid values rejected with 400 (negative control).
+# Effort: reasoning_effort = none|high|max
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: none|high|max → 200; low|xhigh → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "zhipuai/glm-5.2"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh"]
+values = ["none", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 1
 output = 4
+cache_read = 0.3
 
 [limit]
 context = 1_048_576
diff --git a/providers/aiand/models/zai-org/glm-5.3.toml b/providers/aiand/models/zai-org/glm-5.3.toml
index 9bdd6c737ee..d7d6ad36c6d 100644
--- a/providers/aiand/models/zai-org/glm-5.3.toml
+++ b/providers/aiand/models/zai-org/glm-5.3.toml
@@ -1,16 +1,16 @@
-# Source: GET https://api.aiand.com/v1/models (USD list prices; accessed
-# 2026-08-29). GET /v1/models reports context_window = 1048576 (shared base
-# model uses 1_000_000), so the exact figure is overridden here.
-# Lab / peer relays (zhipuai, zai, OpenRouter, Kilo) use always-on effort
-# low|high|max. ai&'s deployment differs: live reasoning_effort probe on
-# 2026-08-29 accepts none|low|xhigh|max (200); rejects minimal|medium|high
-# with 400 ("Supported values: none, low, xhigh, max"). effort=none returns
-# no reasoning_content (off); xhigh|max return reasoning_content.
+# Effort: reasoning_effort = low|high|max (always-on; no none)
+# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Docs: https://docs.aiand.com/models/catalog/
+# Live probe 2026-09-11: low|high|max → 200; none|xhigh → 400.
+# Reasoning side channel: message.reasoning_content
 base_model = "zhipuai/glm-5.3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "xhigh", "max"]
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 1

From 08ce85fbdf5da373466cf25f283ba01929a59123 Mon Sep 17 00:00:00 2001
From: Divy <divy@coralbricks.ai>
Date: Mon, 14 Sep 2026 09:39:21 -0700
Subject: [PATCH 1287/2012] feat(coralbricks): add GLM 5.3 Flash
 (glm-5.3-flash-fp4) (#6795)

New model live on the CoralBricks gateway: 1M context, $0.15/M in,
$0.50/M out, cached input free (cache_read=0). Inherits the lab
baseline from zhipuai/glm-5.3-flash (always reasons, effort
low|high|max, vision). Verified in the live /v1/models plus chat and
image-input smokes.

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 .../coralbricks/models/glm-5.3-flash-fp4.toml      | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/coralbricks/models/glm-5.3-flash-fp4.toml

diff --git a/providers/coralbricks/models/glm-5.3-flash-fp4.toml b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
new file mode 100644
index 00000000000..199058e48b0
--- /dev/null
+++ b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash FP4"
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+interleaved = true
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0

From 069fe539263af3562de9449b63116eb15fbf0e0f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 11:43:41 -0500
Subject: [PATCH 1288/2012] chore(sync): update Vercel AI Gateway model catalog
 (#6733)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add reasoning options for new models

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <63023139+rekram1-node@users.noreply.github.com>
---
 .../models/alibaba/qwen3.8-flash-next.toml    | 15 -----------
 .../models/bytedance/seed-2.1-turbo.toml      | 25 +++++++++++++++++++
 .../models/deepseek/deepseek-v4.1-flash.toml  | 18 +++++++------
 .../inclusionai/ling-3.0-flash-vl-free.toml   | 24 ++++++++++++++++++
 .../models/inclusionai/ling-3.0-flash-vl.toml | 24 ++++++++++++++++++
 .../nvidia/nemotron-3-nano-30b-a3b.toml       |  3 ++-
 .../models/openai/gpt-oss-safeguard-20b.toml  | 14 +----------
 providers/vercel/models/sakana/fugu-max.toml  | 25 +++++++++++++++++++
 .../vercel/models/sakana/fugu-ultra-v2.toml   | 25 +++++++++++++++++++
 .../spacexai/grok-4.1-fast-reasoning.toml     |  9 +------
 10 files changed, 137 insertions(+), 45 deletions(-)
 delete mode 100644 providers/vercel/models/alibaba/qwen3.8-flash-next.toml
 create mode 100644 providers/vercel/models/bytedance/seed-2.1-turbo.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
 create mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
 create mode 100644 providers/vercel/models/sakana/fugu-max.toml
 create mode 100644 providers/vercel/models/sakana/fugu-ultra-v2.toml

diff --git a/providers/vercel/models/alibaba/qwen3.8-flash-next.toml b/providers/vercel/models/alibaba/qwen3.8-flash-next.toml
deleted file mode 100644
index a905f1db633..00000000000
--- a/providers/vercel/models/alibaba/qwen3.8-flash-next.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "alibaba/qwen3.8-flash-next"
-name = "Qwen 3.8 Flash Next"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "xhigh"] }]
-
-[cost]
-input = 0.12
-output = 0.4
-cache_read = 0.01
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/vercel/models/bytedance/seed-2.1-turbo.toml b/providers/vercel/models/bytedance/seed-2.1-turbo.toml
new file mode 100644
index 00000000000..515539af756
--- /dev/null
+++ b/providers/vercel/models/bytedance/seed-2.1-turbo.toml
@@ -0,0 +1,25 @@
+# Effort: reasoning.effort = none|minimal|low|medium|high|xhigh|max
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+name = "Seed 2.1 Turbo"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "seed"
+release_date = "2026-06-23"
+last_updated = "2026-06-23"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.5
+output = 2.5
+cache_read = 0.1
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
index 3381d473c24..c2b9c23b86c 100644
--- a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
@@ -1,13 +1,15 @@
 # Toggle: reasoning.enabled = true|false
 # Effort: reasoning.effort = high|xhigh (Vercel gateway aliases)
 base_model = "deepseek/deepseek-v4.1-flash"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "xhigh"] }]
 
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh"]
 
-[limit]
-context = 1_048_576
-output = 393_216
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.015
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
new file mode 100644
index 00000000000..63c96c847cf
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
@@ -0,0 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+name = "Ling 3.0 Flash VL (Free)"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "ling"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
new file mode 100644
index 00000000000..3cf827af685
--- /dev/null
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
@@ -0,0 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+name = "Ling 3.0 Flash VL"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "ling"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 256_000
+output = 32_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/vercel/models/nvidia/nemotron-3-nano-30b-a3b.toml
index 6a539219186..3154112a16e 100644
--- a/providers/vercel/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/vercel/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -6,4 +6,5 @@ type = "toggle"
 
 [cost]
 input = 0.05
-output = 0.24
+output = 0.2
+cache_read = 0.025
diff --git a/providers/vercel/models/openai/gpt-oss-safeguard-20b.toml b/providers/vercel/models/openai/gpt-oss-safeguard-20b.toml
index 8e0bda33d18..f90816d0102 100644
--- a/providers/vercel/models/openai/gpt-oss-safeguard-20b.toml
+++ b/providers/vercel/models/openai/gpt-oss-safeguard-20b.toml
@@ -1,14 +1,6 @@
+base_model = "openai/gpt-oss-safeguard-20b"
 name = "gpt-oss-safeguard-20b"
 description = "Safety model for policy screening, moderation, and risk-aware routing workflows"
-family = "gpt-oss"
-release_date = "2025-10-29"
-last_updated = "2024-12-01"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -22,7 +14,3 @@ output = 0.2
 context = 128_000
 input = 112_000
 output = 16_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/vercel/models/sakana/fugu-max.toml b/providers/vercel/models/sakana/fugu-max.toml
new file mode 100644
index 00000000000..a305e311145
--- /dev/null
+++ b/providers/vercel/models/sakana/fugu-max.toml
@@ -0,0 +1,25 @@
+# Effort: reasoning.effort = high|xhigh|max
+# https://console.sakana.ai/models
+name = "Fugu Max"
+description = "Multi-agent model for routing expert agents across complex analytical tasks"
+family = "fugu"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "effort", values = ["high", "xhigh", "max"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/sakana/fugu-ultra-v2.toml b/providers/vercel/models/sakana/fugu-ultra-v2.toml
new file mode 100644
index 00000000000..f77055d8e6e
--- /dev/null
+++ b/providers/vercel/models/sakana/fugu-ultra-v2.toml
@@ -0,0 +1,25 @@
+# Effort: reasoning.effort = high|xhigh|max
+# https://console.sakana.ai/models
+name = "Fugu Ultra v2"
+description = "Quality-first multi-agent model for hard research, analysis, and competitions"
+family = "fugu"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+reasoning_options = [{ type = "effort", values = ["high", "xhigh", "max"] }]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/vercel/models/spacexai/grok-4.1-fast-reasoning.toml b/providers/vercel/models/spacexai/grok-4.1-fast-reasoning.toml
index 68e87906f8a..c29a5484d0e 100644
--- a/providers/vercel/models/spacexai/grok-4.1-fast-reasoning.toml
+++ b/providers/vercel/models/spacexai/grok-4.1-fast-reasoning.toml
@@ -1,12 +1,6 @@
+base_model = "xai/grok-4.1-fast-reasoning"
 name = "Grok 4.1 Fast Reasoning"
 description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-family = "grok"
-release_date = "2025-11-19"
-last_updated = "2025-11-19"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
 reasoning_options = []
 
 [cost]
@@ -20,4 +14,3 @@ output = 1_000_000
 
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]

From c142a1f8d11298643ebf9313fa889415775b3a4a Mon Sep 17 00:00:00 2001
From: Tarun Pandit <tpandit@gmu.edu>
Date: Mon, 14 Sep 2026 12:46:50 -0400
Subject: [PATCH 1289/2012] feat: add Infer provider with Astra and Sol API
 options (#6925)

---
 providers/infer/logo.svg                      |  4 ++++
 .../models/infer/gpt-5.6-sol:official.toml    | 20 +++++++++++++++++++
 .../models/infer/gpt-6-astra:official.toml    | 20 +++++++++++++++++++
 providers/infer/provider.toml                 |  7 +++++++
 4 files changed, 51 insertions(+)
 create mode 100644 providers/infer/logo.svg
 create mode 100644 providers/infer/models/infer/gpt-5.6-sol:official.toml
 create mode 100644 providers/infer/models/infer/gpt-6-astra:official.toml
 create mode 100644 providers/infer/provider.toml

diff --git a/providers/infer/logo.svg b/providers/infer/logo.svg
new file mode 100644
index 00000000000..514621c5477
--- /dev/null
+++ b/providers/infer/logo.svg
@@ -0,0 +1,4 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1254 1254" fill="currentColor">
+  <path d="M393 704 455 538H531L548 446C568 346 656 270 755 270H889L840 424H781C746 424 722 444 715 477L707 538H895Z"/>
+  <path d="m496 714 199-75-77 313H436Z"/>
+</svg>
diff --git a/providers/infer/models/infer/gpt-5.6-sol:official.toml b/providers/infer/models/infer/gpt-5.6-sol:official.toml
new file mode 100644
index 00000000000..9c6273b9563
--- /dev/null
+++ b/providers/infer/models/infer/gpt-5.6-sol:official.toml
@@ -0,0 +1,20 @@
+# Source: https://infer.flow7.org/api/public/catalog (2026-09-12 UTC)
+# Exact selector: infer/gpt-5.6-sol:official
+# Responses request reasoning.effort follows the underlying OpenAI model.
+base_model = "openai/gpt-5.6-sol"
+base_model_omit = ["limit.input"]
+name = "GPT-5.6 Sol (Official API)"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.5
+output = 12.5
+cache_read = 0.25
+cache_write = 3.125
+
+[limit]
+context = 271999
+output = 128000
+
+[provider]
+shape = "responses"
diff --git a/providers/infer/models/infer/gpt-6-astra:official.toml b/providers/infer/models/infer/gpt-6-astra:official.toml
new file mode 100644
index 00000000000..f8c638d5fff
--- /dev/null
+++ b/providers/infer/models/infer/gpt-6-astra:official.toml
@@ -0,0 +1,20 @@
+# Source: https://infer.flow7.org/api/public/catalog (2026-09-12 UTC)
+# Exact selector: infer/gpt-6-astra:official
+# Responses request reasoning.effort follows the underlying OpenAI model.
+base_model = "openai/gpt-6-astra"
+base_model_omit = ["limit.input"]
+name = "GPT-6 Astra (Official API)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 12.5
+output = 62.5
+cache_read = 1.25
+cache_write = 15.625
+
+[limit]
+context = 271999
+output = 128000
+
+[provider]
+shape = "responses"
diff --git a/providers/infer/provider.toml b/providers/infer/provider.toml
new file mode 100644
index 00000000000..327bd75640f
--- /dev/null
+++ b/providers/infer/provider.toml
@@ -0,0 +1,7 @@
+# Docs: https://infer.flow7.org/docs
+# Public catalog: https://infer.flow7.org/api/public/catalog
+name = "Infer by Flow7"
+npm = "@ai-sdk/openai"
+env = ["INFER_API_KEY"]
+api = "https://infer.flow7.org/v1"
+doc = "https://infer.flow7.org/opencode"

From 121daf0e36e14c1afdc50393fd0b15d466708b3f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 17:22:58 +0000
Subject: [PATCH 1290/2012] chore(sync): update OpenRouter model catalog
 (#7073)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e0b3ef90d2e..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From 9590ab20691cb67ee4dd65b1b128ed3a037d9625 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 17:23:02 +0000
Subject: [PATCH 1291/2012] chore(sync): update Kilo model catalog (#7074)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index bd4537b7471..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 393_216
 
 [modalities]

From b7dc138cdd59c8ef15e803aae6e48d7d0679720b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 17:23:06 +0000
Subject: [PATCH 1292/2012] chore(sync): update Merge Gateway model catalog
 (#7075)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 0264c6bc733..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 74867114aac073b3e0011027e440332c7cc5ec4a Mon Sep 17 00:00:00 2001
From: Lee-Si-Yoon <siyoon@friendli.ai>
Date: Tue, 15 Sep 2026 02:41:10 +0900
Subject: [PATCH 1293/2012] feat(sync): replace Friendli generator with
 SyncProvider (#7006)

* feat(sync): replace Friendli generator with SyncProvider

* fix(friendli): tri-state reasoning, tool_call delta, MiniMax budget verification

* fix(friendli): delete removed/deprecated models, drop budget_tokens, align effort with lab

* fix(friendli): restore unbounded budget_tokens per review

* feat(sync): track selectively skipped models

* fix(friendli): fail closed on catalog gaps

---------

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/script/generate-friendli.ts     | 505 ----------------
 packages/core/src/sync/index.ts               |  23 +-
 packages/core/src/sync/missing-issues.ts      |   2 +-
 packages/core/src/sync/providers/friendli.ts  | 560 ++++++++++++++++++
 packages/core/test/friendli.test.ts           |  41 ++
 packages/core/test/missing-skips.test.ts      |  50 ++
 .../models/MiniMaxAI/MiniMax-M2.5.toml        |  10 +-
 .../models/deepseek-ai/DeepSeek-V3.2.toml     |  28 +-
 .../models/google/gemma-4-31B-it.toml         |  14 +-
 .../friendli/models/zai-org/GLM-5.1.toml      |  16 +-
 .../friendli/models/zai-org/GLM-5.2.toml      |  25 +-
 .../models/zai-org/GLM-5.3-Flash.toml         |  26 +
 .../friendli/models/zai-org/GLM-5.3.toml      |  35 +-
 sync.md                                       |   3 +
 14 files changed, 766 insertions(+), 572 deletions(-)
 delete mode 100644 packages/core/script/generate-friendli.ts
 create mode 100644 packages/core/src/sync/providers/friendli.ts
 create mode 100644 packages/core/test/friendli.test.ts
 create mode 100644 packages/core/test/missing-skips.test.ts
 create mode 100644 providers/friendli/models/zai-org/GLM-5.3-Flash.toml

diff --git a/packages/core/script/generate-friendli.ts b/packages/core/script/generate-friendli.ts
deleted file mode 100644
index d048f3d8369..00000000000
--- a/packages/core/script/generate-friendli.ts
+++ /dev/null
@@ -1,505 +0,0 @@
-#!/usr/bin/env bun
-
-import { mkdir } from "node:fs/promises";
-import path from "node:path";
-import { z } from "zod";
-
-import { inferKimiFamily } from "../src/family.js";
-
-// Friendli API endpoint
-const API_ENDPOINT = "https://api.friendli.ai/serverless/v1/models";
-
-// Zod schemas for API response validation
-const Functionality = z.object({
-  tool_call: z.boolean(),
-  parallel_tool_call: z.boolean(),
-  structured_output: z.boolean(),
-});
-
-const Pricing = z.object({
-  input: z.number(),
-  output: z.number(),
-  response_time: z.number(),
-  unit_type: z.enum(["TOKEN", "SECOND"]),
-});
-
-const FriendliModel = z
-  .object({
-    id: z.string(),
-    name: z.string(),
-    max_completion_tokens: z.number(),
-    context_length: z.number(),
-    functionality: Functionality,
-    pricing: Pricing,
-    hugging_face_url: z.string().optional(),
-    description: z.string().optional(),
-    license: z.string().optional(),
-    policy: z.string().optional().nullable(),
-    created: z.number(), // Unix timestamp
-  })
-  .passthrough();
-
-const FriendliResponse = z.object({
-  data: z.array(FriendliModel),
-});
-
-// Family inference patterns
-const familyPatterns: [RegExp, string][] = [
-  [/qwen3/i, "qwen3"],
-  [/deepseek-r1/i, "deepseek-r1"],
-  [/glm-4/i, "glm-4"],
-  [/glm-5/i, "glm"],
-];
-
-function inferFamily(modelId: string, modelName: string): string | undefined {
-  const kimiFamily = inferKimiFamily(modelId, modelName);
-  if (kimiFamily !== undefined) return kimiFamily;
-
-  for (const [pattern, family] of familyPatterns) {
-    if (pattern.test(modelId) || pattern.test(modelName)) {
-      return family;
-    }
-  }
-  return undefined;
-}
-
-function extractModelName(fullName: string): string {
-  // "meta-llama/Llama-3.3-70B-Instruct" -> "Llama 3.3 70B Instruct"
-  const parts = fullName.split("/");
-  const modelName = parts.at(-1) ?? fullName;
-  return modelName
-    .replace(/-/g, " ")
-    .replace(/\b\w/g, (l) => l.toUpperCase());
-}
-
-// TODO: Replace with functionality.parse_reasoning from API when available
-function isReasoningModel(modelId: string): boolean {
-  const nonReasoningPatterns = [
-    /qwen3.*instruct/i,
-  ];
-
-  for (const pattern of nonReasoningPatterns) {
-    if (pattern.test(modelId)) {
-      return false;
-    }
-  }
-
-  // Everything else is reasoning or hybrid reasoning
-  return true;
-}
-
-function formatNumber(n: number): string {
-  if (n >= 1000) {
-    // Format with underscores for readability (e.g., 131_072)
-    return n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, "_");
-  }
-  return n.toString();
-}
-
-function timestampToDate(timestamp: number): string {
-  const date = new Date(timestamp * 1000);
-  return date.toISOString().slice(0, 10);
-}
-
-function getTodayDate(): string {
-  return new Date().toISOString().slice(0, 10);
-}
-
-interface ExistingModel {
-  name?: string;
-  family?: string;
-  attachment?: boolean;
-  reasoning?: boolean;
-  tool_call?: boolean;
-  structured_output?: boolean;
-  temperature?: boolean;
-  knowledge?: string;
-  release_date?: string;
-  last_updated?: string;
-  open_weights?: boolean;
-  interleaved?: boolean | { field: string };
-  status?: string;
-  cost?: {
-    input?: number;
-    output?: number;
-    reasoning?: number;
-    cache_read?: number;
-    cache_write?: number;
-  };
-  limit?: {
-    context?: number;
-    input?: number;
-    output?: number;
-  };
-  modalities?: {
-    input?: string[];
-    output?: string[];
-  };
-  provider?: {
-    npm?: string;
-    api?: string;
-  };
-}
-
-async function loadExistingModel(
-  filePath: string,
-): Promise<ExistingModel | null> {
-  try {
-    const file = Bun.file(filePath);
-    if (!(await file.exists())) {
-      return null;
-    }
-    const toml = await import(filePath, { with: { type: "toml" } }).then(
-      (mod) => mod.default,
-    );
-    return toml as ExistingModel;
-  } catch (e) {
-    console.warn(`Warning: Failed to parse existing file ${filePath}:`, e);
-    return null;
-  }
-}
-
-interface MergedModel {
-  name: string;
-  family?: string;
-  attachment: boolean;
-  reasoning: boolean;
-  tool_call: boolean;
-  structured_output?: boolean;
-  temperature: boolean;
-  knowledge?: string;
-  release_date: string;
-  last_updated: string;
-  open_weights: boolean;
-  interleaved?: boolean | { field: string };
-  status?: string;
-  cost?: {
-    input: number;
-    output: number;
-  };
-  limit: {
-    context: number;
-    output: number;
-  };
-  modalities: {
-    input: string[];
-    output: string[];
-  };
-}
-
-function mergeModel(
-  apiModel: z.infer<typeof FriendliModel>,
-  existing: ExistingModel | null,
-): MergedModel {
-  const contextTokens = apiModel.context_length;
-  const outputTokens = apiModel.max_completion_tokens;
-
-  const openWeights = Boolean(apiModel.hugging_face_url);
-
-  const merged: MergedModel = {
-    // Always from API
-    name: extractModelName(apiModel.name),
-    attachment: false, // All Friendli models are text-only currently
-    reasoning: isReasoningModel(apiModel.id),
-    tool_call: apiModel.functionality.tool_call,
-    temperature: true,
-    release_date: timestampToDate(apiModel.created),
-    last_updated: getTodayDate(),
-    open_weights: openWeights,
-    limit: {
-      context: contextTokens,
-      output: outputTokens,
-    },
-    modalities: {
-      input: ["text"],
-      output: ["text"],
-    },
-  };
-
-  // structured_output only if true
-  if (apiModel.functionality.structured_output === true) {
-    merged.structured_output = true;
-  }
-
-  // Cost from API - ONLY include if unit_type is TOKEN
-  if (apiModel.pricing.unit_type === "TOKEN") {
-    merged.cost = {
-      input: apiModel.pricing.input,
-      output: apiModel.pricing.output,
-    };
-  } else {
-    console.log(
-      `  Note: ${apiModel.id} uses ${apiModel.pricing.unit_type} pricing - cost section omitted`,
-    );
-  }
-
-  // Preserve from existing OR infer
-  if (existing?.family) {
-    merged.family = existing.family;
-  } else {
-    const inferred = inferFamily(apiModel.id, apiModel.name);
-    if (inferred) {
-      merged.family = inferred;
-    }
-  }
-
-  // Preserve manual fields from existing
-  if (existing?.knowledge) {
-    merged.knowledge = existing.knowledge;
-  }
-  if (existing?.interleaved !== undefined) {
-    merged.interleaved = existing.interleaved;
-  }
-  if (existing?.status !== undefined) {
-    merged.status = existing.status;
-  }
-
-  return merged;
-}
-
-function formatToml(model: MergedModel): string {
-  const lines: string[] = [];
-
-  // Basic fields
-  lines.push(`name = "${model.name.replace(/"/g, '\\"')}"`);
-  if (model.family) {
-    lines.push(`family = "${model.family}"`);
-  }
-  lines.push(`attachment = ${model.attachment}`);
-  lines.push(`reasoning = ${model.reasoning}`);
-  lines.push(`tool_call = ${model.tool_call}`);
-  if (model.structured_output !== undefined) {
-    lines.push(`structured_output = ${model.structured_output}`);
-  }
-  lines.push(`temperature = ${model.temperature}`);
-  if (model.knowledge) {
-    lines.push(`knowledge = "${model.knowledge}"`);
-  }
-  lines.push(`release_date = "${model.release_date}"`);
-  lines.push(`last_updated = "${model.last_updated}"`);
-  lines.push(`open_weights = ${model.open_weights}`);
-  if (model.status) {
-    lines.push(`status = "${model.status}"`);
-  }
-
-  // Interleaved section (if present)
-  if (model.interleaved !== undefined) {
-    lines.push("");
-    if (model.interleaved === true) {
-      lines.push(`interleaved = true`);
-    } else if (typeof model.interleaved === "object") {
-      lines.push(`[interleaved]`);
-      lines.push(`field = "${model.interleaved.field}"`);
-    }
-  }
-
-  // Cost section (only if present)
-  if (model.cost) {
-    lines.push("");
-    lines.push(`[cost]`);
-    lines.push(`input = ${model.cost.input}`);
-    lines.push(`output = ${model.cost.output}`);
-  }
-
-  // Limit section
-  lines.push("");
-  lines.push(`[limit]`);
-  lines.push(`context = ${formatNumber(model.limit.context)}`);
-  lines.push(`output = ${formatNumber(model.limit.output)}`);
-
-  // Modalities section
-  lines.push("");
-  lines.push(`[modalities]`);
-  lines.push(
-    `input = [${model.modalities.input.map((m) => `"${m}"`).join(", ")}]`,
-  );
-  lines.push(
-    `output = [${model.modalities.output.map((m) => `"${m}"`).join(", ")}]`,
-  );
-
-  return lines.join("\n") + "\n";
-}
-
-interface Changes {
-  field: string;
-  oldValue: string;
-  newValue: string;
-}
-
-function detectChanges(
-  existing: ExistingModel | null,
-  merged: MergedModel,
-): Changes[] {
-  if (!existing) return [];
-
-  const changes: Changes[] = [];
-
-  const compare = (field: string, oldVal: unknown, newVal: unknown) => {
-    const oldStr = JSON.stringify(oldVal);
-    const newStr = JSON.stringify(newVal);
-    if (oldStr !== newStr) {
-      changes.push({
-        field,
-        oldValue: formatValue(oldVal),
-        newValue: formatValue(newVal),
-      });
-    }
-  };
-
-  const formatValue = (val: unknown): string => {
-    if (typeof val === "number") return formatNumber(val);
-    if (Array.isArray(val)) return `[${val.join(", ")}]`;
-    if (val === undefined) return "(none)";
-    return String(val);
-  };
-
-  compare("name", existing.name, merged.name);
-  compare("family", existing.family, merged.family);
-  compare("attachment", existing.attachment, merged.attachment);
-  compare("reasoning", existing.reasoning, merged.reasoning);
-  compare("tool_call", existing.tool_call, merged.tool_call);
-  compare(
-    "structured_output",
-    existing.structured_output,
-    merged.structured_output,
-  );
-  compare("open_weights", existing.open_weights, merged.open_weights);
-  compare("release_date", existing.release_date, merged.release_date);
-  compare("cost.input", existing.cost?.input, merged.cost?.input);
-  compare("cost.output", existing.cost?.output, merged.cost?.output);
-  compare("limit.context", existing.limit?.context, merged.limit.context);
-  compare("limit.output", existing.limit?.output, merged.limit.output);
-  compare("modalities.input", existing.modalities?.input, merged.modalities.input);
-
-  return changes;
-}
-
-async function main() {
-  const args = process.argv.slice(2);
-  const dryRun = args.includes("--dry-run");
-
-  const modelsDir = path.join(
-    import.meta.dirname,
-    "..",
-    "..",
-    "..",
-    "providers",
-    "friendli",
-    "models",
-  );
-
-  if (dryRun) {
-    console.log(`[DRY RUN] Fetching Friendli models from API...`);
-  } else {
-    console.log(`Fetching Friendli models from API...`);
-  }
-
-  // Fetch API data
-  const res = await fetch(API_ENDPOINT);
-  if (!res.ok) {
-    console.error(`Failed to fetch API: ${res.status} ${res.statusText}`);
-    process.exit(1);
-  }
-
-  const json = await res.json();
-  const parsed = FriendliResponse.safeParse(json);
-  if (!parsed.success) {
-    console.error("Invalid API response:", parsed.error.errors);
-    process.exit(1);
-  }
-
-  const apiModels = parsed.data.data;
-
-  // Get existing files (recursively)
-  const existingFiles = new Set<string>();
-  try {
-    for await (const file of new Bun.Glob("**/*.toml").scan({
-      cwd: modelsDir,
-      absolute: false,
-    })) {
-      existingFiles.add(file);
-    }
-  } catch {
-    // Directory might not exist yet
-  }
-
-  console.log(
-    `Found ${apiModels.length} models in API, ${existingFiles.size} existing files\n`,
-  );
-
-  // Track API model IDs for orphan detection
-  const apiModelIds = new Set<string>();
-
-  let created = 0;
-  let updated = 0;
-  let unchanged = 0;
-
-  for (const apiModel of apiModels) {
-    const relativePath = `${apiModel.id}.toml`;
-    const filePath = path.join(modelsDir, relativePath);
-    const dirPath = path.dirname(filePath);
-
-    apiModelIds.add(relativePath);
-
-    const existing = await loadExistingModel(filePath);
-    const merged = mergeModel(apiModel, existing);
-    const tomlContent = formatToml(merged);
-
-    if (existing === null) {
-      created++;
-      if (dryRun) {
-        console.log(`[DRY RUN] Would create: ${relativePath}`);
-        console.log(`  name = "${merged.name}"`);
-        if (merged.family) {
-          console.log(`  family = "${merged.family}" (inferred)`);
-        }
-        console.log("");
-      } else {
-        await mkdir(dirPath, { recursive: true });
-        await Bun.write(filePath, tomlContent);
-        console.log(`Created: ${relativePath}`);
-      }
-    } else {
-      const changes = detectChanges(existing, merged);
-
-      if (changes.length > 0) {
-        updated++;
-        if (dryRun) {
-          console.log(`[DRY RUN] Would update: ${relativePath}`);
-        } else {
-          await Bun.write(filePath, tomlContent);
-          console.log(`Updated: ${relativePath}`);
-        }
-        for (const change of changes) {
-          console.log(`  ${change.field}: ${change.oldValue} → ${change.newValue}`);
-        }
-        console.log("");
-      } else {
-        unchanged++;
-      }
-    }
-  }
-
-  // Check for orphaned files
-  const orphaned: string[] = [];
-  for (const file of existingFiles) {
-    if (!apiModelIds.has(file)) {
-      orphaned.push(file);
-      console.log(`Warning: Orphaned file (not in API): ${file}`);
-    }
-  }
-
-  // Summary
-  console.log("");
-  if (dryRun) {
-    console.log(
-      `Summary: ${created} would be created, ${updated} would be updated, ${unchanged} unchanged, ${orphaned.length} orphaned`,
-    );
-  } else {
-    console.log(
-      `Summary: ${created} created, ${updated} updated, ${unchanged} unchanged, ${orphaned.length} orphaned`,
-    );
-  }
-}
-
-await main();
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index e6c30874235..39d87a5817e 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -18,6 +18,7 @@ import { deepinfra } from "./providers/deepinfra.js";
 import { digitalocean } from "./providers/digitalocean.js";
 import { edenai } from "./providers/edenai.js";
 import { empiriolabs } from "./providers/empiriolabs.js";
+import { friendli } from "./providers/friendli.js";
 import { githubCopilot } from "./providers/github-copilot.js";
 import { google } from "./providers/google.js";
 import { hyper } from "./providers/hyper.js";
@@ -96,6 +97,12 @@ export interface SyncProvider<SourceModel> {
    * undefined to skip silently (no notice, no missing-model issue).
    */
   sourceID?(model: SourceModel): string | undefined;
+  /**
+   * Return the ID when a source model skipped by translateModel needs a
+   * missing-model issue. Existing local metadata for that ID is preserved.
+   * Return undefined for intentional skips.
+   */
+  missingModelID?(model: SourceModel): string | undefined;
   skippedNotice?(ids: string[]): string[];
   fetchModels(): Promise<unknown>;
   parseModels(raw: unknown): SourceModel[];
@@ -143,6 +150,7 @@ export const providers: {
   digitalocean: SyncProvider<any>;
   edenai: SyncProvider<any>;
   empiriolabs: SyncProvider<any>;
+  friendli: SyncProvider<any>;
   "github-copilot": SyncProvider<any>;
   google: SyncProvider<any>;
   hyper: SyncProvider<any>;
@@ -179,6 +187,7 @@ export const providers: {
   digitalocean,
   edenai,
   empiriolabs,
+  friendli,
   "github-copilot": githubCopilot,
   google,
   hyper,
@@ -222,7 +231,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "friendli", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
@@ -255,6 +264,7 @@ export async function syncProvider<SourceModel>(
   const caseNormalizedDesiredPaths = new Map<string, string>();
   const desiredMetadata = new Map<string, { model: z.infer<typeof ModelMetadata>; content: string }>();
   const skippedRemote: string[] = [];
+  const missingRemote = new Set<string>();
   const missingReasoning = new Map<string, string>();
 
   for (const sourceModel of sourceModels) {
@@ -277,6 +287,8 @@ export async function syncProvider<SourceModel>(
     if (translated === undefined) {
       const skippedID = provider.sourceID?.(sourceModel);
       if (skippedID !== undefined) skippedRemote.push(skippedID);
+      const missingID = provider.missingModelID?.(sourceModel);
+      if (missingID !== undefined) missingRemote.add(missingID);
       continue;
     }
 
@@ -456,6 +468,10 @@ export async function syncProvider<SourceModel>(
   const missingLocal: string[] = [];
   for (const relativePath of new Set([...existing.keys(), ...brokenSymlinks])) {
     if (desired.has(relativePath)) continue;
+    if (missingRemote.has(relativePath.slice(0, -5))) {
+      unchanged++;
+      continue;
+    }
     if (missingReasoning.has(relativePath.slice(0, -5))) {
       unchanged++;
       continue;
@@ -487,10 +503,11 @@ export async function syncProvider<SourceModel>(
     ...provider.missingNotice?.(missingLocal) ?? [],
   ];
 
-  const issueModels = [
+  const issueModels = [...new Set([
+    ...missingRemote.values(),
     ...(provider.skipCreates === true ? skippedRemote : []),
     ...missingReasoning.keys(),
-  ];
+  ])];
   if (
     provider.trackMissingModels !== false
     && issueModels.length > 0
diff --git a/packages/core/src/sync/missing-issues.ts b/packages/core/src/sync/missing-issues.ts
index 81a01a3742e..3e841f4aec0 100644
--- a/packages/core/src/sync/missing-issues.ts
+++ b/packages/core/src/sync/missing-issues.ts
@@ -26,7 +26,7 @@ function issueBody(provider: MissingModelIssueTarget, modelId: string, reason?:
     `| Expected path | \`${provider.modelsDir}/${modelId}.toml\` |`,
     "",
     reason === undefined
-      ? "This provider uses `skipCreates` because the remote source is not enough to auto-author a full TOML."
+      ? "Automatic creation was skipped because the remote source is not enough to auto-author a complete catalog entry."
       : `Sync diagnostic: ${reason}`,
     "Add the model manually (prefer `base_model` when matching `models/` metadata exists).",
     ...(reason === undefined ? [] : [
diff --git a/packages/core/src/sync/providers/friendli.ts b/packages/core/src/sync/providers/friendli.ts
new file mode 100644
index 00000000000..d021f8ed484
--- /dev/null
+++ b/packages/core/src/sync/providers/friendli.ts
@@ -0,0 +1,560 @@
+import path from "node:path";
+import { readdirSync } from "node:fs";
+import { z } from "zod";
+
+import { describeModel } from "../../describe.js";
+import { inferKimiFamily, ModelFamilyValues } from "../../family.js";
+import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
+import { factorBaseModel } from "./openrouter.js";
+
+const API_ENDPOINT = "https://api.friendli.ai/serverless/v1/models";
+const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+
+// Friendli catalog pricing is USD per-token; catalog cost is USD per-million.
+const PER_TOKEN_TO_PER_MILLION = 1_000_000;
+
+const InterleavedField = z.enum(["reasoning_content", "reasoning_details"]);
+
+// Friendli's /v1/models `interleaved` flag is unreliable for some models: it
+// reports `false` for deepseek-ai/DeepSeek-V3.2 even though a live
+// POST /chat/completions request (chat_template_kwargs.enable_thinking=true)
+// returns both `reasoning` and `reasoning_content` in the response. Relying on
+// `existing?.interleaved` to carry this forward is fragile — if the on-disk
+// file ever loses the field for any reason, the live verification is silently
+// forgotten on the next sync with no trace. This map is the durable source of
+// truth for models where a live request has verified a real field the
+// catalog API misreports; translateInterleaved consults it before falling
+// back to the existing on-disk value.
+const VERIFIED_INTERLEAVED_OVERRIDES: Record<string, SyncedFullModel["interleaved"]> = {
+  "deepseek-ai/DeepSeek-V3.2": { field: "reasoning_content" },
+};
+
+// Raw API reasoning_options shape, including budget_tokens (a real
+// reasoning-budget control on Friendli: min = -1 means unlimited, max
+// corresponds to max_completion_tokens). Confirmed via the live /v1/models
+// response and https://friendli.ai/docs/openapi/model-apis/chat-completions
+// (reasoning_budget is a documented request field). The catalog's min/max
+// are not safe published bounds (see translateReasoningOptions below), so
+// they are parsed but never carried into the synced model.
+const FriendliReasoningOption = z
+  .discriminatedUnion("type", [
+    z.object({ type: z.literal("toggle") }).passthrough(),
+    z
+      .object({ type: z.literal("effort"), values: z.array(z.string()) })
+      .passthrough(),
+    z
+      .object({
+        type: z.literal("budget_tokens"),
+        min: z.number().optional(),
+        max: z.number().optional(),
+      })
+      .passthrough(),
+  ])
+  .optional();
+
+export const FriendliModel = z
+  .object({
+    id: z.string(),
+    hugging_face_id: z.string().optional(),
+    name: z.string(),
+    created: z.number(),
+    context_length: z.number(),
+    max_completion_tokens: z.number(),
+    functionality: z
+      .object({
+        tool_call: z.boolean(),
+        parallel_tool_call: z.boolean().optional(),
+        structured_output: z.boolean(),
+        tool_choice: z.boolean().optional(),
+        system_messages: z.boolean().optional(),
+      })
+      .passthrough(),
+    pricing: z
+      .object({
+        input: z.union([z.string(), z.number()]),
+        output: z.union([z.string(), z.number()]),
+        prompt: z.union([z.string(), z.number()]).optional(),
+        completion: z.union([z.string(), z.number()]).optional(),
+        input_cache_read: z.union([z.string(), z.number()]).optional(),
+        input_cache_write: z.union([z.string(), z.number()]).optional(),
+        // The pre-SyncProvider generator validated this field and authored
+        // cost only for TOKEN pricing. The current catalog always omits it
+        // for the 7 live models, but Friendli has served SECOND-priced
+        // entries before — passthrough would silently x1,000,000 a
+        // per-second rate into the catalog's USD/MTok cost.
+        unit_type: z.enum(["TOKEN", "SECOND"]).optional(),
+      })
+      .passthrough(),
+    description: z.string().optional(),
+    hugging_face_url: z.string().optional(),
+    license: z.string().optional(),
+    policy: z.string().nullable().optional(),
+    deprecation_date: z.string().nullable().optional(),
+    reasoning: z.boolean().optional(),
+    reasoning_options: z.array(FriendliReasoningOption).optional(),
+    interleaved: z.union([InterleavedField, z.boolean()]).optional(),
+    input_modalities: z.array(z.string()).optional(),
+    output_modalities: z.array(z.string()).optional(),
+    base_model: z.string().optional(),
+    mode: z.string().optional(),
+  })
+  .passthrough();
+
+export const FriendliResponse = z
+  .object({
+    data: z.array(FriendliModel),
+  })
+  .passthrough();
+
+export type FriendliModel = z.infer<typeof FriendliModel>;
+
+// HuggingFace-style API orgs that are not catalog lab ids. Map them onto the
+// catalog metadata tree so self-referential or HF-style base_model values
+// resolve to the right lab directory.
+const LAB_PREFIX_MAP: Record<string, string> = {
+  "zai-org": "zhipuai",
+  "deepseek-ai": "deepseek",
+  "LGAI-EXAONE": "lgai-exaone",
+  "MiniMaxAI": "minimax",
+  "meta-llama": "meta",
+  "mistralai": "mistral",
+  "Qwen": "alibaba",
+};
+
+// Resolve an API `base_model` id to the on-disk `models/<lab>/<file>.toml` id.
+// Friendli declares a base_model for most entries, but only models with an
+// existing lab metadata file can be factored (override-only). Self-referential
+// base_model values (==id) resolve to the model's own lab id when a metadata
+// file exists under the mapped lab prefix.
+//
+// Case-insensitive lookup: the API lowercases some ids (e.g.
+// "minimax/minimax-m2.5") that exist on disk as mixed-case
+// ("minimax/MiniMax-M2.5.toml"), so we never trust a raw API id and always
+// read the directory.
+const baseModelCache = new Map<string, string | null>();
+
+function resolveBaseModelID(baseModel: string | undefined): string | undefined {
+  if (baseModel === undefined || baseModel.length === 0) return undefined;
+  const cached = baseModelCache.get(baseModel);
+  if (cached !== undefined) return cached ?? undefined;
+
+  let resolved = lookupLabFile(baseModel);
+  if (resolved === undefined) {
+    const [org, ...parts] = baseModel.split("/");
+    const mapped = org !== undefined ? LAB_PREFIX_MAP[org] : undefined;
+    if (mapped !== undefined && parts.length > 0) {
+      resolved = lookupLabFile(`${mapped}/${parts.join("/")}`);
+    }
+  }
+
+  baseModelCache.set(baseModel, resolved ?? null);
+  return resolved;
+}
+
+// Resolve a Friendli entry to its catalog lab metadata id.
+// 1) API-declared base_model (handles HF id → catalog slug mismatches)
+// 2) self-referential fallback: some entries (e.g. deepseek-ai/DeepSeek-V3.2)
+//    omit base_model entirely even though a matching lab metadata file
+//    exists under the mapped lab prefix — resolve against the model's own id.
+function resolveLabModelSync(model: FriendliModel): string | undefined {
+  return resolveBaseModelID(model.base_model) ?? resolveBaseModelID(model.id);
+}
+
+function lookupLabFile(baseModel: string): string | undefined {
+  const [lab, ...modelParts] = baseModel.split("/");
+  const modelSlug = modelParts.join("/");
+  if (lab === undefined || modelSlug.length === 0) return undefined;
+
+  let labDir: string | undefined;
+  try {
+    const dirs = readdirSync(MODELS_DIR, { withFileTypes: true })
+      .filter((entry) => entry.isDirectory())
+      .map((entry) => entry.name);
+    labDir = dirs.find((dir) => dir.toLowerCase() === lab.toLowerCase());
+  } catch {
+    return undefined;
+  }
+  if (labDir === undefined) return undefined;
+
+  const expected = `${modelSlug}.toml`.toLowerCase();
+  let fileMatch: string | undefined;
+  try {
+    fileMatch = readdirSync(path.join(MODELS_DIR, labDir))
+      .filter((file) => file.endsWith(".toml"))
+      .find((file) => file.toLowerCase() === expected);
+  } catch {
+    // fall through
+  }
+  if (fileMatch === undefined) return undefined;
+
+  return `${labDir}/${fileMatch.slice(0, -".toml".length)}`;
+}
+
+export const friendli = {
+  id: "friendli",
+  name: "Friendli",
+  modelsDir: "providers/friendli/models",
+  // Friendli's /v1/models is authoritative for what this host serves: a model
+  // absent from the catalog (or past its deprecation_date) must not stay in
+  // the catalog as a live-looking route, so missing files are deleted rather
+  // than retained. Deprecation marking below only applies while the model is
+  // still listed; once it disappears, the file goes with it.
+  deleteMissing: true,
+  // Friendli's catalog describes real reasoning controls and limits directly;
+  // do not carry over a stale base_model when a model switches lab → full inline.
+  preserveBaseModels: false,
+  // The runner's default preserveDescription re-injects the resolved base
+  // description when the translator omits it, recreating an identical
+  // override. Friendli descriptions come from the API verbatim and match the
+  // lab's, so drop the re-injection.
+  preserveDescriptions: false,
+  // Leading wire-path comments (Toggle/Effort/Budget + doc URLs) always
+  // refresh from reasoningHeader() below instead of freezing whatever
+  // comment happened to be on disk the first time a file was created.
+  authoritativeHeaders: true,
+  async fetchModels() {
+    const response = await fetch(API_ENDPOINT);
+    if (!response.ok) {
+      throw new Error(
+        `Friendli request failed: ${response.status} ${response.statusText}`,
+      );
+    }
+    return response.json();
+  },
+  parseModels(raw: unknown) {
+    const models = FriendliResponse.parse(raw).data;
+    if (models.length === 0) {
+      throw new Error("Friendli returned an empty model catalog; refusing destructive sync");
+    }
+    return models;
+  },
+  translateModel(model: FriendliModel, context) {
+    const existing = context.existing(model.id);
+    const authored = context.authored(model.id);
+    // A model past its deprecation_date is skipped outright (tracked or not):
+    // with deleteMissing enabled, skipping removes an already-tracked file on
+    // the next sync, so the catalog never keeps serving a dead route as a
+    // live-looking entry. Source-of-truth policy: a deprecation_date in the
+    // catalog means the same thing as the model disappearing from it.
+    if (isDeprecated(model)) return undefined;
+    const factorBase = resolveLabModelSync(model);
+    // Friendli is a multi-lab relay, so models that need a canonical lab entry
+    // are handled by the missing-model issue flow. If an existing factored
+    // entry becomes temporarily unresolvable, skip it as well: the runner
+    // preserves its TOML rather than expanding or deleting it. Existing true
+    // host-unique full-inline entries can still update normally.
+    if (
+      factorBase === undefined
+      && (existing === undefined || authored?.base_model !== undefined)
+    ) return undefined;
+    const built = buildFriendliModel(
+      model,
+      existing,
+      factorBase,
+    );
+    return {
+      id: model.id,
+      model: built,
+      header: reasoningHeader(built),
+    };
+  },
+  sourceID(model: FriendliModel) {
+    return model.id;
+  },
+  missingModelID(model: FriendliModel) {
+    // Active models only reach the skip path when their provider-agnostic lab
+    // metadata is missing. Deprecated models are intentional removals.
+    return isDeprecated(model) ? undefined : model.id;
+  },
+  skippedNotice(ids: string[]) {
+    if (ids.length === 0) return [];
+    return [
+      `${ids.length} remote model(s) skipped: no provider-agnostic lab metadata to factor onto (full-inline creates are not authored for a multi-lab relay — add models/<lab>/<model>.toml, then re-sync) or deprecation_date passed: ${ids.join(", ")}`,
+    ];
+  },
+  missingNotice(paths: string[]) {
+    if (paths.length === 0) return [];
+    return [
+      `${paths.length} local model(s) deleted after being removed from the Friendli API (or past their deprecation_date): ${paths.join(", ")}`,
+    ];
+  },
+} satisfies SyncProvider<FriendliModel>;
+
+// Leading wire-path comments for every reasoning control type this host
+// authors on a file, matching the wire paths documented in
+// providers/friendli/provider.toml. With authoritativeHeaders enabled, this
+// header always replaces whatever was on disk, so it never goes stale.
+const REASONING_GUIDE_URL = "https://friendli.ai/docs/guides/reasoning";
+const EFFORT_DOC_URL =
+  "https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0";
+const BUDGET_DOC_URL =
+  "https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0";
+
+function reasoningHeader(model: SyncedModel): string | undefined {
+  const options = model.reasoning_options;
+  if (options === undefined || options.length === 0) return undefined;
+  const lines: string[] = [];
+  for (const option of options) {
+    if (option.type === "toggle") {
+      lines.push("# Toggle: chat_template_kwargs.enable_thinking = true | false");
+      lines.push(`# ${REASONING_GUIDE_URL}`);
+    }
+    if (option.type === "effort") {
+      if (option.values.length > 0) {
+        const values = option.values.map((value) => `"${value}"`).join(" | ");
+        lines.push(`# Effort: reasoning_effort = ${values}`);
+      } else {
+        lines.push("# Effort: reasoning_effort (model-specific accepted values)");
+      }
+      lines.push(`# ${EFFORT_DOC_URL}`);
+    }
+    if (option.type === "budget_tokens") {
+      lines.push(
+        "# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)",
+      );
+      lines.push(`# ${BUDGET_DOC_URL}`);
+    }
+  }
+  return lines.length > 0 ? `${lines.join("\n")}\n` : undefined;
+}
+
+type Modality = "text" | "audio" | "image" | "video" | "pdf";
+
+const ALLOWED_MODALITIES: Record<string, true> = {
+  text: true,
+  audio: true,
+  image: true,
+  video: true,
+  pdf: true,
+};
+
+function translateModalities(values: string[] | undefined): Modality[] {
+  const result = [...new Set(
+    (values ?? ["text"])
+      .map((value) => value.toLowerCase())
+      .filter((value): value is Modality => ALLOWED_MODALITIES[value] === true),
+  )];
+  return result.length > 0 ? result : ["text"];
+}
+
+// Skip models whose deprecation_date has passed. Friendli returns an ISO
+// timestamp (e.g. "2026-08-20T00:00:00Z"); we compare against now at sync time.
+function isDeprecated(model: FriendliModel): boolean {
+  if (model.deprecation_date === undefined || model.deprecation_date === null) return false;
+  const dep = Date.parse(model.deprecation_date);
+  return Number.isFinite(dep) && dep <= Date.now();
+}
+
+function perMillion(value: string | number | undefined): number | undefined {
+  if (value === undefined) return undefined;
+  const number = Number(value);
+  if (!Number.isFinite(number) || number < 0) return undefined;
+  const perM = number * PER_TOKEN_TO_PER_MILLION;
+  return Math.round(perM * 1_000_000) / 1_000_000;
+}
+
+function buildCost(
+  model: FriendliModel,
+  existing: ExistingModel["cost"] | undefined,
+): NonNullable<ExistingModel["cost"]> | undefined {
+  // TOKEN-priced per-token USD rates are converted to USD/MTok. Any other
+  // unit (e.g. SECOND) is not a token rate: do not author a cost section for
+  // it instead of publishing an invented per-million price.
+  if (model.pricing.unit_type !== undefined && model.pricing.unit_type !== "TOKEN") {
+    return existing;
+  }
+  const input = perMillion(model.pricing.input);
+  const output = perMillion(model.pricing.output);
+  if (input === undefined || output === undefined) return existing;
+  return {
+    input,
+    output,
+    cache_read: perMillion(model.pricing.input_cache_read) ?? existing?.cache_read,
+    cache_write: perMillion(model.pricing.input_cache_write) ?? existing?.cache_write,
+  };
+}
+
+// Translate API reasoning_options into host-accurate catalog options.
+// budget_tokens is kept as an unbounded `{ type = "budget_tokens" }`:
+// reasoning_budget is a real, independently enforced Friendli control
+// (live-verified on GLM-5.3, gemma-4-31B-it, DeepSeek-V3.2, and
+// MiniMax-M2.5 — small budgets truncate reasoning_content mid-sentence while
+// completion continues), and peers such as OpenRouter/Requesty publish it
+// when the host supports it. The catalog's min/max values are not safe
+// published range constraints — GLM-5.3 accepted reasoning_budget=1_048_577
+// despite reporting max=1_048_576 — so the capability is preserved without
+// authoring bounds. A budget-only reasoner (MiniMax-M2.5) therefore publishes
+// `[{ type = "budget_tokens" }]`, not []: [] would falsely claim no caller
+// control on a host that documents reasoning_budget.
+function translateReasoningOptions(
+  api: FriendliModel["reasoning_options"],
+): SyncedFullModel["reasoning_options"] {
+  if (api === undefined) return undefined;
+  const options: NonNullable<SyncedFullModel["reasoning_options"]> = [];
+  for (const option of api) {
+    if (option === undefined) continue;
+    if (option.type === "budget_tokens") {
+      options.push({ type: "budget_tokens" });
+      continue;
+    }
+    options.push(option as NonNullable<SyncedFullModel["reasoning_options"]>[number]);
+  }
+  return options.length > 0 ? options : [];
+}
+
+function translateInterleaved(
+  modelID: string,
+  value: FriendliModel["interleaved"],
+  existing: SyncedFullModel["interleaved"] | undefined,
+): SyncedFullModel["interleaved"] {
+  const verified = VERIFIED_INTERLEAVED_OVERRIDES[modelID];
+  if (verified !== undefined) return verified;
+  if (value === undefined) return existing;
+  // The models endpoint can be stale/wrong for this field (verified live
+  // against deepseek-ai/DeepSeek-V3.2, see VERIFIED_INTERLEAVED_OVERRIDES) —
+  // trust an existing authored value over an API false rather than clearing it.
+  if (value === false) return existing;
+  if (value === true) return true;
+  return { field: value };
+}
+
+function inferFamily(modelID: string, name: string): SyncedFullModel["family"] {
+  const kimiFamily = inferKimiFamily(modelID, name);
+  if (kimiFamily !== undefined) return kimiFamily;
+  const target = `${modelID} ${name}`.toLowerCase();
+  return [...ModelFamilyValues]
+    .sort((a, b) => b.length - a.length)
+    .find((family) => {
+      const escaped = family.toLowerCase().replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+      if (family === "o") {
+        return new RegExp(`(^|[^a-z0-9])${escaped}(?=\\d|$|[^a-z0-9])`).test(target);
+      }
+      return new RegExp(`(^|[^a-z0-9])${escaped}(?=$|[^a-z0-9])`).test(target);
+    });
+}
+
+function buildFriendliModel(
+  model: FriendliModel,
+  existing: ExistingModel | undefined,
+  factorBase: string | undefined,
+): SyncedModel {
+  // translateModel already skips models past their deprecation_date, so every
+  // model reaching this point is live. Carry hand-authored lifecycle statuses
+  // (e.g. beta) through unchanged.
+  const status = existing?.status;
+
+  // Only override modalities when the API explicitly provides them; otherwise
+  // omit the override so lab metadata (e.g. gemma vision) is inherited.
+  const apiInput = model.input_modalities !== undefined ? translateModalities(model.input_modalities) : undefined;
+  const apiOutput = model.output_modalities !== undefined ? translateModalities(model.output_modalities) : undefined;
+  // For a factored provider entry, only write the modality sides Friendli
+  // actually supplied. Plain-object inheritance deep-merges, so an omitted
+  // side must remain omitted to preserve the lab's canonical modality list
+  // rather than replacing it with an empty array.
+  const modalities = apiInput !== undefined || apiOutput !== undefined
+    ? {
+      ...(apiInput !== undefined ? { input: apiInput } : {}),
+      ...(apiOutput !== undefined ? { output: apiOutput } : {}),
+    }
+    : undefined;
+  // undefined when the API omits input_modalities so factorBaseModel
+  // inherits the lab attachment; only override when explicitly provided.
+  const attachment = apiInput !== undefined ? apiInput.some((value) => value !== "text") : undefined;
+  // Completion-length cap: when a base_model exists, defer to the lab's own
+  // limit.output instead of forcing Friendli's max_completion_tokens onto it.
+  // Friendli's max_completion_tokens equals context_length for every one of
+  // the 7 live models, and blindly asserting that as the completion cap would
+  // overwrite lab-verified, genuinely tighter completion limits (e.g.
+  // DeepSeek-V3.2's lab file documents output=64_000 out of a 128_000
+  // context, not "same as context"). Only fall back to Friendli's own
+  // reported value when there is no base_model to inherit a real
+  // completion-cap policy from (full-inline entries).
+  const limit = {
+    context: model.context_length,
+    input: existing?.limit?.input,
+    output: factorBase !== undefined ? undefined : model.max_completion_tokens,
+  };
+  // Reasoning is tri-state: Friendli omits the flag for some reasoners, and
+  // treating "absent" as `false` would publish an explicit reasoning=false
+  // override on factored entries and strip their reasoning_options. Only
+  // override when the API is authoritative; otherwise the lab wins.
+  const reasoning = model.reasoning === true ? true : model.reasoning === false ? false : undefined;
+  const reasoningOptions = reasoning === false ? undefined : translateReasoningOptions(model.reasoning_options);
+  const interleaved = translateInterleaved(model.id, model.interleaved, existing?.interleaved);
+  const structuredOutput = model.functionality.structured_output;
+  const cost = buildCost(model, existing?.cost);
+  const releaseDate = existing?.release_date ?? new Date(model.created * 1000).toISOString().slice(0, 10);
+  const today = new Date().toISOString().slice(0, 10);
+  const lastUpdated = existing?.last_updated ?? today;
+
+  if (factorBase !== undefined) {
+    return factorBaseModel(
+      factorBase,
+      {
+        attachment,
+        reasoning,
+        reasoning_options: reasoningOptions,
+        interleaved,
+        // Friendli is authoritative for this host's tool-call surface; a
+        // real delta vs the lab (either direction) must be published.
+        tool_call: model.functionality.tool_call,
+        structured_output: structuredOutput,
+        // A factored entry inherits the lab description. Friendli's catalog
+        // description is host metadata, not a new model identity, and its
+        // generic text can be weaker than the lab's canonical description.
+        // Keep it only for full-inline entries below.
+        description: undefined,
+        limit,
+        modalities,
+        cost,
+        status,
+      },
+      limit,
+      existing?.base_model === factorBase ? existing.base_model_omit : undefined,
+    );
+  }
+
+  const name = existing?.name ?? (model.name.split("/").at(-1) ?? model.name);
+  // Full-inline has no lab to inherit from: a reasoning flag the API omits
+  // defaults to false here (describeModel needs a boolean), while factored
+  // entries above leave it unset so the lab's value stands.
+  const inlineReasoning = reasoning ?? false;
+  return {
+    name,
+    description:
+      existing?.description ??
+      model.description ??
+      describeModel({
+        id: model.id,
+        providerId: "friendli",
+        name,
+        family: existing?.family,
+        reasoning: inlineReasoning,
+        tool_call: model.functionality.tool_call,
+        structured_output: structuredOutput,
+        open_weights: Boolean(model.hugging_face_url),
+        // Full-inline entries have no lab modalities to inherit. Default only
+        // sides omitted by the API to text rather than constructing empty
+        // arrays, which would advertise an impossible no-output/no-input model.
+        modalities: { input: apiInput ?? ["text"], output: apiOutput ?? ["text"] },
+      }),
+    family: existing?.family ?? inferFamily(model.id, name),
+    // Full-inline has no lab to inherit from; default text-only when the API
+    // omits modalities. Earlier `attachment` is undefined in that case.
+    attachment: attachment ?? false,
+    reasoning: inlineReasoning,
+    reasoning_options: reasoningOptions,
+    tool_call: model.functionality.tool_call,
+    structured_output: structuredOutput,
+    temperature: existing?.temperature ?? true,
+    release_date: releaseDate,
+    last_updated: lastUpdated,
+    open_weights: Boolean(model.hugging_face_url),
+    interleaved,
+    knowledge: existing?.knowledge,
+    cost,
+    limit: { context: model.context_length, output: model.max_completion_tokens },
+    modalities: { input: apiInput ?? ["text"], output: apiOutput ?? ["text"] },
+    status,
+  };
+}
diff --git a/packages/core/test/friendli.test.ts b/packages/core/test/friendli.test.ts
new file mode 100644
index 00000000000..7c6afa49a91
--- /dev/null
+++ b/packages/core/test/friendli.test.ts
@@ -0,0 +1,41 @@
+import { expect, test } from "bun:test";
+
+import { friendli, FriendliModel } from "../src/sync/providers/friendli.js";
+
+const model = FriendliModel.parse({
+  id: "example/model",
+  name: "Example Model",
+  created: 1_775_088_000,
+  context_length: 128_000,
+  max_completion_tokens: 128_000,
+  functionality: {
+    tool_call: true,
+    structured_output: true,
+  },
+  pricing: {
+    input: "0.000001",
+    output: "0.000002",
+  },
+});
+
+test("tracks active Friendli models missing lab metadata", () => {
+  expect(friendli.missingModelID(model)).toBe(model.id);
+});
+
+test("rejects an empty Friendli catalog", () => {
+  expect(() => friendli.parseModels({ data: [] })).toThrow("empty model catalog");
+});
+
+test("skips a factored model when its lab metadata cannot be resolved", () => {
+  expect(friendli.translateModel(model, {
+    existing: () => ({ base_model: "example/missing" }),
+    authored: () => ({ base_model: "example/missing" }),
+  })).toBeUndefined();
+});
+
+test("does not track deprecated Friendli models as missing", () => {
+  expect(friendli.missingModelID({
+    ...model,
+    deprecation_date: "2000-01-01T00:00:00Z",
+  })).toBeUndefined();
+});
diff --git a/packages/core/test/missing-skips.test.ts b/packages/core/test/missing-skips.test.ts
new file mode 100644
index 00000000000..ddb3351186f
--- /dev/null
+++ b/packages/core/test/missing-skips.test.ts
@@ -0,0 +1,50 @@
+import { expect, spyOn, test } from "bun:test";
+import { mkdir, mkdtemp, rm } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import path from "node:path";
+
+import { syncProvider, type SyncProvider } from "../src/sync/index.js";
+import * as missingIssues from "../src/sync/missing-issues.js";
+
+test("opens issues for selectively skipped missing models", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "sync-missing-model-"));
+  const modelsDir = path.join(dir, "providers", "example", "models");
+  await mkdir(modelsDir, { recursive: true });
+  const existingPath = path.join(modelsDir, "needs-metadata.toml");
+  await Bun.write(existingPath, 'name = "Keep me"\n');
+  const issues = spyOn(missingIssues, "openMissingModelIssues").mockResolvedValue([]);
+  const provider: SyncProvider<{ id: string; missing: boolean }> = {
+    id: "example",
+    name: "Example",
+    modelsDir,
+    async fetchModels() {
+      return [
+        { id: "needs-metadata", missing: true },
+        { id: "intentional-skip", missing: false },
+      ];
+    },
+    parseModels(raw) {
+      return raw as { id: string; missing: boolean }[];
+    },
+    translateModel() {
+      return undefined;
+    },
+    sourceID(model) {
+      return model.id;
+    },
+    missingModelID(model) {
+      return model.missing ? model.id : undefined;
+    },
+  };
+
+  try {
+    const result = await syncProvider(provider, { openIssues: true });
+    expect(result).toMatchObject({ deleted: 0, unchanged: 1 });
+    expect(await Bun.file(existingPath).text()).toBe('name = "Keep me"\n');
+    expect(issues).toHaveBeenCalledTimes(1);
+    expect(issues.mock.calls[0]?.[1]).toEqual(["needs-metadata"]);
+  } finally {
+    issues.mockRestore();
+    await rm(dir, { recursive: true, force: true });
+  }
+});
diff --git a/providers/friendli/models/MiniMaxAI/MiniMax-M2.5.toml b/providers/friendli/models/MiniMaxAI/MiniMax-M2.5.toml
index 9fa904e270b..82bcc78bb46 100644
--- a/providers/friendli/models/MiniMaxAI/MiniMax-M2.5.toml
+++ b/providers/friendli/models/MiniMaxAI/MiniMax-M2.5.toml
@@ -1,13 +1,14 @@
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
 base_model = "minimax/MiniMax-M2.5"
-name = "MiniMax-M2.5"
-release_date = "2026-02-12"
-last_updated = "2026-02-12"
 structured_output = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.3
 output = 1.2
@@ -15,4 +16,3 @@ cache_read = 0.06
 
 [limit]
 context = 196_608
-output = 196_608
diff --git a/providers/friendli/models/deepseek-ai/DeepSeek-V3.2.toml b/providers/friendli/models/deepseek-ai/DeepSeek-V3.2.toml
index 4714efddca6..4dfa7122d90 100644
--- a/providers/friendli/models/deepseek-ai/DeepSeek-V3.2.toml
+++ b/providers/friendli/models/deepseek-ai/DeepSeek-V3.2.toml
@@ -1,19 +1,18 @@
-name = "DeepSeek-V3.2"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-family = "deepseek"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-tool_call = true
-structured_output = true
-temperature = true
-release_date = "2025-12-01"
-last_updated = "2025-12-01"
-open_weights = true
+# Toggle: chat_template_kwargs.enable_thinking = true | false
+# https://friendli.ai/docs/guides/reasoning
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
+base_model = "deepseek/deepseek-v3.2"
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 0.5
 output = 1.5
@@ -21,8 +20,3 @@ cache_read = 0.25
 
 [limit]
 context = 163_840
-output = 163_840
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/friendli/models/google/gemma-4-31B-it.toml b/providers/friendli/models/google/gemma-4-31B-it.toml
index cd945119017..e6094afcc7e 100644
--- a/providers/friendli/models/google/gemma-4-31B-it.toml
+++ b/providers/friendli/models/google/gemma-4-31B-it.toml
@@ -1,5 +1,17 @@
+# Toggle: chat_template_kwargs.enable_thinking = true | false
+# https://friendli.ai/docs/guides/reasoning
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
 base_model = "google/gemma-4-31b-it"
-reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 0.14
diff --git a/providers/friendli/models/zai-org/GLM-5.1.toml b/providers/friendli/models/zai-org/GLM-5.1.toml
index 1ee4131636b..d97a8d26dda 100644
--- a/providers/friendli/models/zai-org/GLM-5.1.toml
+++ b/providers/friendli/models/zai-org/GLM-5.1.toml
@@ -1,13 +1,18 @@
+# Toggle: chat_template_kwargs.enable_thinking = true | false
+# https://friendli.ai/docs/guides/reasoning
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
 base_model = "zhipuai/glm-5.1"
-name = "GLM-5.1"
-release_date = "2026-04-07"
-last_updated = "2026-04-07"
-structured_output = true
-reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
 [cost]
 input = 1.4
 output = 4.4
@@ -15,4 +20,3 @@ cache_read = 0.26
 
 [limit]
 context = 202_752
-output = 202_752
diff --git a/providers/friendli/models/zai-org/GLM-5.2.toml b/providers/friendli/models/zai-org/GLM-5.2.toml
index 81fbbb90c49..187d5de39d0 100644
--- a/providers/friendli/models/zai-org/GLM-5.2.toml
+++ b/providers/friendli/models/zai-org/GLM-5.2.toml
@@ -1,17 +1,28 @@
-name = "GLM-5.2"
-# Friendli documents only $.chat_template_kwargs.enable_thinking = true | false
-# for this model, not the configured effort values "high" and "max".
-# https://friendli.ai/docs/guides/reasoning (accessed 2026-06-25)
+# Toggle: chat_template_kwargs.enable_thinking = true | false
+# https://friendli.ai/docs/guides/reasoning
+# Effort: reasoning_effort = "high" | "max"
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
 base_model = "zhipuai/glm-5.2"
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
 [[reasoning_options]]
 type = "effort"
 values = ["high", "max"]
 
-[interleaved]
-field = "reasoning_content"
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
 input = 1.4
 output = 4.4
-cache_read = 0.26
\ No newline at end of file
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
diff --git a/providers/friendli/models/zai-org/GLM-5.3-Flash.toml b/providers/friendli/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..7d25fca9c7b
--- /dev/null
+++ b/providers/friendli/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,26 @@
+# Effort: reasoning_effort = "low" | "high" | "max"
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
+# https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
+base_model = "zhipuai/glm-5.3-flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/friendli/models/zai-org/GLM-5.3.toml b/providers/friendli/models/zai-org/GLM-5.3.toml
index bf8c8ce9c16..baad00a0786 100644
--- a/providers/friendli/models/zai-org/GLM-5.3.toml
+++ b/providers/friendli/models/zai-org/GLM-5.3.toml
@@ -1,42 +1,23 @@
-# Friendli's chat-completions endpoint documents reasoning_effort (generic
-# enum) and reasoning_budget (integer token cap) as real request fields; GLM-5.3
-# on Z.AI's own API always reasons and only exposes low|high|max, so we mirror
-# those graded levels here instead of Friendli's full generic effort enum.
-#
-# budget_tokens is a REAL, independently-enforced reasoning-token budget, not
-# derived from max_tokens/context capacity: two live POST /chat/completions
-# requests against this exact model (reasoning_budget=30 and =50, generous
-# max_tokens=3000/4000) cut reasoning_content mid-sentence at the requested
-# cap while completion_tokens continued to 1553/177 respectively — proof the
-# reasoning phase and the completion phase are capped independently.
-# min=-1/max=1_048_576 are Friendli's own reported values, read verbatim from
-# GET /serverless/v1/models -> reasoning_options -> {type: budget_tokens} for
-# zai-org/GLM-5.3 (not computed by us from max_completion_tokens; that field
-# happens to equal this model's max_completion_tokens/context_length, but we
-# pass through the catalog's own budget_tokens object, we do not derive it).
+# Effort: reasoning_effort = "low" | "high" | "max"
 # https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-effort-one-of-0
+# Budget: reasoning_budget = positive integer reasoning-token cap (-1 = unlimited)
 # https://friendli.ai/docs/openapi/model-apis/chat-completions#body-reasoning-budget-one-of-0
-# (accessed 2026-08-29)
 base_model = "zhipuai/glm-5.3"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
 [[reasoning_options]]
 type = "budget_tokens"
-min = -1
-max = 1_048_576
-
-[interleaved]
-field = "reasoning_content"
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.26
+output = 3.96
+cache_read = 0.234
 
 [limit]
 context = 1_048_576
-output = 1_048_576
-
diff --git a/sync.md b/sync.md
index 06ceb1b2a42..70a06dab55b 100644
--- a/sync.md
+++ b/sync.md
@@ -46,6 +46,7 @@ Sync runs also write `.sync/model-sync-report.md` for the automation workflow PR
 - Removes existing files that are no longer present in the desired synced set.
 - Writes `.sync/model-sync-report.md` for GitHub Actions.
 - When `skipCreates` is set and issue opens are enabled, opens one deduped GitHub issue per remote model missing from the local catalog (via `gh`).
+- When a provider selectively skips only some models, `missingModelID` can preserve existing metadata and mark those skips for the same deduped issue flow without disabling safe automatic creates.
 
 Because the runner removes files missing from the desired set, a provider module should only skip source models when deleting existing local files for those skipped IDs is intentional.
 
@@ -59,6 +60,8 @@ Providers that cannot safely auto-create TOMLs set `skipCreates: true`. In GitHu
 4. Dispatches the Issue Fixer explicitly so issues created with `GITHUB_TOKEN` can still produce PRs
 5. If listing fails, creates nothing (fail closed)
 
+Providers that can auto-create most models may instead return an ID from `missingModelID` only for `translateModel` skips that need manual metadata. The runner preserves an existing local entry for that ID while the issue is handled. Intentional skips return `undefined` and do not open issues.
+
 Requires `GH_TOKEN` on the sync workflow step. Local runs are notice-only unless `--open-issues`. Use `--no-issues` / `--dry-run` to skip creates. Each newly opened issue explicitly dispatches the issue-fixer workflow so an agent can research the missing metadata and open a model PR.
 
 The first Actions run may open a batch of issues per provider, including remote IDs the catalog intentionally omits (e.g. OpenAI whisper/tts/moderation surfaces, dated snapshots). This one-time volume is accepted by design: close unwanted issues once and the closed-title dedupe suppresses them permanently. If the dedupe list window (1000 labeled issues per provider) ever fills, the sync fails closed and creates nothing rather than risk duplicates.

From 88b37f329a371d00faef8fa5d8143f6465a82c1a Mon Sep 17 00:00:00 2001
From: Joost <15650945+jope35@users.noreply.github.com>
Date: Mon, 14 Sep 2026 19:41:42 +0200
Subject: [PATCH 1294/2012] feat(melious): add Melious provider with the GLM,
 DeepSeek and Kimi families (#7046)

* feat(melious): add Melious provider with the GLM, DeepSeek and Kimi families

Melious AI GmbH is a European OpenAI-compatible relay of open-weight models.
This is the first of two PRs: it adds the provider and the 15 models whose lab
metadata already exists under models/, so it touches providers/melious/ only --
no models/, labs/ or render.tsx changes. The remaining catalog follows once this
lands.

Every entry uses base_model and is override-only.

Data: GET https://api.melious.ai/v1/models?include_meta=true, cross-checked against
the per-model hub pages. Prices are the default `balanced` route (a bare chat model
ID resolves to `balanced`), converted at 1.1592 USD/EUR (ECB euro reference rate,
2026-09-11). Each file records its EUR figures, the rate and the date so the
arithmetic is checkable from the diff.

Reasoning: reasoning_effort is the only control this host exposes -- no lab-native
thinking field, no reasoning budget, so no toggle or budget_tokens. The host accepts
any effort string without a 4xx, so the authored values are the documented ladder
(low|medium|high) plus none, which was measured to be a real off state: it returns
usage.reasoning_tokens = 0 and no reasoning_content on glm-5.3, deepseek-v4-pro and
kimi-k2.7-code. Because none is reachable this is an effort list containing none,
not a toggle. Reasoning text is returned in message.reasoning_content.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(melious): derive reasoning_options per model from measurement

Replaces the uniform effort ["none","low","medium","high"] on all 15 entries,
which asserted a graded ladder this host does not implement.

Every model was probed against api.melious.ai/v1/chat/completions on 2026-09-14
at temperature 0 with the same prompt, comparing emitted reasoning_content across
none/low/medium/high (repeat samples where the result was load-bearing):

- 12 models: reasoning_effort="none" is a real off state (reasoning_tokens = 0,
  no reasoning_content) while low/medium/high all reason with no ordered
  difference in depth -> effort ["none", "high"]. An effort list containing none
  rather than a toggle, since off is reachable through this field.
- glm-5, deepseek-r1-0528, kimi-k2.5: "none" does not suppress reasoning and no
  level grades -> reasoning_options = []. An affirmative measurement of no caller
  control, not an untested default.

No model showed graded effort, so no graded levels are published. The host accepts
any effort string without a 4xx, so acceptance was not treated as evidence. The
catalog's reasoning_type is not a reliable guide either: glm-5.3-flash is typed
always-on yet honours "none", glm-5 is typed hybrid yet ignores it.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 providers/melious/logo.svg                    |  3 ++
 .../melious/models/deepseek-r1-0528.toml      | 31 +++++++++++++++
 providers/melious/models/deepseek-v3.2.toml   | 33 ++++++++++++++++
 .../models/deepseek-v4-flash-0731.toml        | 28 ++++++++++++++
 .../melious/models/deepseek-v4-pro-0813.toml  | 28 ++++++++++++++
 providers/melious/models/deepseek-v4-pro.toml | 28 ++++++++++++++
 .../melious/models/deepseek-v4.1-flash.toml   | 28 ++++++++++++++
 providers/melious/models/glm-5.1.toml         | 33 ++++++++++++++++
 providers/melious/models/glm-5.2.toml         | 28 ++++++++++++++
 providers/melious/models/glm-5.3-flash.toml   | 33 ++++++++++++++++
 providers/melious/models/glm-5.3.toml         | 28 ++++++++++++++
 providers/melious/models/glm-5.toml           | 31 +++++++++++++++
 providers/melious/models/kimi-k2.5.toml       | 38 +++++++++++++++++++
 providers/melious/models/kimi-k2.6.toml       | 35 +++++++++++++++++
 providers/melious/models/kimi-k2.7-code.toml  | 28 ++++++++++++++
 providers/melious/models/kimi-k3.toml         | 33 ++++++++++++++++
 providers/melious/provider.toml               | 33 ++++++++++++++++
 17 files changed, 499 insertions(+)
 create mode 100644 providers/melious/logo.svg
 create mode 100644 providers/melious/models/deepseek-r1-0528.toml
 create mode 100644 providers/melious/models/deepseek-v3.2.toml
 create mode 100644 providers/melious/models/deepseek-v4-flash-0731.toml
 create mode 100644 providers/melious/models/deepseek-v4-pro-0813.toml
 create mode 100644 providers/melious/models/deepseek-v4-pro.toml
 create mode 100644 providers/melious/models/deepseek-v4.1-flash.toml
 create mode 100644 providers/melious/models/glm-5.1.toml
 create mode 100644 providers/melious/models/glm-5.2.toml
 create mode 100644 providers/melious/models/glm-5.3-flash.toml
 create mode 100644 providers/melious/models/glm-5.3.toml
 create mode 100644 providers/melious/models/glm-5.toml
 create mode 100644 providers/melious/models/kimi-k2.5.toml
 create mode 100644 providers/melious/models/kimi-k2.6.toml
 create mode 100644 providers/melious/models/kimi-k2.7-code.toml
 create mode 100644 providers/melious/models/kimi-k3.toml
 create mode 100644 providers/melious/provider.toml

diff --git a/providers/melious/logo.svg b/providers/melious/logo.svg
new file mode 100644
index 00000000000..a2ff326807a
--- /dev/null
+++ b/providers/melious/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 800" fill="currentColor">
+  <path d="M378.57,345.92c0,18.04-14.62,32.65-32.65,32.65H0V32.65C0,14.62,14.62,0,32.65,0h345.92v345.92ZM800,421.43h-345.92c-18.04,0-32.65,14.62-32.65,32.65v345.92h345.92c18.04,0,32.65-14.62,32.65-32.65v-345.92ZM378.57,799.12c0-208.59-169.1-377.69-377.69-377.69H0v345.92C0,785.38,14.62,800,32.65,800h345.92v-.88ZM800,32.65c0-18.04-14.62-32.65-32.65-32.65h-345.92v.88c0,208.59,169.1,377.69,377.69,377.69h.88V32.65Z"/>
+</svg>
diff --git a/providers/melious/models/deepseek-r1-0528.toml b/providers/melious/models/deepseek-r1-0528.toml
new file mode 100644
index 00000000000..78d2c96463c
--- /dev/null
+++ b/providers/melious/models/deepseek-r1-0528.toml
@@ -0,0 +1,31 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.60 in / EUR 2.40
+# out / EUR 0.16 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-r1-0528 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget. Measured against this
+# deployment on 2026-09-14, it has no effect on this model: reasoning_effort="none"
+# still returns reasoning (reasoning_content 2_786/3_010 chars across repeat samples),
+# and low/medium/high produce no ordered difference. reasoning_options is therefore []
+# -- this model reasons, with no caller control on this host. That is an affirmative
+# measurement, not an untested default.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 164_000 for this deployment, not the lab's
+# 128_000.
+base_model = "deepseek/deepseek-r1"
+reasoning_options = []
+
+[cost]
+input = 0.69552
+output = 2.78208
+cache_read = 0.185472
+
+[limit]
+context = 164_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/deepseek-v3.2.toml b/providers/melious/models/deepseek-v3.2.toml
new file mode 100644
index 00000000000..87e4436bc09
--- /dev/null
+++ b/providers/melious/models/deepseek-v3.2.toml
@@ -0,0 +1,33 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.30 in / EUR 0.50
+# out / EUR 0.08 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-v3.2 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 164_000 for this deployment, not the lab's
+# 128_000.
+base_model = "deepseek/deepseek-v3.2"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.34776
+output = 0.5796
+cache_read = 0.092736
+
+[limit]
+context = 164_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/deepseek-v4-flash-0731.toml b/providers/melious/models/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..6f5c2d8fa23
--- /dev/null
+++ b/providers/melious/models/deepseek-v4-flash-0731.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.10 in / EUR 0.25
+# out / EUR 0.02 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-v4-flash-0731 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.11592
+output = 0.2898
+cache_read = 0.023184
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/deepseek-v4-pro-0813.toml b/providers/melious/models/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..0cdb0c82645
--- /dev/null
+++ b/providers/melious/models/deepseek-v4-pro-0813.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 1.00 in / EUR 3.00
+# out / EUR 0.10 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-v4-pro-0813 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 1.1592
+output = 3.4776
+cache_read = 0.11592
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/deepseek-v4-pro.toml b/providers/melious/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..ca2515e3ebf
--- /dev/null
+++ b/providers/melious/models/deepseek-v4-pro.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 1.60 in / EUR 3.20
+# out / EUR 0.40 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-v4-pro (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "deepseek/deepseek-v4-pro"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 1.85472
+output = 3.70944
+cache_read = 0.46368
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/deepseek-v4.1-flash.toml b/providers/melious/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..679d9879d8b
--- /dev/null
+++ b/providers/melious/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.20 in / EUR 1.00
+# out / EUR 0.01 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/deepseek-v4.1-flash (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.23184
+output = 1.1592
+cache_read = 0.011592
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/glm-5.1.toml b/providers/melious/models/glm-5.1.toml
new file mode 100644
index 00000000000..63766b09ff8
--- /dev/null
+++ b/providers/melious/models/glm-5.1.toml
@@ -0,0 +1,33 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 1.30 in / EUR 4.00
+# out / EUR 0.32 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/glm-5.1 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 203_000 for this deployment, not the lab's
+# 200_000.
+base_model = "zhipuai/glm-5.1"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 1.50696
+output = 4.6368
+cache_read = 0.370944
+
+[limit]
+context = 203_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/glm-5.2.toml b/providers/melious/models/glm-5.2.toml
new file mode 100644
index 00000000000..8ca4dbe4c3e
--- /dev/null
+++ b/providers/melious/models/glm-5.2.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 1.00 in / EUR 4.00
+# out / EUR 0.25 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/glm-5.2 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 1.1592
+output = 4.6368
+cache_read = 0.2898
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/glm-5.3-flash.toml b/providers/melious/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..d25a740b3f3
--- /dev/null
+++ b/providers/melious/models/glm-5.3-flash.toml
@@ -0,0 +1,33 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.10 in / EUR 0.40
+# out / EUR 0.02 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/glm-5.3-flash (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# modalities.input: this host routes ["text", "image", "video"], narrower than the
+# lab's ["text", "image", "video", "pdf"].
+base_model = "zhipuai/glm-5.3-flash"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.11592
+output = 0.46368
+cache_read = 0.023184
+
+[modalities]
+input = ["text", "image", "video"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/glm-5.3.toml b/providers/melious/models/glm-5.3.toml
new file mode 100644
index 00000000000..03eac7197c0
--- /dev/null
+++ b/providers/melious/models/glm-5.3.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 1.00 in / EUR 3.00
+# out / EUR 0.20 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/glm-5.3 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 1.1592
+output = 3.4776
+cache_read = 0.23184
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/glm-5.toml b/providers/melious/models/glm-5.toml
new file mode 100644
index 00000000000..19ac9f0e728
--- /dev/null
+++ b/providers/melious/models/glm-5.toml
@@ -0,0 +1,31 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.95 in / EUR 2.90
+# out / EUR 0.23 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/glm-5 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget. Measured against this
+# deployment on 2026-09-14, it has no effect on this model: reasoning_effort="none"
+# still returns reasoning (reasoning_content 2_444/2_239/686/3_907 chars across repeat
+# samples), and low/medium/high produce no ordered difference. reasoning_options is
+# therefore [] -- this model reasons, with no caller control on this host. That is an
+# affirmative measurement, not an untested default.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 203_000 for this deployment, not the lab's
+# 204_800.
+base_model = "zhipuai/glm-5"
+reasoning_options = []
+
+[cost]
+input = 1.10124
+output = 3.36168
+cache_read = 0.266616
+
+[limit]
+context = 203_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/kimi-k2.5.toml b/providers/melious/models/kimi-k2.5.toml
new file mode 100644
index 00000000000..fabc024c1a0
--- /dev/null
+++ b/providers/melious/models/kimi-k2.5.toml
@@ -0,0 +1,38 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.50 in / EUR 2.55
+# out / EUR 0.12 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/kimi-k2.5 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget. Measured against this
+# deployment on 2026-09-14, it has no effect on this model: reasoning_effort="none"
+# still returns reasoning (reasoning_content 1_177/1_714/1_100 chars across repeat
+# samples), and low/medium/high produce no ordered difference. reasoning_options is
+# therefore [] -- this model reasons, with no caller control on this host. That is an
+# affirmative measurement, not an untested default.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 262_000 for this deployment, not the lab's
+# 262_144. limit.output: max_output_tokens is null for this model, so output is capped
+# at the served context rather than inheriting the lab's larger 262_144.
+# modalities.input: this host routes ["text", "image"], narrower than the lab's
+# ["text", "image", "video"].
+base_model = "moonshotai/kimi-k2.5"
+reasoning_options = []
+
+[cost]
+input = 0.5796
+output = 2.95596
+cache_read = 0.139104
+
+[limit]
+context = 262_000
+output = 262_000
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/kimi-k2.6.toml b/providers/melious/models/kimi-k2.6.toml
new file mode 100644
index 00000000000..4784a6bc4c8
--- /dev/null
+++ b/providers/melious/models/kimi-k2.6.toml
@@ -0,0 +1,35 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.70 in / EUR 3.50
+# out / EUR 0.23 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/kimi-k2.6 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 256_000 for this deployment, not the lab's
+# 262_144. limit.output: max_output_tokens is null for this model, so output is capped
+# at the served context rather than inheriting the lab's larger 262_144.
+base_model = "moonshotai/kimi-k2.6"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.81144
+output = 4.0572
+cache_read = 0.266616
+
+[limit]
+context = 256_000
+output = 256_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/kimi-k2.7-code.toml b/providers/melious/models/kimi-k2.7-code.toml
new file mode 100644
index 00000000000..a38cc7a360b
--- /dev/null
+++ b/providers/melious/models/kimi-k2.7-code.toml
@@ -0,0 +1,28 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 0.70 in / EUR 3.00
+# out / EUR 0.19 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/kimi-k2.7-code (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+base_model = "moonshotai/kimi-k2.7-code"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.81144
+output = 3.4776
+cache_read = 0.220248
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/models/kimi-k3.toml b/providers/melious/models/kimi-k3.toml
new file mode 100644
index 00000000000..6f9d4f8021b
--- /dev/null
+++ b/providers/melious/models/kimi-k3.toml
@@ -0,0 +1,33 @@
+# Cost: Melious list price for the default `balanced` route -- EUR 2.75 in / EUR 13.75
+# out / EUR 0.68 cache read per 1M tokens -- converted at 1.1592 USD/EUR (ECB euro
+# reference rate, 2026-09-11).
+# Sources: GET https://api.melious.ai/v1/models?include_meta=true and
+# https://melious.ai/hub/models/kimi-k3 (both accessed 2026-09-14)
+# FX: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/eurofxref-graph-usd.en.html
+# Effort: reasoning_effort is the only reasoning control this host exposes, and it
+# forwards no lab-native thinking field and no reasoning budget, so no toggle or
+# budget_tokens. Measured against this deployment on 2026-09-14:
+# reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0 and no
+# reasoning_content), while low/medium/high all return reasoning with no ordered
+# difference in depth, so they are not published as separate grades. The host also
+# accepts undocumented values without a 4xx, so acceptance is not evidence. Because
+# off is reachable through the effort field this is an effort list containing none,
+# not a toggle.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Reasoning text is returned in choices[].message.reasoning_content. cache_read:
+# usage.cached_tokens are billed at Melious's discounted prefix-cache rate.
+# limit.context: the catalog reports 1_000_000 for this deployment, not the lab's
+# 1_048_576.
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 3.1878
+output = 15.939
+cache_read = 0.788256
+
+[limit]
+context = 1_000_000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/melious/provider.toml b/providers/melious/provider.toml
new file mode 100644
index 00000000000..db50ccc00f3
--- /dev/null
+++ b/providers/melious/provider.toml
@@ -0,0 +1,33 @@
+# Melious AI GmbH relays open-weight models built by other labs on European
+# infrastructure. It is not the lab for any model it serves, so every model entry
+# here uses base_model. Raw HTTP is POST /v1/chat/completions; the OpenAI SDK works
+# by pointing base_url at the API.
+# Reasoning: the only control on this surface is `reasoning_effort`. No lab-native
+# field is forwarded -- no thinking object, no enable_thinking, no reasoning budget --
+# so no entry here uses `toggle` or `budget_tokens`. The host also accepts any effort
+# string without a 4xx, including undocumented ones, so acceptance is not evidence.
+# Every model here was measured against this deployment on 2026-09-14 instead. Two
+# behaviours were found, and neither matches the vendor docs' low|medium|high ladder:
+# on 12 models reasoning_effort="none" is a real off state (usage.reasoning_tokens = 0
+# and no reasoning_content) while low, medium and high all reason with no ordered
+# difference in depth -- those are authored as effort ["none", "high"], an effort list
+# containing none rather than a toggle, because off is reachable through this field.
+# On glm-5, deepseek-r1-0528 and kimi-k2.5, "none" does not suppress reasoning and no
+# level grades, so those carry reasoning_options = [] -- an affirmative measurement of
+# no caller control, not an untested default. The catalog's own reasoning_type is not
+# a reliable guide: glm-5.3-flash is typed always-on yet honours "none", and glm-5 is
+# typed hybrid yet ignores it. Reasoning text is returned in
+# choices[].message.reasoning_content.
+# https://melious.ai/docs/reference/chat-completions (accessed 2026-09-14)
+# Routing: a flavour is selected with a model-ID suffix (`:balanced`, `:speed`,
+# `:price`, `:eco`, `:batch`). A bare chat model ID resolves to `balanced`, which is
+# what the ids and prices here describe; only bare ids are published. A `preset`
+# field biases routing, not reasoning depth.
+# https://melious.ai/docs/concepts/routing (accessed 2026-09-14)
+# Caching: usage.cached_tokens are prompt tokens served from the prefix cache and
+# are billed at a cheaper cache-read rate, recorded as cost.cache_read.
+name = "Melious"
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.melious.ai/v1"
+env = ["MELIOUS_API_KEY"]
+doc = "https://melious.ai/docs/reference/models"

From b7cd7e2e557165424fb53c6b6439f5972c3ca10e Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 13:08:29 -0500
Subject: [PATCH 1295/2012] fix: [missing-model] tinfoil: deepseek-v4-1-flash
 (#6922)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../tinfoil/models/deepseek-v4-1-flash.toml   | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/tinfoil/models/deepseek-v4-1-flash.toml

diff --git a/providers/tinfoil/models/deepseek-v4-1-flash.toml b/providers/tinfoil/models/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..23963af6996
--- /dev/null
+++ b/providers/tinfoil/models/deepseek-v4-1-flash.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-11):
+# - https://inference.tinfoil.sh/v1/models (id deepseek-v4-1-flash; experimental; pricing; context 1048576; multimodal; reasoning)
+# - https://docs.tinfoil.sh/guides/reasoning.md (host effort surface; sibling deepseek-v4-flash: none|minimal|low|medium|high|xhigh|max)
+# - Lab metadata: models/deepseek/deepseek-v4.1-flash.toml
+# Effort: reasoning_effort on Tinfoil; off via none (API also exposes thinking disable). Same surface as deepseek-v4-flash.
+# status beta: remote catalog marks experimental=true.
+base_model = "deepseek/deepseek-v4.1-flash"
+status = "beta"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.75
+output = 3
+cache_read = 0.19
+
+[limit]
+context = 1_048_576

From d2a07674d68a8fefbf9aa50d699600a45ebd8962 Mon Sep 17 00:00:00 2001
From: Dawncoral <48824982+Dawncoral@users.noreply.github.com>
Date: Tue, 15 Sep 2026 02:10:03 +0800
Subject: [PATCH 1296/2012] Add Wallaby provider (wallaby/kimi-k3) (#7004)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Add Wallaby provider (wallaby/kimi-k3)

* Declare host-verified reasoning effort for wallaby/kimi-k3

reasoning_effort low/medium/high verified effective on this host
(2026-09-14): graded reduction in reasoning tokens. none/minimal are
accepted but do not disable reasoning, and omitting the parameter runs
full effort, so declare effort values without none and no toggle.
Also restore interleaved reasoning_content and align doc URL.

* Address review: unfixed logo size, add host-verified max effort tier

- logo.svg: drop fixed width/height, keep square viewBox + currentColor
  (merge blocker for new providers)
- reasoning_options: re-probed host 2026-09-14, same-prompt median
  reasoning tokens over 6 runs per level: low 61 < medium 85 < high 105
  < max 123. max is accepted and highest; medium is a distinct graded
  intermediate tier on this wire. No wire path disables reasoning
  (none/minimal still reason; reasoning.enabled=false and
  thinking.type=disabled ignored), so no toggle and no none value.

---------

Co-authored-by: 袁铭鸿 <yuanminghong@MacBook-Pro.local>
Co-authored-by: Dawncoral <dawncoral@users.noreply.github.com>
---
 providers/wallaby/logo.svg                     |  3 +++
 .../wallaby/models/moonshotai/kimi-k3.toml     | 18 ++++++++++++++++++
 providers/wallaby/provider.toml                |  9 +++++++++
 3 files changed, 30 insertions(+)
 create mode 100644 providers/wallaby/logo.svg
 create mode 100644 providers/wallaby/models/moonshotai/kimi-k3.toml
 create mode 100644 providers/wallaby/provider.toml

diff --git a/providers/wallaby/logo.svg b/providers/wallaby/logo.svg
new file mode 100644
index 00000000000..2eddcadc840
--- /dev/null
+++ b/providers/wallaby/logo.svg
@@ -0,0 +1,3 @@
+<svg viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path d="M3 5L6.5 19L10 9L12 15L14 9L17.5 19L21 5" stroke="currentColor" stroke-width="2.2" stroke-linecap="round" stroke-linejoin="round"/>
+</svg>
diff --git a/providers/wallaby/models/moonshotai/kimi-k3.toml b/providers/wallaby/models/moonshotai/kimi-k3.toml
new file mode 100644
index 00000000000..ed5e6e18fe3
--- /dev/null
+++ b/providers/wallaby/models/moonshotai/kimi-k3.toml
@@ -0,0 +1,18 @@
+# Effort: reasoning_effort = low|medium|high|max (verified on this host
+# 2026-09-14; same-prompt median reasoning tokens over 6 runs per level:
+# low 61 < medium 85 < high 105 < max 123 — four graded tiers, "max" accepted
+# and highest, "medium" a distinct intermediate tier). No wire path disables
+# reasoning: "none"/"minimal" are accepted but still reason, and
+# reasoning.enabled=false / thinking.type="disabled" are ignored — hence no
+# "none" in values and no toggle.
+# Pricing: https://wallabytoken.com/pricing.json (accessed 2026-09-14), USD per 1M tokens.
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.70
+output = 13.50
+cache_read = 0.27
diff --git a/providers/wallaby/provider.toml b/providers/wallaby/provider.toml
new file mode 100644
index 00000000000..ffa9498fe24
--- /dev/null
+++ b/providers/wallaby/provider.toml
@@ -0,0 +1,9 @@
+# OpenAI-compatible API focused on Kimi K3. Reasoning effort is adjustable via
+# `reasoning_effort` (low/medium/high) and reasoning streams in the
+# `reasoning_content` field. `temperature` is accepted but normalized to 1.
+# https://wallabytoken.com/docs (accessed 2026-09-14)
+name = "Wallaby"
+npm = "@ai-sdk/openai-compatible"
+env = ["WALLABY_API_KEY"]
+api = "https://api.wallabytoken.com/v1"
+doc = "https://wallabytoken.com/docs"

From c7befb0f3eb16a6ff7eaf032563627c21b1e40f2 Mon Sep 17 00:00:00 2001
From: "aki.io team" <toine@aki.io>
Date: Mon, 14 Sep 2026 20:11:35 +0200
Subject: [PATCH 1297/2012] AKI.IO glm5.3 added, kimit k2.7-coderemoved (#7008)

---
 providers/aki-io/models/glm5.3-754b.toml       | 13 +++++++++++++
 .../aki-io/models/kimi-k2.7-code-1100b.toml    | 18 ------------------
 2 files changed, 13 insertions(+), 18 deletions(-)
 create mode 100644 providers/aki-io/models/glm5.3-754b.toml
 delete mode 100644 providers/aki-io/models/kimi-k2.7-code-1100b.toml

diff --git a/providers/aki-io/models/glm5.3-754b.toml b/providers/aki-io/models/glm5.3-754b.toml
new file mode 100644
index 00000000000..fc20d83dccf
--- /dev/null
+++ b/providers/aki-io/models/glm5.3-754b.toml
@@ -0,0 +1,13 @@
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+# cost as described on the AKI.IO pricing page at https://aki.io/#pricing
+[cost]
+input = 1.00
+output = 3.50
+cache_read = 0.25
+
+[limit]
+context = 524_288
+output = 81_920
+
diff --git a/providers/aki-io/models/kimi-k2.7-code-1100b.toml b/providers/aki-io/models/kimi-k2.7-code-1100b.toml
deleted file mode 100644
index a814eda61af..00000000000
--- a/providers/aki-io/models/kimi-k2.7-code-1100b.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-# attachment are not supported on AKI.IO for security reasons and therefore are disabled
-# AKI.IO supports a reduced set of input modalities, thats why the description of the base model are overridden
-base_model = "moonshotai/kimi-k2.7-code"
-attachment = false
-reasoning_options = []
-
-# cost as described on the AKI.IO pricing page at https://aki.io/#pricing
-[cost]
-input = 0.86
-output = 3.00
-cache_read = 0.18
-
-[limit]
-context = 262_144
-output = 81_920
-
-[modalities]
-input = ["text", "image"]

From fbc1ba4d8f79d73605b8d1b6eceb554b49c225e3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:30:23 +0000
Subject: [PATCH 1298/2012] chore(sync): update OpenRouter model catalog
 (#7079)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576

From a0a53fba078e63408f52aed2528f25c879de2057 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:30:31 +0000
Subject: [PATCH 1299/2012] chore(sync): update LLM Gateway model catalog
 (#7080)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/atria/atria-dawn-preview.toml      | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/atria/atria-dawn-preview.toml

diff --git a/providers/llmgateway-providers/models/atria/atria-dawn-preview.toml b/providers/llmgateway-providers/models/atria/atria-dawn-preview.toml
new file mode 100644
index 00000000000..ff54b19c572
--- /dev/null
+++ b/providers/llmgateway-providers/models/atria/atria-dawn-preview.toml
@@ -0,0 +1,26 @@
+name = "Atria Dawn Preview (Atria)"
+description = "Preview model for early access evaluation, prototyping, and compatibility testing"
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

From d31209b9ada0a26ee10484c0984324f1bcf47552 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:30:39 +0000
Subject: [PATCH 1300/2012] chore(sync): update Tinfoil model catalog (#7082)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/deepseek-v4-1-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/tinfoil/models/deepseek-v4-1-flash.toml b/providers/tinfoil/models/deepseek-v4-1-flash.toml
index 23963af6996..e3ef163f515 100644
--- a/providers/tinfoil/models/deepseek-v4-1-flash.toml
+++ b/providers/tinfoil/models/deepseek-v4-1-flash.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.75
-output = 3
-cache_read = 0.19
+input = 0.65
+output = 1.45
+cache_read = 0.13
 
 [limit]
 context = 1_048_576

From 237b345e527d496f3ebe951fb4d4b7865d7216ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:30:41 +0000
Subject: [PATCH 1301/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7083)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway/models/atria-dawn-preview.toml | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100644 providers/llmgateway/models/atria-dawn-preview.toml

diff --git a/providers/llmgateway/models/atria-dawn-preview.toml b/providers/llmgateway/models/atria-dawn-preview.toml
new file mode 100644
index 00000000000..29a41c97425
--- /dev/null
+++ b/providers/llmgateway/models/atria-dawn-preview.toml
@@ -0,0 +1,26 @@
+name = "Atria Dawn Preview"
+description = "Preview model for early access evaluation, prototyping, and compatibility testing"
+release_date = "2026-09-12"
+last_updated = "2026-09-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 378c14f6e888da31342e893a925ac7e8e80467e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:30:44 +0000
Subject: [PATCH 1302/2012] chore(sync): update Merge Gateway model catalog
 (#7081)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 +++----
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index f966f69d01c..42aa3ae2951 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 84592d1b8dd..0264c6bc733 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From f9d3fe1575c5d92bff09e513f1c739b9cee6ffe6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 19:23:20 +0000
Subject: [PATCH 1303/2012] chore(sync): update OpenRouter model catalog
 (#7084)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..70e672577c6 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.975
+output = 11.06
+cache_read = 0.2291
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index d6c9825dd85..018aa6c317f 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.91
-output = 2.86
-cache_read = 0.169
+input = 0.9
+output = 3.069
+cache_read = 0.18
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 385d8f5f5c4d6040581ac123113d7630acd071a5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 19:23:29 +0000
Subject: [PATCH 1304/2012] chore(sync): update Kilo model catalog (#7085)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 53a08317945..7c4f6bc8390 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.91
-output = 2.86
-cache_read = 0.169
+input = 0.9
+output = 3.069
+cache_read = 0.18
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 54d73a7a4137719ff8ccfcf8b43345d8953c2474 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 20:25:15 +0000
Subject: [PATCH 1305/2012] chore(sync): update OpenRouter model catalog
 (#7089)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576

From 9b784475ecb4c7b3956490de88c4469f27858d76 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 20:25:20 +0000
Subject: [PATCH 1306/2012] chore(sync): update Kilo model catalog (#7091)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..c7e3a421910 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.975
+output = 11.06
+cache_read = 0.2291
 
 [limit]
 context = 1_048_576

From b39c9f46e4c1ab337124d7b315f03a2737d64c50 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 20:25:23 +0000
Subject: [PATCH 1307/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7088)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-1-flash.toml            | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/empiriolabs/models/deepseek-v4-1-flash.toml

diff --git a/providers/empiriolabs/models/deepseek-v4-1-flash.toml b/providers/empiriolabs/models/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..ca0af8dd884
--- /dev/null
+++ b/providers/empiriolabs/models/deepseek-v4-1-flash.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 393_216
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.3
+
+[limit]
+output = 393_216

From 9b9fad2cef1103bbdaa130aa0cce5e3ab60b698b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 20:25:28 +0000
Subject: [PATCH 1308/2012] chore(sync): update Merge Gateway model catalog
 (#7090)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml            | 6 +++---
 .../merge-gateway/models/deepseek/deepseek-v4-pro.toml   | 9 ++++++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
index bf23b6fc065..063b14797cb 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-pro-0813.toml
@@ -8,6 +8,6 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.74
+output = 3.48
+cache_read = 0.003625
diff --git a/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml b/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
index 79cfd112ba1..c68cf8e2951 100644
--- a/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/merge-gateway/models/deepseek/deepseek-v4-pro.toml
@@ -9,6 +9,9 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.74
+output = 3.48
+cache_read = 0.145
+
+[limit]
+output = 262_000

From 5c21a5da3893d90abbb59c22bbf17de306828f30 Mon Sep 17 00:00:00 2001
From: Divy <divy@coralbricks.ai>
Date: Mon, 14 Sep 2026 14:00:34 -0700
Subject: [PATCH 1309/2012] fix(coralbricks): drop PDF input from GLM 5.3 Flash
 (#7086)

The zhipuai/glm-5.3-flash baseline lists PDF input, which the
CoralBricks endpoint does not accept for this model. Override
[modalities] to the text, image and video input it serves.

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 providers/coralbricks/models/glm-5.3-flash-fp4.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/coralbricks/models/glm-5.3-flash-fp4.toml b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
index 199058e48b0..a7836490ddd 100644
--- a/providers/coralbricks/models/glm-5.3-flash-fp4.toml
+++ b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
@@ -12,3 +12,7 @@ output = 131_072
 input = 0.15
 output = 0.5
 cache_read = 0
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

From c28ce719c50a6de8135eb51c8508a28853b45065 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 21:23:06 +0000
Subject: [PATCH 1310/2012] chore(sync): update OpenRouter model catalog
 (#7092)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576

From da88c6b195dfb4e3b43bf92d74d5dc6f904e9e24 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 22:25:10 +0000
Subject: [PATCH 1311/2012] chore(sync): update Kilo model catalog (#7096)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../google/gemini-2.5-pro-preview-05-06.toml  | 30 -------------------
 .../models/openai/gpt-4-turbo-preview.toml    | 23 --------------
 2 files changed, 53 deletions(-)
 delete mode 100644 providers/kilo/models/google/gemini-2.5-pro-preview-05-06.toml
 delete mode 100644 providers/kilo/models/openai/gpt-4-turbo-preview.toml

diff --git a/providers/kilo/models/google/gemini-2.5-pro-preview-05-06.toml b/providers/kilo/models/google/gemini-2.5-pro-preview-05-06.toml
deleted file mode 100644
index 5d6ee47e3c9..00000000000
--- a/providers/kilo/models/google/gemini-2.5-pro-preview-05-06.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "Google: Gemini 2.5 Pro Preview 05-06"
-description = "Advanced Gemini model for complex reasoning, coding, and multimodal analysis"
-family = "gemini"
-release_date = "2025-05-07"
-last_updated = "2025-05-07"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["high"]
-
-[cost]
-input = 1.25
-output = 10
-reasoning = 10
-cache_read = 0.125
-cache_write = 0.375
-
-[limit]
-context = 1_048_576
-output = 65_535
-
-[modalities]
-input = ["text", "image", "pdf", "audio", "video"]
-output = ["text"]
diff --git a/providers/kilo/models/openai/gpt-4-turbo-preview.toml b/providers/kilo/models/openai/gpt-4-turbo-preview.toml
deleted file mode 100644
index 25a840098c0..00000000000
--- a/providers/kilo/models/openai/gpt-4-turbo-preview.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "OpenAI: GPT-4 Turbo Preview ($$$$)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-25"
-last_updated = "2024-01-25"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 10
-output = 30
-
-[limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text", "pdf"]
-output = ["text"]

From fd30f18c3130e8f1619edb7a39b76fbdac77d188 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 22:25:34 +0000
Subject: [PATCH 1312/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7093)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/alibaba/qwen3.8-flash.toml      | 9 +++++++--
 providers/vercel/models/inclusionai/ling-3.0-flash.toml | 6 +++---
 2 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/providers/vercel/models/alibaba/qwen3.8-flash.toml b/providers/vercel/models/alibaba/qwen3.8-flash.toml
index 4f06d10abd2..bd0aba476f5 100644
--- a/providers/vercel/models/alibaba/qwen3.8-flash.toml
+++ b/providers/vercel/models/alibaba/qwen3.8-flash.toml
@@ -6,10 +6,15 @@
 # Budget: providerOptions.alibaba.thinkingBudget -> thinking_budget (reasoning tokens)
 base_model = "alibaba/qwen3.8-flash"
 name = "Qwen 3.8 Flash"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.16
+input = 0.15
 output = 0.47
 cache_read = 0.016
 cache_write = 0.2
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash.toml b/providers/vercel/models/inclusionai/ling-3.0-flash.toml
index 5f0c3f52fde..dd3bac573c4 100644
--- a/providers/vercel/models/inclusionai/ling-3.0-flash.toml
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash.toml
@@ -10,9 +10,9 @@ open_weights = false
 reasoning_options = []
 
 [cost]
-input = 0.06
-output = 0.18
-cache_read = 0.012
+input = 0.021
+output = 0.063
+cache_read = 0.0042
 
 [limit]
 context = 256_000

From c2131f68b19f1b36372a7646fcfcb4d4881fc83c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 22:25:41 +0000
Subject: [PATCH 1313/2012] chore(sync): update OpenRouter model catalog
 (#7095)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +--
 .../google/gemini-2.5-pro-preview-05-06.toml  | 38 -------------------
 .../models/openai/gpt-4-turbo-preview.toml    | 24 ------------
 3 files changed, 3 insertions(+), 65 deletions(-)
 delete mode 100644 providers/openrouter/models/google/gemini-2.5-pro-preview-05-06.toml
 delete mode 100644 providers/openrouter/models/openai/gpt-4-turbo-preview.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..f7dc249d032 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.0763
+output = 0.1526
+cache_read = 0.01526
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemini-2.5-pro-preview-05-06.toml b/providers/openrouter/models/google/gemini-2.5-pro-preview-05-06.toml
deleted file mode 100644
index 658c2b8b4b9..00000000000
--- a/providers/openrouter/models/google/gemini-2.5-pro-preview-05-06.toml
+++ /dev/null
@@ -1,38 +0,0 @@
-name = "Gemini 2.5 Pro Preview 05-06"
-description = "Advanced Gemini model for complex reasoning, coding, and multimodal analysis"
-family = "gemini-pro"
-release_date = "2025-05-07"
-last_updated = "2025-05-07"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2025-01-31"
-open_weights = false
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 128
-max = 32_768
-
-[cost]
-input = 1.25
-output = 10
-reasoning = 10
-cache_read = 0.125
-cache_write = 0.375
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 2.5
-output = 15
-cache_read = 0.25
-
-[limit]
-context = 1_048_576
-output = 65_535
-
-[modalities]
-input = ["text", "image", "pdf", "audio", "video"]
-output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-4-turbo-preview.toml b/providers/openrouter/models/openai/gpt-4-turbo-preview.toml
deleted file mode 100644
index f1e4784d5dd..00000000000
--- a/providers/openrouter/models/openai/gpt-4-turbo-preview.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "GPT-4 Turbo Preview"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-01-25"
-last_updated = "2024-01-25"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2023-12-31"
-open_weights = false
-
-[cost]
-input = 10
-output = 30
-
-[limit]
-context = 128_000
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 8aed31f1ab8c41df4e0ef4b196bc9d29983e9f99 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 23:22:17 +0000
Subject: [PATCH 1314/2012] chore(sync): update Kilo model catalog (#7098)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index c7e3a421910..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.975
-output = 11.06
-cache_read = 0.2291
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7c4f6bc8390..556169b19ef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -16,12 +16,12 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.9
-output = 3.069
-cache_read = 0.18
+output = 3
+cache_read = 0.09
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 00557bb267a0a311caf37af35520f971482cd79e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 23:22:27 +0000
Subject: [PATCH 1315/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7099)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/alibaba/qwen-3.6-max-preview.toml   | 11 +++++++++--
 providers/vercel/models/alibaba/qwen3.5-flash.toml    |  2 +-
 providers/vercel/models/alibaba/qwen3.5-plus.toml     | 11 +++++++++--
 providers/vercel/models/alibaba/qwen3.6-plus.toml     |  2 +-
 4 files changed, 20 insertions(+), 6 deletions(-)

diff --git a/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml b/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
index 957f43d1d02..2978e3e117c 100644
--- a/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
+++ b/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
@@ -5,15 +5,22 @@ release_date = "2026-04-20"
 last_updated = "2026-04-24"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1, max = 131_072 }]
 temperature = true
 tool_call = true
 open_weights = true
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 131_072
+
 [cost]
 input = 1.3
 output = 7.8
-cache_read = 0.26
+cache_read = 0.13
 cache_write = 1.625
 
 [limit]
diff --git a/providers/vercel/models/alibaba/qwen3.5-flash.toml b/providers/vercel/models/alibaba/qwen3.5-flash.toml
index f7d9fba9ad9..ad8d97ce9c5 100644
--- a/providers/vercel/models/alibaba/qwen3.5-flash.toml
+++ b/providers/vercel/models/alibaba/qwen3.5-flash.toml
@@ -13,7 +13,7 @@ max = 81_920
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.001
+cache_read = 0.01
 cache_write = 0.125
 
 [limit]
diff --git a/providers/vercel/models/alibaba/qwen3.5-plus.toml b/providers/vercel/models/alibaba/qwen3.5-plus.toml
index e3a91399921..fab6054e8ca 100644
--- a/providers/vercel/models/alibaba/qwen3.5-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3.5-plus.toml
@@ -1,11 +1,18 @@
 base_model = "alibaba/qwen3.5-plus"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1, max = 81_920 }]
 name = "Qwen 3.5 Plus"
 attachment = true
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 81_920
+
 [cost]
 input = 0.4
-output = 2.4
+output = 2.5
 cache_read = 0.04
 cache_write = 0.5
 
diff --git a/providers/vercel/models/alibaba/qwen3.6-plus.toml b/providers/vercel/models/alibaba/qwen3.6-plus.toml
index ff598b0417f..1b29656202b 100644
--- a/providers/vercel/models/alibaba/qwen3.6-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3.6-plus.toml
@@ -12,7 +12,7 @@ max = 131_072
 [cost]
 input = 0.5
 output = 3
-cache_read = 0.1
+cache_read = 0.05
 cache_write = 0.625
 
 [limit]

From b6dfc25160a2bf484d46446ab79bc3ee6f4e8206 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 23:22:56 +0000
Subject: [PATCH 1316/2012] chore(sync): update OpenRouter model catalog
 (#7097)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f7dc249d032..d6a823fa222 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0763
-output = 0.1526
-cache_read = 0.01526
+input = 0.07476
+output = 0.14952
+cache_read = 0.014952
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 70e672577c6..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.975
-output = 11.06
-cache_read = 0.2291
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 018aa6c317f..5c6d5ba351d 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -16,12 +16,12 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.9
-output = 3.069
-cache_read = 0.18
+output = 3
+cache_read = 0.09
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 44128c26e3d5753cb8ba92350fc43db45bb9ba31 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 18:38:38 -0500
Subject: [PATCH 1317/2012] chore(sync): update DigitalOcean model catalog
 (#7094)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4.1-flash.toml            | 18 ++++++++++++++++++
 .../digitalocean/models/glm-5.3-flash.toml     |  2 +-
 2 files changed, 19 insertions(+), 1 deletion(-)
 create mode 100644 providers/digitalocean/models/deepseek-v4.1-flash.toml

diff --git a/providers/digitalocean/models/deepseek-v4.1-flash.toml b/providers/digitalocean/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..736d7e55f2d
--- /dev/null
+++ b/providers/digitalocean/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,18 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+attachment = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text"]
diff --git a/providers/digitalocean/models/glm-5.3-flash.toml b/providers/digitalocean/models/glm-5.3-flash.toml
index f0bf95a9448..ce40eb73fd2 100644
--- a/providers/digitalocean/models/glm-5.3-flash.toml
+++ b/providers/digitalocean/models/glm-5.3-flash.toml
@@ -3,7 +3,7 @@ name = "GLM5.3 Flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["medium", "high", "xhigh"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.15

From 7abebf64f33a07b1a7331d34e6b01035e17bd1bc Mon Sep 17 00:00:00 2001
From: Hugues-Antoine <ha@rabany.eu>
Date: Tue, 15 Sep 2026 01:38:55 +0200
Subject: [PATCH 1318/2012] chore(sync): update cost ovhcloud model qwen*27b
 (#7056)

Co-authored-by: Hugues-Antoine Rabany <ha@aclccp.com>
---
 providers/ovhcloud/models/qwen3.6-27b.toml | 4 ++--
 providers/ovhcloud/models/qwen3.8-27b.toml | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/ovhcloud/models/qwen3.6-27b.toml b/providers/ovhcloud/models/qwen3.6-27b.toml
index b0decce5dc7..c1d7bd3cb28 100644
--- a/providers/ovhcloud/models/qwen3.6-27b.toml
+++ b/providers/ovhcloud/models/qwen3.6-27b.toml
@@ -16,8 +16,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.47
-output = 3.19
+input = 0.46
+output = 3.11
 
 [limit]
 context = 262_144
diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 45ae7920a40..626aa0d7f89 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -22,8 +22,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0
-output = 0
+input = 0.46
+output = 3.11
 
 [limit]
 context = 262_144

From 21cfd9a2c5dd2684d6b9a6b17727b668789501f4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:00:49 +0000
Subject: [PATCH 1319/2012] chore(sync): update Kilo model catalog (#7101)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 3 +--
 .../kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml     | 4 ----
 providers/kilo/models/qwen/qwen3.5-35b-a3b.toml           | 4 ----
 providers/kilo/models/tencent/hy3.toml                    | 6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                   | 1 +
 .../kilo/models/~deepseek/deepseek-flash-latest.toml      | 2 +-
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 7 files changed, 10 insertions(+), 18 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 5d67b671b7a..7c76fb082af 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -9,7 +9,3 @@ values = ["high"]
 [cost]
 input = 0.15
 output = 1.2
-
-[limit]
-context = 262_144
-output = 235_929
diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index bdebd95519c..96076003f63 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -9,9 +9,5 @@ values = ["none", "high"]
 input = 0.1625
 output = 1.3
 
-[limit]
-context = 256_000
-output = 16_384
-
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bb89edf75cf..80b000693ec 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -12,3 +12,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
+output = 128_000
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 51ee78461f9..673b730eb39 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -17,7 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.15
 output = 0.6
-cache_read = 0.003
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From 262e508371b41428c5767b6c82cebad1aaabf1f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:00:52 +0000
Subject: [PATCH 1320/2012] chore(sync): update LLM Gateway model catalog
 (#7100)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/together-ai/gemma-4-31b-it.toml        | 15 ---------------
 .../models/together-ai/gpt-oss-20b.toml           | 11 -----------
 2 files changed, 26 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/together-ai/gemma-4-31b-it.toml
 delete mode 100644 providers/llmgateway-providers/models/together-ai/gpt-oss-20b.toml

diff --git a/providers/llmgateway-providers/models/together-ai/gemma-4-31b-it.toml b/providers/llmgateway-providers/models/together-ai/gemma-4-31b-it.toml
deleted file mode 100644
index 1eaec5c565d..00000000000
--- a/providers/llmgateway-providers/models/together-ai/gemma-4-31b-it.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "google/gemma-4-31b-it"
-name = "Gemma 4 31B IT (Together AI)"
-attachment = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 0.39
-output = 0.97
-
-[modalities]
-input = ["text"]
diff --git a/providers/llmgateway-providers/models/together-ai/gpt-oss-20b.toml b/providers/llmgateway-providers/models/together-ai/gpt-oss-20b.toml
deleted file mode 100644
index ca4ba5e8361..00000000000
--- a/providers/llmgateway-providers/models/together-ai/gpt-oss-20b.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "openai/gpt-oss-20b"
-name = "GPT OSS 20B (Together AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.05
-output = 0.2

From bbe7c0602a628ab5b1d24425afa14281a8fc1c94 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:00:55 +0000
Subject: [PATCH 1321/2012] chore(sync): update NanoGPT model catalog (#7104)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From e8075c03da95b9f4f8996a47267fbad9bc6d86d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:00:58 +0000
Subject: [PATCH 1322/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#7103)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ovhcloud/models/qwen3.6-27b.toml | 4 ++--
 providers/ovhcloud/models/qwen3.8-27b.toml | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/ovhcloud/models/qwen3.6-27b.toml b/providers/ovhcloud/models/qwen3.6-27b.toml
index c1d7bd3cb28..b0decce5dc7 100644
--- a/providers/ovhcloud/models/qwen3.6-27b.toml
+++ b/providers/ovhcloud/models/qwen3.6-27b.toml
@@ -16,8 +16,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.46
-output = 3.11
+input = 0.47
+output = 3.19
 
 [limit]
 context = 262_144
diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 626aa0d7f89..45ae7920a40 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -22,8 +22,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.46
-output = 3.11
+input = 0
+output = 0
 
 [limit]
 context = 262_144

From 33b2a1a33092b0758bac9c5da22356f3ed6c6d6b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:01:01 +0000
Subject: [PATCH 1323/2012] chore(sync): update Deep Infra model catalog
 (#7105)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/Qwen/Qwen3.8-27B.toml | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/providers/deepinfra/models/Qwen/Qwen3.8-27B.toml b/providers/deepinfra/models/Qwen/Qwen3.8-27B.toml
index 75f14120f8d..fd86f3b9a1a 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.8-27B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.8-27B.toml
@@ -8,9 +8,6 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.4
-output = 3
-cache_read = 0.04
-
-[modalities]
-input = ["text", "image"]
+input = 0.2
+output = 2.5
+cache_read = 0.05

From 604a0c19922842002a0865c8cb5a573c1b269a5f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:01:03 +0000
Subject: [PATCH 1324/2012] chore(sync): update OpenRouter model catalog
 (#7102)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 .../models/qwen/qwen3-next-80b-a3b-thinking.toml          | 1 -
 providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml     | 8 ++------
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 7 ++++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 8 files changed, 20 insertions(+), 25 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d6a823fa222..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07476
-output = 0.14952
-cache_read = 0.014952
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..451e7c3012c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.9834
+output = 2.9502
+cache_read = 0.03278
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 35f13bee686..14e222d894d 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -8,4 +8,3 @@ output = 1.2
 
 [limit]
 context = 262_144
-output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index 8257248c271..34aa7a00097 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -6,12 +6,8 @@ base_model = "alibaba/qwen3.5-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.3125
-output = 1.25
-cache_read = 0.15625
-
-[limit]
-output = 16_384
+input = 0.1625
+output = 1.3
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 3adac002861..cf2eefb64f7 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.6832
-output = 2.1472
-cache_read = 0.12688
+input = 1.4
+output = 4.4
+cache_read = 0.14
 
 [limit]
 context = 1_048_576
+output = 128_000
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 182582c86a5..dc34ba0ccf7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.15
 output = 0.6
-cache_read = 0.003
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From d5f38f55abab60b8c32f1d024a9d584907574369 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:37:34 +0000
Subject: [PATCH 1325/2012] chore(sync): update OpenRouter model catalog
 (#7106)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..93d4a9d7ab9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 2acb5dba9962ab34372786b3fa762d037cb578c2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 01:37:39 +0000
Subject: [PATCH 1326/2012] chore(sync): update Kilo model catalog (#7107)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..a43a707c98d 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 9ed35e8bcfc7ef85cac4b19374917084ded5edfd Mon Sep 17 00:00:00 2001
From: Julio Monteiro <julio@monteiro.eti.br>
Date: Mon, 14 Sep 2026 23:20:30 -0300
Subject: [PATCH 1327/2012] Add published per-token pricing for Ollama Cloud
 models (#7108)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Add published per-token pricing for Ollama Cloud models

Ollama publishes per-1M-token rates on each model's library page
(e.g. https://ollama.com/library/glm-5.3) but the ollama-cloud
provider entries carried no cost data, so consumers computing spend
from this catalog (opencode, ccusage, etc.) report /bin/zsh.00 for these
models.

Adds [cost] blocks to the 16 models with published rates, sourced
from each model's library page on 2026-09-14. Cross-checked against
opencode-go entries for every overlapping model (glm-5.2, glm-5.3,
kimi-k3, kimi-k2.6, minimax-m2.7, deepseek-v4-pro, deepseek-v4.1-flash)
and against the zai first-party entries for the GLM family — all match.

DeepSeek models bill Base (off-peak) and Peak (2x) rates; the schema's
cost tiers only express context-based pricing, so the Base rate is
authored with a top-of-file comment noting the Peak multiplier.

Models with no published price (gpt-oss, gemma4, qwen3.5,
nemotron-3-super, nemotron-3-nano) and retired models (kimi-k2.5,
minimax-m2.5) are left without cost, matching the AGENTS.md guidance
for request-only / no public price entries.

* Add pricing for the 6 remaining served Ollama Cloud models

Completes the cost pass for every model Ollama's cloud API serves:
gemma4:31b, gpt-oss:120b, gpt-oss:20b, nemotron-3-super,
nemotron-3-nano:30b, and qwen3.5:397b. Also cites the canonical
https://ollama.com/pricing page in every leading header instead of
per-model library pages, and references issue #5913 on the DeepSeek
entries, whose Base/Peak rates the cost schema cannot yet express.
---
 providers/ollama-cloud/models/deepseek-v4-flash.toml  |  9 +++++++++
 .../ollama-cloud/models/deepseek-v4-flash:0731.toml   | 11 ++++++++++-
 providers/ollama-cloud/models/deepseek-v4-pro.toml    |  9 +++++++++
 .../ollama-cloud/models/deepseek-v4-pro:0813.toml     |  9 +++++++++
 .../ollama-cloud/models/deepseek-v4.1-flash.toml      | 11 ++++++++++-
 providers/ollama-cloud/models/gemma4:31b.toml         |  6 ++++++
 providers/ollama-cloud/models/glm-5.1.toml            |  6 ++++++
 providers/ollama-cloud/models/glm-5.2.toml            |  8 +++++++-
 providers/ollama-cloud/models/glm-5.3-flash.toml      |  6 ++++++
 providers/ollama-cloud/models/glm-5.3.toml            |  8 +++++++-
 providers/ollama-cloud/models/gpt-oss:120b.toml       |  6 ++++++
 providers/ollama-cloud/models/gpt-oss:20b.toml        |  6 ++++++
 providers/ollama-cloud/models/kimi-k2.6.toml          |  6 ++++++
 providers/ollama-cloud/models/kimi-k2.7-code.toml     |  6 ++++++
 providers/ollama-cloud/models/kimi-k3.toml            |  6 ++++++
 providers/ollama-cloud/models/minimax-m2.7.toml       |  6 ++++++
 providers/ollama-cloud/models/minimax-m3.toml         |  6 ++++++
 .../ollama-cloud/models/mistral-large-3:675b.toml     |  5 +++++
 .../ollama-cloud/models/nemotron-3-nano:30b.toml      |  5 +++++
 providers/ollama-cloud/models/nemotron-3-super.toml   |  6 ++++++
 providers/ollama-cloud/models/nemotron-3-ultra.toml   |  6 ++++++
 providers/ollama-cloud/models/qwen3.5:397b.toml       |  5 +++++
 22 files changed, 148 insertions(+), 4 deletions(-)

diff --git a/providers/ollama-cloud/models/deepseek-v4-flash.toml b/providers/ollama-cloud/models/deepseek-v4-flash.toml
index 75c065b672c..515c3a01a0e 100644
--- a/providers/ollama-cloud/models/deepseek-v4-flash.toml
+++ b/providers/ollama-cloud/models/deepseek-v4-flash.toml
@@ -1,3 +1,7 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15). Ollama bills
+# Base (off-peak) and Peak (2x, 12:00-18:00 UTC Mon-Fri) rates; the schema
+# cannot express time-based tiers, so the Base rate is recorded. See
+# https://github.com/anomalyco/models.dev/issues/5913.
 name = "deepseek-v4-flash"
 description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
 family = "deepseek-flash"
@@ -16,3 +20,8 @@ output = 1048576
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
diff --git a/providers/ollama-cloud/models/deepseek-v4-flash:0731.toml b/providers/ollama-cloud/models/deepseek-v4-flash:0731.toml
index 9666e6cb2dd..08a5f37bd8e 100644
--- a/providers/ollama-cloud/models/deepseek-v4-flash:0731.toml
+++ b/providers/ollama-cloud/models/deepseek-v4-flash:0731.toml
@@ -1,3 +1,7 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15). Ollama bills
+# Base (off-peak) and Peak (2x, 12:00-18:00 UTC Mon-Fri) rates; the schema
+# cannot express time-based tiers, so the Base rate is recorded. See
+# https://github.com/anomalyco/models.dev/issues/5913.
 # Toggle: native POST /api/chat and /api/generate use `think = true|false`.
 # Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
 # or `reasoning.effort` with high|max (DeepSeek V4 maps low/medium -> high).
@@ -9,4 +13,9 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "
 
 [limit]
 context = 1048576
-output = 1048576
\ No newline at end of file
+output = 1048576
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
diff --git a/providers/ollama-cloud/models/deepseek-v4-pro.toml b/providers/ollama-cloud/models/deepseek-v4-pro.toml
index 09643bf6663..95855eac75c 100644
--- a/providers/ollama-cloud/models/deepseek-v4-pro.toml
+++ b/providers/ollama-cloud/models/deepseek-v4-pro.toml
@@ -1,3 +1,7 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15). Ollama bills
+# Base (off-peak) and Peak (2x, 12:00-18:00 UTC Mon-Fri) rates; the schema
+# cannot express time-based tiers, so the Base rate is recorded. See
+# https://github.com/anomalyco/models.dev/issues/5913.
 name = "deepseek-v4-pro"
 description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
 family = "deepseek-thinking"
@@ -16,3 +20,8 @@ output = 1048576
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.66
+output = 1.98
+cache_read = 0.022
diff --git a/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml b/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml
index 9ae3ed43255..97cd7c67de3 100644
--- a/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml
+++ b/providers/ollama-cloud/models/deepseek-v4-pro:0813.toml
@@ -1,3 +1,7 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15). Ollama bills
+# Base (off-peak) and Peak (2x, 12:00-18:00 UTC Mon-Fri) rates; the schema
+# cannot express time-based tiers, so the Base rate is recorded. See
+# https://github.com/anomalyco/models.dev/issues/5913.
 # Toggle: native POST /api/chat and /api/generate use `think = true|false`.
 # Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
 # or `reasoning.effort` with high|max (DeepSeek V4 maps low/medium -> high).
@@ -12,3 +16,8 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "
 [limit]
 context = 1048576
 output = 1048576
+
+[cost]
+input = 0.66
+output = 1.98
+cache_read = 0.022
diff --git a/providers/ollama-cloud/models/deepseek-v4.1-flash.toml b/providers/ollama-cloud/models/deepseek-v4.1-flash.toml
index 84ecfb2ff46..298a6d57684 100644
--- a/providers/ollama-cloud/models/deepseek-v4.1-flash.toml
+++ b/providers/ollama-cloud/models/deepseek-v4.1-flash.toml
@@ -1,3 +1,7 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15). Ollama bills
+# Base (off-peak) and Peak (2x, 12:00-18:00 UTC Mon-Fri) rates; the schema
+# cannot express time-based tiers, so the Base rate is recorded. See
+# https://github.com/anomalyco/models.dev/issues/5913.
 # Toggle: native POST /api/chat and /api/generate use `think = true|false`.
 # Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
 # or `reasoning.effort` with low|high|max (Flash maps requested low -> low).
@@ -11,4 +15,9 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "h
 field = "reasoning_content"
 
 [limit]
-context = 1048576
\ No newline at end of file
+context = 1048576
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.003
diff --git a/providers/ollama-cloud/models/gemma4:31b.toml b/providers/ollama-cloud/models/gemma4:31b.toml
index ff9564944cd..f6cec87c879 100644
--- a/providers/ollama-cloud/models/gemma4:31b.toml
+++ b/providers/ollama-cloud/models/gemma4:31b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "gemma4:31b"
 description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
 family = "gemma"
@@ -17,3 +18,8 @@ output = 262144
 [modalities]
 input = ["text","image"]
 output = ["text"]
+
+[cost]
+input = 0.14
+output = 0.4
+cache_read = 0.05
diff --git a/providers/ollama-cloud/models/glm-5.1.toml b/providers/ollama-cloud/models/glm-5.1.toml
index d546ca5ed92..0ee8e039d72 100644
--- a/providers/ollama-cloud/models/glm-5.1.toml
+++ b/providers/ollama-cloud/models/glm-5.1.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "glm-5.1"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 family = "glm"
@@ -19,3 +20,8 @@ output = 131072
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 1.0
+output = 3.2
+cache_read = 0.2
diff --git a/providers/ollama-cloud/models/glm-5.2.toml b/providers/ollama-cloud/models/glm-5.2.toml
index dcb62d38c8a..f2dfc95e184 100644
--- a/providers/ollama-cloud/models/glm-5.2.toml
+++ b/providers/ollama-cloud/models/glm-5.2.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 base_model = "zhipuai/glm-5.2"
 open_weights = true
 
@@ -9,4 +10,9 @@ values = ["high", "max"]
 field = "reasoning_content"
 
 [limit]
-context = 976_000
\ No newline at end of file
+context = 976_000
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/ollama-cloud/models/glm-5.3-flash.toml b/providers/ollama-cloud/models/glm-5.3-flash.toml
index fa66dc84ecc..acff7eff693 100644
--- a/providers/ollama-cloud/models/glm-5.3-flash.toml
+++ b/providers/ollama-cloud/models/glm-5.3-flash.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 base_model = "zhipuai/glm-5.3-flash"
 open_weights = true
 
@@ -7,3 +8,8 @@ values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
diff --git a/providers/ollama-cloud/models/glm-5.3.toml b/providers/ollama-cloud/models/glm-5.3.toml
index d16b23a5f89..b2d0787d7df 100644
--- a/providers/ollama-cloud/models/glm-5.3.toml
+++ b/providers/ollama-cloud/models/glm-5.3.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 # Effort: OpenAI POST /v1/chat/completions accepts top-level `reasoning_effort`
 # or `reasoning.effort` with low|high|max (GLM-5.3 always reasons; thinking
 # cannot be disabled, so there is no toggle).
@@ -12,4 +13,9 @@ reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 field = "reasoning_content"
 
 [limit]
-context = 1_048_576
\ No newline at end of file
+context = 1_048_576
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
diff --git a/providers/ollama-cloud/models/gpt-oss:120b.toml b/providers/ollama-cloud/models/gpt-oss:120b.toml
index 45d64801792..67a40918ed6 100644
--- a/providers/ollama-cloud/models/gpt-oss:120b.toml
+++ b/providers/ollama-cloud/models/gpt-oss:120b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "gpt-oss:120b"
 description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
 family = "gpt-oss"
@@ -16,3 +17,8 @@ output = 32768
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.014
diff --git a/providers/ollama-cloud/models/gpt-oss:20b.toml b/providers/ollama-cloud/models/gpt-oss:20b.toml
index 7e355ac8e7e..da92e1826d5 100644
--- a/providers/ollama-cloud/models/gpt-oss:20b.toml
+++ b/providers/ollama-cloud/models/gpt-oss:20b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "gpt-oss:20b"
 description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
 family = "gpt-oss"
@@ -16,3 +17,8 @@ output = 32768
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.07
+output = 0.3
+cache_read = 0.035
diff --git a/providers/ollama-cloud/models/kimi-k2.6.toml b/providers/ollama-cloud/models/kimi-k2.6.toml
index 8f5f6011b45..644ab635138 100644
--- a/providers/ollama-cloud/models/kimi-k2.6.toml
+++ b/providers/ollama-cloud/models/kimi-k2.6.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "kimi-k2.6"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 family = "kimi-k2"
@@ -16,3 +17,8 @@ output = 262144
 [modalities]
 input = ["text", "image"]
 output = ["text"]
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.16
diff --git a/providers/ollama-cloud/models/kimi-k2.7-code.toml b/providers/ollama-cloud/models/kimi-k2.7-code.toml
index e5ddc85c644..faac2d00fb3 100644
--- a/providers/ollama-cloud/models/kimi-k2.7-code.toml
+++ b/providers/ollama-cloud/models/kimi-k2.7-code.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 base_model = "moonshotai/kimi-k2.7-code"
 name = "kimi-k2.7-code"
 reasoning_options = [{ type = "toggle" }]
@@ -5,3 +6,8 @@ reasoning_options = [{ type = "toggle" }]
 [modalities]
 input = ["text", "image"]
 output = ["text"]
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.19
diff --git a/providers/ollama-cloud/models/kimi-k3.toml b/providers/ollama-cloud/models/kimi-k3.toml
index 8c5b5b050bb..da32429a156 100644
--- a/providers/ollama-cloud/models/kimi-k3.toml
+++ b/providers/ollama-cloud/models/kimi-k3.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 # Ollama Cloud direct API model ID verified from https://docs.ollama.com/api/tags.
 # Library page documents Kimi K3 as a cloud thinking model with 1M context
 # and text/image input.
@@ -13,3 +14,8 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "h
 
 [modalities]
 input = ["text", "image"]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
diff --git a/providers/ollama-cloud/models/minimax-m2.7.toml b/providers/ollama-cloud/models/minimax-m2.7.toml
index 88e4a404a96..fe6f862b2a7 100644
--- a/providers/ollama-cloud/models/minimax-m2.7.toml
+++ b/providers/ollama-cloud/models/minimax-m2.7.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "minimax-m2.7"
 description = "MiniMax model for chat, coding, office work, and agentic tasks"
 family = "minimax"
@@ -16,3 +17,8 @@ output = 196608
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
diff --git a/providers/ollama-cloud/models/minimax-m3.toml b/providers/ollama-cloud/models/minimax-m3.toml
index c8155d651db..1505acb8787 100644
--- a/providers/ollama-cloud/models/minimax-m3.toml
+++ b/providers/ollama-cloud/models/minimax-m3.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "minimax-m3"
 description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
 family = "minimax-m3"
@@ -19,3 +20,8 @@ output = 131_072
 [modalities]
 input = ["text", "image", "video"]
 output = ["text"]
+
+[cost]
+input = 0.6
+output = 2.4
+cache_read = 0.12
diff --git a/providers/ollama-cloud/models/mistral-large-3:675b.toml b/providers/ollama-cloud/models/mistral-large-3:675b.toml
index 905503bb479..d49e67e4ef0 100644
--- a/providers/ollama-cloud/models/mistral-large-3:675b.toml
+++ b/providers/ollama-cloud/models/mistral-large-3:675b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "mistral-large-3:675b"
 description = "Flagship Mistral model for advanced reasoning, coding, and multilingual work"
 family = "mistral-large"
@@ -15,3 +16,7 @@ output = 262144
 [modalities]
 input = ["text","image"]
 output = ["text"]
+
+[cost]
+input = 0.5
+output = 1.5
diff --git a/providers/ollama-cloud/models/nemotron-3-nano:30b.toml b/providers/ollama-cloud/models/nemotron-3-nano:30b.toml
index 179fd8e2beb..184a873907e 100644
--- a/providers/ollama-cloud/models/nemotron-3-nano:30b.toml
+++ b/providers/ollama-cloud/models/nemotron-3-nano:30b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "nemotron-3-nano:30b"
 base_model = "nvidia/nemotron-3-nano-30b-a3b"
 family = "nemotron"
@@ -16,3 +17,7 @@ output = 131072
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.06
+output = 0.24
diff --git a/providers/ollama-cloud/models/nemotron-3-super.toml b/providers/ollama-cloud/models/nemotron-3-super.toml
index b72a6d91e5d..8a4ac22dec6 100644
--- a/providers/ollama-cloud/models/nemotron-3-super.toml
+++ b/providers/ollama-cloud/models/nemotron-3-super.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "nemotron-3-super"
 base_model = "nvidia/nemotron-3-super-120b-a12b"
 family = "nemotron"
@@ -16,3 +17,8 @@ output = 65536
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[cost]
+input = 0.015
+output = 0.6
+cache_read = 0.015
diff --git a/providers/ollama-cloud/models/nemotron-3-ultra.toml b/providers/ollama-cloud/models/nemotron-3-ultra.toml
index 23cbabaecfb..6490eeeca84 100644
--- a/providers/ollama-cloud/models/nemotron-3-ultra.toml
+++ b/providers/ollama-cloud/models/nemotron-3-ultra.toml
@@ -1,6 +1,12 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "nemotron-3-ultra"
 base_model = "nvidia/nemotron-3-ultra-550b-a55b"
 reasoning_options = [{ type = "toggle" }]
 
 [limit]
 context = 262144
+
+[cost]
+input = 0.1
+output = 3
+cache_read = 0.1
diff --git a/providers/ollama-cloud/models/qwen3.5:397b.toml b/providers/ollama-cloud/models/qwen3.5:397b.toml
index 25442c93f33..956aa90c51c 100644
--- a/providers/ollama-cloud/models/qwen3.5:397b.toml
+++ b/providers/ollama-cloud/models/qwen3.5:397b.toml
@@ -1,3 +1,4 @@
+# Cost from https://ollama.com/pricing (accessed 2026-09-15).
 name = "qwen3.5:397b"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
@@ -19,3 +20,7 @@ output = 65536
 [modalities]
 input = ["text", "image"]
 output = ["text"]
+
+[cost]
+input = 0.6
+output = 3.6

From b7b26bfd734b31a7f99f5655e60a84406afbfe74 Mon Sep 17 00:00:00 2001
From: Vansh <72547809+vansh-one@users.noreply.github.com>
Date: Tue, 15 Sep 2026 07:51:23 +0530
Subject: [PATCH 1328/2012] feat(provider): add Vispark Lab with Vision
 Small/Medium/Large (#6997)

* feat(provider): add Vispark Lab with Vision Small/Medium/Large

* use neutral capability-focused model descriptions
---
 labs/vispark/lab.toml                         |  1 +
 labs/vispark/logo.svg                         |  3 ++
 models/vispark/vision-large.toml              | 26 ++++++++++++++
 models/vispark/vision-medium.toml             | 26 ++++++++++++++
 models/vispark/vision-small.toml              | 26 ++++++++++++++
 providers/vispark/logo.svg                    |  3 ++
 .../vispark/models/vispark/vision-large.toml  | 36 +++++++++++++++++++
 .../vispark/models/vispark/vision-medium.toml | 36 +++++++++++++++++++
 .../vispark/models/vispark/vision-small.toml  | 36 +++++++++++++++++++
 providers/vispark/provider.toml               | 11 ++++++
 10 files changed, 204 insertions(+)
 create mode 100644 labs/vispark/lab.toml
 create mode 100644 labs/vispark/logo.svg
 create mode 100644 models/vispark/vision-large.toml
 create mode 100644 models/vispark/vision-medium.toml
 create mode 100644 models/vispark/vision-small.toml
 create mode 100644 providers/vispark/logo.svg
 create mode 100644 providers/vispark/models/vispark/vision-large.toml
 create mode 100644 providers/vispark/models/vispark/vision-medium.toml
 create mode 100644 providers/vispark/models/vispark/vision-small.toml
 create mode 100644 providers/vispark/provider.toml

diff --git a/labs/vispark/lab.toml b/labs/vispark/lab.toml
new file mode 100644
index 00000000000..3aebe3fe5da
--- /dev/null
+++ b/labs/vispark/lab.toml
@@ -0,0 +1 @@
+description = "Vispark builds India-centered foundational multimodal intelligence (Vision) with 1M-token context, served via Vispark Lab's API."
diff --git a/labs/vispark/logo.svg b/labs/vispark/logo.svg
new file mode 100644
index 00000000000..2348c5a2f7f
--- /dev/null
+++ b/labs/vispark/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path fill-rule="evenodd" d="M0.33,0.22L0.22,0.33L0.15,0.52L0.17,0.7L0.68,1.66L0.85,2.03L0.98,2.24L1.29,2.88L2.09,4.36L2.22,4.66L2.35,4.86L2.39,4.97L3.72,7.49L4.03,8.14L4.16,8.34L4.33,8.71L4.59,9.15L4.75,9.52L5.12,10.19L5.42,10.8L5.55,11L5.69,11.33L6.25,12.35L6.29,12.46L7.19,14.16L7.51,14.81L7.62,14.98L7.67,15.12L8.56,16.79L8.88,17.44L9,17.64L9.32,18.29L9.43,18.45L9.6,18.82L9.97,19.49L10.26,20.1L10.41,20.34L10.47,20.49L10.65,20.8L10.96,21.45L11.09,21.65L11.41,22.3L11.52,22.47L11.83,23.11L12.07,23.5L12.26,23.69L12.35,23.74L12.61,23.83L12.76,23.83L13,23.74L13.16,23.61L13.39,23.26L13.44,23.09L13.98,21.98L14.05,21.78L14.11,21.71L14.14,21.58L14.22,21.47L14.27,21.3L14.68,20.47L14.76,20.27L14.81,20.19L14.87,20.02L14.92,19.95L14.96,19.82L15.51,18.67L15.53,18.58L15.64,18.4L15.66,18.31L15.7,18.27L15.79,18.03L16.22,17.16L16.31,16.92L16.35,16.88L16.35,16.83L16.47,16.6L16.49,16.51L16.53,16.47L16.62,16.23L17.03,15.4L17.18,15.03L17.23,14.96L17.33,14.72L17.84,13.65L17.9,13.48L17.95,13.41L18.01,13.24L18.56,12.09L18.73,11.69L18.79,11.61L18.84,11.45L19.27,10.58L19.32,10.41L19.4,10.3L19.43,10.17L19.51,10.06L19.54,9.93L20.91,7.03L20.93,6.93L21.04,6.75L21.06,6.66L21.1,6.62L21.21,6.34L21.61,5.51L21.63,5.42L21.87,4.96L21.89,4.86L22.32,3.99L22.41,3.75L22.45,3.72L22.47,3.62L22.58,3.44L22.59,3.35L22.63,3.31L22.72,3.07L22.87,2.79L23.24,1.96L23.28,1.92L23.3,1.83L23.35,1.76L23.56,1.28L23.7,1.02L23.83,0.61L23.83,0.48L23.8,0.35L23.67,0.22L23.56,0.18L14.87,0.18L14.68,0.3L14.51,0.48L14.44,0.63L14.4,0.91L14.29,1.31L14.29,1.42L14.18,1.83L14.13,2.18L14.02,2.61L14.02,2.7L13.74,3.88L13.74,3.99L13.61,4.47L13.59,4.66L13.53,4.84L13.53,5.07L13.59,5.2L13.72,5.31L13.79,5.34L14.02,5.34L14.98,5.12L15.18,5.1L15.68,4.97L15.79,4.97L16.29,4.84L16.4,4.84L16.99,4.7L17.14,4.68L17.27,4.7L17.47,4.86L17.53,5.01L17.53,5.12L17.45,5.36L17.34,5.6L17.34,5.66L17.21,5.95L17.2,6.06L17.1,6.25L17.03,6.51L16.94,6.69L16.88,6.9L16.79,7.1L16.79,7.16L16.64,7.49L16.64,7.54L16.53,7.78L16.51,7.9L16.36,8.23L16.36,8.28L16.03,9.15L16.03,9.37L16.09,9.49L16.18,9.58L16.36,9.65L16.46,9.65L16.55,9.61L16.77,9.63L16.86,9.69L16.96,9.8L17.01,9.93L17.01,10.08L16.23,11.63L15.98,12.09L15.81,12.46L15.7,12.63L15.11,13.83L14.9,14.18L14.42,15.16L13.48,16.96L13.31,17.33L13.2,17.49L13.04,17.84L12.89,17.95L12.65,17.97L12.5,17.9L12.41,17.79L12.35,17.66L12.35,17.51L12.46,16.73L12.46,16.51L12.57,15.73L12.63,15.05L12.74,14.26L12.74,14.05L12.83,13.42L12.81,13.11L12.74,13L12.63,12.91L12.35,12.89L10.96,13.31L10.72,13.33L10.59,13.28L10.47,13.13L10.43,13.04L10.45,12.78L10.52,12.59L10.52,12.52L10.65,12.13L10.69,11.93L10.76,11.74L10.8,11.54L10.93,11.17L10.95,11.02L11.06,10.71L11.08,10.56L11.2,10.19L11.24,9.98L11.32,9.8L11.35,9.6L11.45,9.32L11.45,9.21L11.37,9.02L11.28,8.93L11.17,8.88L10.95,8.88L10.63,9L10.02,9.19L9.82,9.28L9.71,9.3L8.04,9.87L7.88,9.87L7.77,9.84L7.65,9.73L7.58,9.56L7.62,9.21L7.75,8.67L7.75,8.56L7.86,8.12L7.9,7.82L8.02,7.29L8.02,7.17L8.17,6.56L8.17,6.45L8.3,5.92L8.3,5.81L8.41,5.36L8.45,5.07L8.56,4.62L8.63,4.16L8.73,3.81L8.75,3.61L8.86,3.16L8.88,2.96L9.13,1.78L9.13,1.66L9.36,0.67L9.36,0.52L9.26,0.31L9.13,0.2L8.99,0.15L0.5,0.15Z"/>
+</svg>
diff --git a/models/vispark/vision-large.toml b/models/vispark/vision-large.toml
new file mode 100644
index 00000000000..0c850177b0c
--- /dev/null
+++ b/models/vispark/vision-large.toml
@@ -0,0 +1,26 @@
+# Lab metadata for Vispark Vision Large.
+# Provider-agnostic facts about the model Vispark built.
+# Served first-party by Vispark Lab via OpenAI-compatible API: https://api.lab.vispark.in/v1
+# Model list: GET https://api.lab.vispark.in/v1/models returns vispark/vision-large
+# (verified live 2026-09-13; context_length 1000000, max_output_length 65536 per entry).
+# Chat: POST https://api.lab.vispark.in/v1/chat/completions.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Large"
+description = "Most capable Vision model for complex reasoning, detailed media analysis, and structured output over a 1M-token context window."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/models/vispark/vision-medium.toml b/models/vispark/vision-medium.toml
new file mode 100644
index 00000000000..89b8b000c7f
--- /dev/null
+++ b/models/vispark/vision-medium.toml
@@ -0,0 +1,26 @@
+# Lab metadata for Vispark Vision Medium.
+# Provider-agnostic facts about the model Vispark built.
+# Served first-party by Vispark Lab via OpenAI-compatible API: https://api.lab.vispark.in/v1
+# Model list: GET https://api.lab.vispark.in/v1/models returns vispark/vision-medium
+# (verified live 2026-09-13; context_length 1000000, max_output_length 65536 per entry).
+# Chat: POST https://api.lab.vispark.in/v1/chat/completions.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Medium"
+description = "Balanced multimodal model pairing a 1M-token context window with deeper reasoning for analysis, content creation, and tool use across text, image, audio, video, and PDF inputs."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/models/vispark/vision-small.toml b/models/vispark/vision-small.toml
new file mode 100644
index 00000000000..2dce89d0c4f
--- /dev/null
+++ b/models/vispark/vision-small.toml
@@ -0,0 +1,26 @@
+# Lab metadata for Vispark Vision Small.
+# Provider-agnostic facts about the model Vispark built.
+# Served first-party by Vispark Lab via OpenAI-compatible API: https://api.lab.vispark.in/v1
+# Model list: GET https://api.lab.vispark.in/v1/models returns vispark/vision-small
+# (verified live 2026-09-13; context_length 1000000, max_output_length 65536 per entry).
+# Chat: POST https://api.lab.vispark.in/v1/chat/completions.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Small"
+description = "Fast, low-cost multimodal model for understanding text, images, audio, video, and PDFs, with tool calling and a 1M-token context window."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/providers/vispark/logo.svg b/providers/vispark/logo.svg
new file mode 100644
index 00000000000..2348c5a2f7f
--- /dev/null
+++ b/providers/vispark/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path fill-rule="evenodd" d="M0.33,0.22L0.22,0.33L0.15,0.52L0.17,0.7L0.68,1.66L0.85,2.03L0.98,2.24L1.29,2.88L2.09,4.36L2.22,4.66L2.35,4.86L2.39,4.97L3.72,7.49L4.03,8.14L4.16,8.34L4.33,8.71L4.59,9.15L4.75,9.52L5.12,10.19L5.42,10.8L5.55,11L5.69,11.33L6.25,12.35L6.29,12.46L7.19,14.16L7.51,14.81L7.62,14.98L7.67,15.12L8.56,16.79L8.88,17.44L9,17.64L9.32,18.29L9.43,18.45L9.6,18.82L9.97,19.49L10.26,20.1L10.41,20.34L10.47,20.49L10.65,20.8L10.96,21.45L11.09,21.65L11.41,22.3L11.52,22.47L11.83,23.11L12.07,23.5L12.26,23.69L12.35,23.74L12.61,23.83L12.76,23.83L13,23.74L13.16,23.61L13.39,23.26L13.44,23.09L13.98,21.98L14.05,21.78L14.11,21.71L14.14,21.58L14.22,21.47L14.27,21.3L14.68,20.47L14.76,20.27L14.81,20.19L14.87,20.02L14.92,19.95L14.96,19.82L15.51,18.67L15.53,18.58L15.64,18.4L15.66,18.31L15.7,18.27L15.79,18.03L16.22,17.16L16.31,16.92L16.35,16.88L16.35,16.83L16.47,16.6L16.49,16.51L16.53,16.47L16.62,16.23L17.03,15.4L17.18,15.03L17.23,14.96L17.33,14.72L17.84,13.65L17.9,13.48L17.95,13.41L18.01,13.24L18.56,12.09L18.73,11.69L18.79,11.61L18.84,11.45L19.27,10.58L19.32,10.41L19.4,10.3L19.43,10.17L19.51,10.06L19.54,9.93L20.91,7.03L20.93,6.93L21.04,6.75L21.06,6.66L21.1,6.62L21.21,6.34L21.61,5.51L21.63,5.42L21.87,4.96L21.89,4.86L22.32,3.99L22.41,3.75L22.45,3.72L22.47,3.62L22.58,3.44L22.59,3.35L22.63,3.31L22.72,3.07L22.87,2.79L23.24,1.96L23.28,1.92L23.3,1.83L23.35,1.76L23.56,1.28L23.7,1.02L23.83,0.61L23.83,0.48L23.8,0.35L23.67,0.22L23.56,0.18L14.87,0.18L14.68,0.3L14.51,0.48L14.44,0.63L14.4,0.91L14.29,1.31L14.29,1.42L14.18,1.83L14.13,2.18L14.02,2.61L14.02,2.7L13.74,3.88L13.74,3.99L13.61,4.47L13.59,4.66L13.53,4.84L13.53,5.07L13.59,5.2L13.72,5.31L13.79,5.34L14.02,5.34L14.98,5.12L15.18,5.1L15.68,4.97L15.79,4.97L16.29,4.84L16.4,4.84L16.99,4.7L17.14,4.68L17.27,4.7L17.47,4.86L17.53,5.01L17.53,5.12L17.45,5.36L17.34,5.6L17.34,5.66L17.21,5.95L17.2,6.06L17.1,6.25L17.03,6.51L16.94,6.69L16.88,6.9L16.79,7.1L16.79,7.16L16.64,7.49L16.64,7.54L16.53,7.78L16.51,7.9L16.36,8.23L16.36,8.28L16.03,9.15L16.03,9.37L16.09,9.49L16.18,9.58L16.36,9.65L16.46,9.65L16.55,9.61L16.77,9.63L16.86,9.69L16.96,9.8L17.01,9.93L17.01,10.08L16.23,11.63L15.98,12.09L15.81,12.46L15.7,12.63L15.11,13.83L14.9,14.18L14.42,15.16L13.48,16.96L13.31,17.33L13.2,17.49L13.04,17.84L12.89,17.95L12.65,17.97L12.5,17.9L12.41,17.79L12.35,17.66L12.35,17.51L12.46,16.73L12.46,16.51L12.57,15.73L12.63,15.05L12.74,14.26L12.74,14.05L12.83,13.42L12.81,13.11L12.74,13L12.63,12.91L12.35,12.89L10.96,13.31L10.72,13.33L10.59,13.28L10.47,13.13L10.43,13.04L10.45,12.78L10.52,12.59L10.52,12.52L10.65,12.13L10.69,11.93L10.76,11.74L10.8,11.54L10.93,11.17L10.95,11.02L11.06,10.71L11.08,10.56L11.2,10.19L11.24,9.98L11.32,9.8L11.35,9.6L11.45,9.32L11.45,9.21L11.37,9.02L11.28,8.93L11.17,8.88L10.95,8.88L10.63,9L10.02,9.19L9.82,9.28L9.71,9.3L8.04,9.87L7.88,9.87L7.77,9.84L7.65,9.73L7.58,9.56L7.62,9.21L7.75,8.67L7.75,8.56L7.86,8.12L7.9,7.82L8.02,7.29L8.02,7.17L8.17,6.56L8.17,6.45L8.3,5.92L8.3,5.81L8.41,5.36L8.45,5.07L8.56,4.62L8.63,4.16L8.73,3.81L8.75,3.61L8.86,3.16L8.88,2.96L9.13,1.78L9.13,1.66L9.36,0.67L9.36,0.52L9.26,0.31L9.13,0.2L8.99,0.15L0.5,0.15Z"/>
+</svg>
diff --git a/providers/vispark/models/vispark/vision-large.toml b/providers/vispark/models/vispark/vision-large.toml
new file mode 100644
index 00000000000..9238a65cb89
--- /dev/null
+++ b/providers/vispark/models/vispark/vision-large.toml
@@ -0,0 +1,36 @@
+# Vispark: Vision Large — first-party host (provider IS the lab, full inline allowed).
+# API model ID: vispark/vision-large — GET https://api.lab.vispark.in/v1/models (verified live 2026-09-13).
+# Cost source, USD/MTok, used verbatim: the provider's own models endpoint publishes
+# pricing.prompt = "0.00000737" and pricing.completion = "0.00002211" USD per token
+# (per-token USD strings) = input 7.37, output 22.11 per MTok.
+# Reasoning: always-on, no caller control — accepted params are only
+# temperature/max_tokens/top_p/top_k/stop/seed/frequency_penalty/presence_penalty/tools/
+# tool_choice/response_format — no reasoning_effort/thinking/enable_thinking/budget field;
+# Capabilities: temperature, top_p/top_k/stop/seed, tools, response_format json_object/json_schema,
+# multimodal input text+image+audio+video+pdf (base64/URL), text output, streaming + usage.
+# Limits: context 1_000_000 (context_length), output 65_536 (max_output_length) — same endpoint.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Large"
+description = "Most capable Vision model for complex reasoning, detailed media analysis, and structured output over a 1M-token context window."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 7.37
+output = 22.11
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/providers/vispark/models/vispark/vision-medium.toml b/providers/vispark/models/vispark/vision-medium.toml
new file mode 100644
index 00000000000..08a3a70579d
--- /dev/null
+++ b/providers/vispark/models/vispark/vision-medium.toml
@@ -0,0 +1,36 @@
+# Vispark: Vision Medium — first-party host (provider IS the lab, full inline allowed).
+# API model ID: vispark/vision-medium (default model) — GET https://api.lab.vispark.in/v1/models (verified live 2026-09-13).
+# Cost source, USD/MTok, used verbatim: the provider's own models endpoint publishes
+# pricing.prompt = "0.00000421" and pricing.completion = "0.00001263" USD per token
+# (per-token USD strings) = input 4.21, output 12.63 per MTok.
+# Reasoning: always-on, no caller control — accepted params are only
+# temperature/max_tokens/top_p/top_k/stop/seed/frequency_penalty/presence_penalty/tools/
+# tool_choice/response_format — no reasoning_effort/thinking/enable_thinking/budget field;
+# Capabilities: temperature, top_p/top_k/stop/seed, tools, response_format json_object/json_schema,
+# multimodal input text+image+audio+video+pdf (base64/URL), text output, streaming + usage.
+# Limits: context 1_000_000 (context_length), output 65_536 (max_output_length) — same endpoint.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Medium"
+description = "Balanced multimodal model pairing a 1M-token context window with deeper reasoning for analysis, content creation, and tool use across text, image, audio, video, and PDF inputs."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 4.21
+output = 12.63
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/providers/vispark/models/vispark/vision-small.toml b/providers/vispark/models/vispark/vision-small.toml
new file mode 100644
index 00000000000..c93c050eb49
--- /dev/null
+++ b/providers/vispark/models/vispark/vision-small.toml
@@ -0,0 +1,36 @@
+# Vispark: Vision Small — first-party host (provider IS the lab, full inline allowed).
+# API model ID: vispark/vision-small — GET https://api.lab.vispark.in/v1/models (verified live 2026-09-13).
+# Cost source, USD/MTok, used verbatim: the provider's own models endpoint publishes
+# pricing.prompt = "0.00000105" and pricing.completion = "0.00000316" USD per token
+# (per-token USD strings) = input 1.05, output 3.16 per MTok.
+# Reasoning: always-on, no caller control — accepted params are only
+# temperature/max_tokens/top_p/top_k/stop/seed/frequency_penalty/presence_penalty/tools/
+# tool_choice/response_format — no reasoning_effort/thinking/enable_thinking/budget field;
+# Capabilities: temperature, top_p/top_k/stop/seed, tools, response_format json_object/json_schema,
+# multimodal input text+image+audio+video+pdf (base64/URL), text output, streaming + usage.
+# Limits: context 1_000_000 (context_length), output 65_536 (max_output_length) — same endpoint.
+# Release 2024-05-15: live endpoint returns created = 1715731200 (= 2024-05-15) per model,
+# Last updated 2026-09: current lineup as verified live 2026-09-13 (month precision).
+name = "Vision Small"
+description = "Fast, low-cost multimodal model for understanding text, images, audio, video, and PDFs, with tool calling and a 1M-token context window."
+release_date = "2024-05-15"
+last_updated = "2026-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 1.05
+output = 3.16
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
diff --git a/providers/vispark/provider.toml b/providers/vispark/provider.toml
new file mode 100644
index 00000000000..ac7e1634846
--- /dev/null
+++ b/providers/vispark/provider.toml
@@ -0,0 +1,11 @@
+# Vispark Lab — first-party host for Vispark Vision models.
+# API: https://api.lab.vispark.in/v1
+# Endpoints: GET /v1/models, POST /v1/chat/completions, POST /v1/completions
+# Auth: X-API-Key header or Authorization: Bearer <key> (keys like vl_...)
+# Platform: https://lab.vispark.in/ — get API key: https://lab.vispark.in/profile#api
+# Vision docs: https://lab.vispark.in/#vision — company: https://vispark.in
+name = "Vispark"
+npm = "@ai-sdk/openai-compatible"
+env = ["VISPARK_LAB_API_KEY"]
+api = "https://api.lab.vispark.in/v1"
+doc = "https://lab.vispark.in/#vision"

From b15ca12b8a137a6fa65ac457010d08167a94d63f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 02:35:25 +0000
Subject: [PATCH 1329/2012] chore(sync): update Deep Infra model catalog
 (#7110)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/zai-org/GLM-5.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/deepinfra/models/zai-org/GLM-5.3.toml b/providers/deepinfra/models/zai-org/GLM-5.3.toml
index b9e733f0b6f..f038885dbc0 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.3.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.3.toml
@@ -12,7 +12,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 1.2
 output = 4
-cache_read = 0.12
+cache_read = 0.2
 
 [limit]
 context = 1_048_576

From 12a1889c2f1f4c50b6c172809133e902d015688b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 02:35:32 +0000
Subject: [PATCH 1330/2012] chore(sync): update Kilo model catalog (#7109)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 556169b19ef..662428c45f6 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -17,7 +17,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.9
 output = 3
-cache_read = 0.09
+cache_read = 0.15
 
 [limit]
 context = 1_048_576

From 5eacd8811bd60c0309f999d9bfe943789e97d48f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 02:35:37 +0000
Subject: [PATCH 1331/2012] chore(sync): update OpenRouter model catalog
 (#7111)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 5c6d5ba351d..0a7b3acfd67 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -17,7 +17,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.9
 output = 3
-cache_read = 0.09
+cache_read = 0.15
 
 [limit]
 context = 1_310_720

From f55512687d17410f5a7bd9c6fcf989dcf7d9b504 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 03:31:17 +0000
Subject: [PATCH 1332/2012] chore(sync): update OpenRouter model catalog
 (#7113)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index df31c0bbb32..1273df92943 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.12
-cache_read = 0.012
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0a7b3acfd67..3e3ce108967 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 235_929
 
 [modalities]
 input = ["text"]

From 47d5282c3f61489d5d420b1bc960e27b48b9f6ce Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 03:31:30 +0000
Subject: [PATCH 1333/2012] chore(sync): update Kilo model catalog (#7114)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 662428c45f6..38aa9fed430 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
-context = 1_048_576
-output = 131_072
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 04d931bfeb644cdfdb1f4615bd6d7d5bb56feec6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 14 Sep 2026 23:23:06 -0500
Subject: [PATCH 1334/2012] feat(sync): add Fireworks AI catalog sync (#7112)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .github/workflows/sync-models.yml             |   1 +
 package.json                                  |   1 +
 packages/core/src/sync/index.ts               |   5 +-
 .../core/src/sync/providers/fireworks-ai.ts   | 268 ++++++++++++++++++
 packages/core/test/fireworks-ai-sync.test.ts  | 219 ++++++++++++++
 .../models/deepseek-v4-flash-0731.toml        |  10 +-
 .../models/deepseek-v4-pro-0813.toml          |  10 +-
 .../fireworks/models/deepseek-v4p1-flash.toml |  10 +-
 .../accounts/fireworks/models/glm-5p2.toml    |  16 +-
 .../fireworks/models/glm-5p3-flash.toml       |  15 +-
 .../accounts/fireworks/models/glm-5p3.toml    |  14 +-
 .../fireworks/models/gpt-oss-120b.toml        |   8 +-
 .../accounts/fireworks/models/kimi-k2p6.toml  |  16 +-
 .../fireworks/models/kimi-k2p7-code.toml      |  16 +-
 .../accounts/fireworks/models/kimi-k3.toml    |  14 +-
 .../accounts/fireworks/models/minimax-m3.toml |  20 +-
 .../models/nemotron-3-ultra-nvfp4.toml        |   6 +-
 .../fireworks/models/qwen3p8-max.toml         |  17 +-
 sync.md                                       |  12 +
 19 files changed, 614 insertions(+), 64 deletions(-)
 create mode 100644 packages/core/src/sync/providers/fireworks-ai.ts
 create mode 100644 packages/core/test/fireworks-ai-sync.test.ts

diff --git a/.github/workflows/sync-models.yml b/.github/workflows/sync-models.yml
index ac9f3690bcd..3ec03eb942d 100644
--- a/.github/workflows/sync-models.yml
+++ b/.github/workflows/sync-models.yml
@@ -77,6 +77,7 @@ jobs:
           DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
           DIGITALOCEAN_API_TOKEN: ${{ secrets.DIGITALOCEAN_API_TOKEN }}
           DIGITALOCEAN_ACCESS_TOKEN: ${{ secrets.DIGITALOCEAN_ACCESS_TOKEN }}
+          FIREWORKS_API_KEY: ${{ secrets.FIREWORKS_API_KEY }}
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
           OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
diff --git a/package.json b/package.json
index 4a7b90393f7..a1de7f5bb8e 100644
--- a/package.json
+++ b/package.json
@@ -38,6 +38,7 @@
     "vercel:generate": "bun ./packages/core/script/sync-models.ts vercel",
     "wandb:generate": "bun ./packages/core/script/sync-models.ts wandb",
     "digitalocean:sync": "bun ./packages/core/script/sync-models.ts digitalocean",
+    "fireworks:sync": "bun ./packages/core/script/sync-models.ts fireworks-ai",
     "ambient:sync": "bun ./packages/core/script/sync-models.ts ambient",
     "models:sync": "bun ./packages/core/script/sync-models.ts",
     "sync:models": "bun ./packages/core/script/sync-models.ts",
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 39d87a5817e..9156827c689 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -18,6 +18,7 @@ import { deepinfra } from "./providers/deepinfra.js";
 import { digitalocean } from "./providers/digitalocean.js";
 import { edenai } from "./providers/edenai.js";
 import { empiriolabs } from "./providers/empiriolabs.js";
+import { fireworksAi } from "./providers/fireworks-ai.js";
 import { friendli } from "./providers/friendli.js";
 import { githubCopilot } from "./providers/github-copilot.js";
 import { google } from "./providers/google.js";
@@ -150,6 +151,7 @@ export const providers: {
   digitalocean: SyncProvider<any>;
   edenai: SyncProvider<any>;
   empiriolabs: SyncProvider<any>;
+  "fireworks-ai": SyncProvider<any>;
   friendli: SyncProvider<any>;
   "github-copilot": SyncProvider<any>;
   google: SyncProvider<any>;
@@ -187,6 +189,7 @@ export const providers: {
   digitalocean,
   edenai,
   empiriolabs,
+  "fireworks-ai": fireworksAi,
   friendli,
   "github-copilot": githubCopilot,
   google,
@@ -231,7 +234,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "friendli", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "fireworks-ai", "friendli", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
diff --git a/packages/core/src/sync/providers/fireworks-ai.ts b/packages/core/src/sync/providers/fireworks-ai.ts
new file mode 100644
index 00000000000..aa1d366ca9e
--- /dev/null
+++ b/packages/core/src/sync/providers/fireworks-ai.ts
@@ -0,0 +1,268 @@
+import { z } from "zod";
+
+import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
+import { factorBaseModel } from "./openrouter.js";
+
+const API_ENDPOINT = "https://api.fireworks.ai/v1/serverless/models";
+
+const FireworksPrice = z.object({
+  sku: z.string().min(1),
+  amount: z.string().regex(/^\d+(?:\.\d+)?$/),
+  unit: z.literal("1M tokens"),
+}).passthrough();
+
+export const FireworksModel = z.object({
+  id: z.string().min(1),
+  object: z.literal("model"),
+  serverless_mode: z.string().min(1),
+  service_tier: z.string().min(1).optional(),
+  usage_identifier: z.string().min(1).optional(),
+  aliases: z.array(z.string().min(1)).optional(),
+  pricing: z.array(FireworksPrice),
+  display_name: z.string().min(1),
+  description: z.string(),
+  context_length: z.number().int().positive().optional(),
+  use_cases: z.array(z.string()).optional(),
+  input_modalities: z.array(z.string()),
+  output_modalities: z.array(z.string()),
+  created: z.number().int().nonnegative(),
+}).passthrough();
+
+export const FireworksResponse = z.object({
+  object: z.literal("list"),
+  data: z.array(FireworksModel),
+}).passthrough();
+
+export type FireworksModel = z.infer<typeof FireworksModel>;
+export type FireworksCatalogModel = FireworksModel & {
+  catalogId: string;
+  flagModes: FireworksModel[];
+};
+
+export const fireworksAi = {
+  id: "fireworks-ai",
+  name: "Fireworks AI",
+  modelsDir: "providers/fireworks-ai/models",
+  skipCreates: true,
+  // The endpoint describes the public serverless catalog, but it still lacks
+  // enough intrinsic metadata and reasoning controls to create safe entries.
+  deleteMissing: false,
+  sourceID(model) {
+    return supportsCatalogModel(model) ? model.catalogId : undefined;
+  },
+  skippedNotice(ids) {
+    if (ids.length === 0) return [];
+    return [
+      `${ids.length} Fireworks serverless text/vision IDs were not created because the endpoint does not yet provide output limits, reasoning controls, tool support, or open-weight status. Existing models are still updated from API-authoritative fields.`,
+      `Skipped remote IDs: ${ids.map((id) => `\`${id}\``).join(", ")}`,
+    ];
+  },
+  missingNotice(paths) {
+    if (paths.length === 0) return [];
+    return [
+      `${paths.length} local Fireworks models were absent from the serverless catalog and were retained for manual lifecycle review.`,
+      `Retained local paths: ${paths.map((item) => `\`${item}\``).join(", ")}`,
+    ];
+  },
+  async fetchModels() {
+    const key = process.env.FIREWORKS_API_KEY;
+    if (key === undefined) throw new Error("Fireworks AI sync requires FIREWORKS_API_KEY");
+    return fetchFireworksModels(key);
+  },
+  parseModels(raw) {
+    return expandFireworksModels(FireworksResponse.parse(raw).data);
+  },
+  translateModel(model, context) {
+    if (!supportsCatalogModel(model)) return undefined;
+    const existing = context.existing(model.catalogId);
+    if (existing === undefined) return undefined;
+    return {
+      id: model.catalogId,
+      model: buildFireworksModel(model, existing),
+    };
+  },
+} satisfies SyncProvider<FireworksCatalogModel>;
+
+export async function fetchFireworksModels(
+  key: string,
+  fetcher: typeof fetch = fetch,
+) {
+  const response = await fetcher(API_ENDPOINT, {
+    headers: { Authorization: `Bearer ${key}` },
+  });
+  if (!response.ok) {
+    throw new Error(`Fireworks AI models request failed: ${response.status} ${response.statusText}`);
+  }
+  return FireworksResponse.parse(await response.json());
+}
+
+export function expandFireworksModels(models: FireworksModel[]): FireworksCatalogModel[] {
+  const expanded = new Map<string, FireworksCatalogModel>();
+  const grouped = Map.groupBy(models, (model) => model.id);
+  for (const rows of grouped.values()) {
+    const defaultRow = rows.find((model) =>
+      model.usage_identifier === undefined && model.service_tier === undefined
+    );
+    const flagModes = rows.filter((model) => model.service_tier !== undefined);
+
+    // A default row owns the base model ID and exposes flag-based paths such as
+    // Priority as experimental modes. A priority-only model still needs to be
+    // discoverable, so its service-tier recipe becomes the base invocation.
+    const baseRow = defaultRow ?? flagModes[0];
+    if (baseRow !== undefined) add(baseRow.id, baseRow, defaultRow === undefined ? [] : flagModes);
+
+    for (const model of rows) {
+      if (model.usage_identifier !== undefined) add(model.usage_identifier, model, []);
+      for (const alias of model.aliases ?? []) add(alias, model, []);
+    }
+  }
+  return [...expanded.values()];
+
+  function add(catalogId: string, model: FireworksModel, flagModes: FireworksModel[]) {
+    if (!expanded.has(catalogId)) expanded.set(catalogId, { ...model, catalogId, flagModes });
+  }
+}
+
+function supportsCatalogModel(model: FireworksCatalogModel) {
+  return model.output_modalities.includes("text");
+}
+
+type Modality = SyncedFullModel["modalities"]["input"][number];
+
+const MODALITIES = new Set<Modality>(["text", "audio", "image", "video", "pdf"]);
+
+function catalogModalities(values: string[], fallback: Modality[]): Modality[] {
+  const modalities = values.filter((value): value is Modality => MODALITIES.has(value as Modality));
+  return modalities.length === 0 ? fallback : modalities;
+}
+
+function pricing(model: Pick<FireworksModel, "pricing">, existing: NonNullable<SyncedFullModel["cost"]>) {
+  const bySku = new Map(model.pricing.map((price) => [price.sku, Number(price.amount)]));
+  return {
+    ...existing,
+    input: bySku.get("LLM input tokens (uncached)") ?? existing.input,
+    cache_read: bySku.get("LLM input tokens (cached)") ?? existing.cache_read,
+    output: bySku.get("LLM output tokens") ?? existing.output,
+  };
+}
+
+function provider(
+  model: FireworksCatalogModel,
+  existing: ExistingModel["provider"],
+): ExistingModel["provider"] {
+  if (model.service_tier !== undefined) {
+    return {
+      ...existing,
+      body: {
+        ...existing?.body,
+        service_tier: model.service_tier,
+      },
+    };
+  }
+  if (existing === undefined) return undefined;
+
+  const body = { ...existing.body };
+  delete body.service_tier;
+  const result = { ...existing };
+  if (Object.keys(body).length === 0) delete result.body;
+  else result.body = body;
+  return Object.keys(result).length === 0 ? undefined : result;
+}
+
+function experimental(
+  model: FireworksCatalogModel,
+  cost: NonNullable<SyncedFullModel["cost"]>,
+  existing: ExistingModel["experimental"],
+): ExistingModel["experimental"] {
+  const modes = { ...existing?.modes };
+  // Priority is currently the only Fireworks flag-based serverless mode. The
+  // endpoint is authoritative for its availability as well as its pricing.
+  delete modes.priority;
+  for (const mode of model.flagModes) {
+    modes[mode.serverless_mode] = {
+      cost: pricing(mode, cost),
+      provider: { body: { service_tier: mode.service_tier! } },
+    };
+  }
+  if (Object.keys(modes).length === 0) return undefined;
+  return {
+    ...existing,
+    modes,
+  };
+}
+
+export function buildFireworksModel(
+  model: FireworksCatalogModel,
+  existing: ExistingModel,
+): SyncedModel {
+  const name = existing.name;
+  const description = existing.description;
+  const releaseDate = existing.release_date;
+  const lastUpdated = existing.last_updated;
+  const reasoning = existing.reasoning;
+  const toolCall = existing.tool_call;
+  const openWeights = existing.open_weights;
+  const limit = existing.limit;
+  const modalities = existing.modalities;
+  const cost = existing.cost;
+
+  if (
+    name === undefined
+    || description === undefined
+    || releaseDate === undefined
+    || lastUpdated === undefined
+    || reasoning === undefined
+    || toolCall === undefined
+    || openWeights === undefined
+    || limit === undefined
+    || limit.context === undefined
+    || limit.output === undefined
+    || modalities === undefined
+    || cost === undefined
+  ) {
+    throw new Error(`Fireworks AI model ${model.catalogId} has incomplete local TOML metadata required for sync`);
+  }
+
+  const input = catalogModalities(model.input_modalities, modalities.input);
+  const outputModalities = catalogModalities(model.output_modalities, modalities.output);
+  // Fireworks reports the advertised context window, while some deployments
+  // reserve a few prompt tokens. Preserve a smaller verified local cap, but
+  // immediately follow any lower ceiling reported by the API.
+  const context = model.context_length === undefined
+    ? limit.context
+    : Math.min(limit.context, model.context_length);
+  const output = Math.min(limit.output, context);
+  const values = {
+    name,
+    description,
+    family: existing.family,
+    release_date: releaseDate,
+    last_updated: lastUpdated,
+    attachment: input.some((modality) => modality !== "text"),
+    reasoning,
+    reasoning_options: existing.reasoning_options,
+    temperature: existing.temperature,
+    tool_call: toolCall,
+    structured_output: existing.structured_output,
+    knowledge: existing.knowledge,
+    open_weights: openWeights,
+    status: existing.status,
+    interleaved: existing.interleaved,
+    cost: pricing(model, cost),
+    limit: {
+      context,
+      input: limit.input,
+      output,
+    },
+    modalities: {
+      input,
+      output: outputModalities,
+    },
+    provider: provider(model, existing.provider),
+    experimental: experimental(model, cost, existing.experimental),
+  } satisfies SyncedFullModel;
+
+  return existing.base_model === undefined
+    ? values
+    : factorBaseModel(existing.base_model, values, values.limit, existing.base_model_omit);
+}
diff --git a/packages/core/test/fireworks-ai-sync.test.ts b/packages/core/test/fireworks-ai-sync.test.ts
new file mode 100644
index 00000000000..54c531336f4
--- /dev/null
+++ b/packages/core/test/fireworks-ai-sync.test.ts
@@ -0,0 +1,219 @@
+import { expect, test } from "bun:test";
+
+import type { ExistingModel } from "../src/sync/index.js";
+import {
+  buildFireworksModel,
+  expandFireworksModels,
+  fetchFireworksModels,
+  FireworksResponse,
+  fireworksAi,
+  type FireworksCatalogModel,
+  type FireworksModel,
+} from "../src/sync/providers/fireworks-ai.js";
+
+test("fetches the Fireworks serverless catalog with bearer auth", async () => {
+  let request: Request | undefined;
+  const fetcher = (async (input: string | URL | Request, init?: RequestInit) => {
+    request = input instanceof Request
+      ? new Request(input, init)
+      : new Request(input.toString(), init);
+    return Response.json({ object: "list", data: [fireworksModel()] });
+  }) as unknown as typeof fetch;
+
+  await fetchFireworksModels("test-key", fetcher);
+
+  expect(request?.url).toBe("https://api.fireworks.ai/v1/serverless/models");
+  expect(request?.headers.get("authorization")).toBe("Bearer test-key");
+});
+
+test("parses the Fireworks serverless model list", () => {
+  const parsed = FireworksResponse.parse({
+    object: "list",
+    data: [fireworksModel()],
+  });
+
+  expect(parsed.data[0]).toMatchObject({
+    id: "accounts/fireworks/models/example",
+    serverless_mode: "standard",
+    context_length: 1_048_576,
+    input_modalities: ["text", "image"],
+  });
+});
+
+test("expands usage identifiers and aliases and attaches flag-only modes", () => {
+  const models = expandFireworksModels([
+    fireworksModel({ aliases: ["accounts/fireworks/routers/example-latest"] }),
+    fireworksModel({
+      serverless_mode: "fast",
+      usage_identifier: "accounts/fireworks/routers/example-fast",
+      aliases: ["accounts/fireworks/routers/example-fast-latest"],
+    }),
+    fireworksModel({ serverless_mode: "priority", service_tier: "priority" }),
+  ]);
+
+  expect(models.map((model) => model.catalogId)).toEqual([
+    "accounts/fireworks/models/example",
+    "accounts/fireworks/routers/example-latest",
+    "accounts/fireworks/routers/example-fast",
+    "accounts/fireworks/routers/example-fast-latest",
+  ]);
+  expect(models[0]?.flagModes).toHaveLength(1);
+  expect(models[0]?.flagModes[0]).toMatchObject({
+    serverless_mode: "priority",
+    service_tier: "priority",
+  });
+});
+
+test("updates Fireworks pricing and modalities while preserving authored facts", () => {
+  const model = buildFireworksModel(
+    catalogModel(),
+    existingModel(),
+  );
+
+  expect(model).toMatchObject({
+    attachment: true,
+    tool_call: true,
+    cost: { input: 1.4, output: 4.4, cache_read: 0.14 },
+    limit: { context: 1_048_573, output: 262_144 },
+    modalities: { input: ["text", "image"], output: ["text"] },
+    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
+    experimental: {
+      modes: {
+        priority: {
+          cost: { input: 1.75, output: 5.5, cache_read: 0.175 },
+          provider: { body: { service_tier: "priority" } },
+        },
+      },
+    },
+  });
+});
+
+test("uses the service-tier recipe for a priority-only model", () => {
+  const priority = fireworksModel({ serverless_mode: "priority", service_tier: "priority" });
+  const [model] = expandFireworksModels([priority]);
+
+  expect(buildFireworksModel(model!, existingModel())).toMatchObject({
+    provider: { body: { service_tier: "priority" } },
+  });
+});
+
+test("clears a stale base service tier when the model returns to standard", () => {
+  const model = buildFireworksModel(
+    catalogModel({ service_tier: undefined }),
+    {
+      ...existingModel(),
+      provider: { body: { service_tier: "priority" } },
+    },
+  );
+
+  expect(model.provider).toBeUndefined();
+});
+
+test("removes a stale priority mode when Fireworks no longer lists it", () => {
+  const model = buildFireworksModel(
+    catalogModel({ flagModes: [] }),
+    {
+      ...existingModel(),
+      experimental: {
+        modes: {
+          priority: {
+            cost: { input: 2, output: 4 },
+            provider: { body: { service_tier: "priority" } },
+          },
+        },
+      },
+    },
+  );
+
+  expect(model.experimental).toBeUndefined();
+});
+
+test("uses serverless modalities as authoritative", () => {
+  const model = buildFireworksModel(
+    catalogModel({ input_modalities: ["text"] }),
+    {
+      ...existingModel(),
+      attachment: true,
+      modalities: { input: ["text", "image", "video"], output: ["text"] },
+    },
+  );
+
+  expect(model.modalities?.input).toEqual(["text"]);
+  expect(model.attachment).toBe(false);
+});
+
+test("uses Fireworks context length only as an upper bound", () => {
+  const model = buildFireworksModel(
+    catalogModel({ context_length: 131_072 }),
+    existingModel(),
+  );
+
+  expect(model.limit?.context).toBe(131_072);
+  expect(model.limit?.output).toBe(131_072);
+});
+
+test("does not report Fireworks embedding rows as missing generation models", () => {
+  const embedding = catalogModel({ output_modalities: ["embeddings"] });
+
+  expect(fireworksAi.sourceID(embedding)).toBeUndefined();
+  expect(fireworksAi.translateModel(embedding, {
+    existing: () => existingModel(),
+    authored: () => existingModel(),
+  })).toBeUndefined();
+});
+
+function fireworksModel(overrides: Partial<FireworksModel> = {}): FireworksModel {
+  return {
+    id: "accounts/fireworks/models/example",
+    object: "model",
+    serverless_mode: "standard",
+    pricing: [
+      { sku: "LLM input tokens (cached)", amount: "0.14", unit: "1M tokens" },
+      { sku: "LLM input tokens (uncached)", amount: "1.4", unit: "1M tokens" },
+      { sku: "LLM output tokens", amount: "4.4", unit: "1M tokens" },
+    ],
+    display_name: "Example",
+    description: "Example reasoning model",
+    context_length: 1_048_576,
+    input_modalities: ["text", "image"],
+    output_modalities: ["text"],
+    created: 1_788_566_400,
+    ...overrides,
+  };
+}
+
+function catalogModel(overrides: Partial<FireworksCatalogModel> = {}): FireworksCatalogModel {
+  const model = fireworksModel(overrides);
+  return {
+    ...model,
+    catalogId: overrides.catalogId ?? model.usage_identifier ?? model.id,
+    flagModes: overrides.flagModes ?? [fireworksModel({
+      serverless_mode: "priority",
+      service_tier: "priority",
+      pricing: [
+        { sku: "LLM input tokens (cached)", amount: "0.175", unit: "1M tokens" },
+        { sku: "LLM input tokens (uncached)", amount: "1.75", unit: "1M tokens" },
+        { sku: "LLM output tokens", amount: "5.5", unit: "1M tokens" },
+      ],
+    })],
+  };
+}
+
+function existingModel(): ExistingModel {
+  return {
+    name: "Example",
+    description: "Example reasoning model",
+    release_date: "2026-09-01",
+    last_updated: "2026-09-01",
+    attachment: false,
+    reasoning: true,
+    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
+    temperature: true,
+    tool_call: true,
+    structured_output: true,
+    open_weights: true,
+    cost: { input: 1, output: 2 },
+    limit: { context: 1_048_573, output: 262_144 },
+    modalities: { input: ["text"], output: ["text"] },
+  };
+}
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
index 51306765904..7bb811f6613 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
@@ -1,5 +1,8 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
@@ -7,10 +10,11 @@ type = "toggle"
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.22
 output = 0.66
 cache_read = 0.007
+
+[experimental.modes.priority]
+cost = { input = 0.275, output = 0.825, cache_read = 0.00875 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
index e98a50a2643..031dd2a013b 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
@@ -4,6 +4,9 @@
 # Pricing: https://fireworks.ai/models/deepseek-ai/deepseek-v4-pro-0813 (accessed 2026-08-14)
 base_model = "deepseek/deepseek-v4-pro-0813"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
@@ -11,10 +14,11 @@ type = "toggle"
 type = "effort"
 values = ["high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 1.32
 output = 3.96
 cache_read = 0.044
+
+[experimental.modes.priority]
+cost = { input = 1.65, output = 4.95, cache_read = 0.055 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml
index d6e8a5fa391..07a24dbd793 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4p1-flash.toml
@@ -9,6 +9,9 @@
 # Released on Fireworks Serverless API 2026-09-10.
 base_model = "deepseek/deepseek-v4.1-flash"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
@@ -16,10 +19,11 @@ type = "toggle"
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.22
 output = 0.66
 cache_read = 0.007
+
+[experimental.modes.priority]
+cost = { input = 0.275, output = 0.825, cache_read = 0.00875 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
index 74a93f7734b..bf70c694ba9 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
@@ -1,7 +1,7 @@
-name = "GLM 5.2"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 # Fireworks' published schema does not document this model's toggle or "max"
 # effort mapping. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
+name = "GLM 5.2"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 family = "glm"
 release_date = "2026-06-16"
 last_updated = "2026-06-16"
@@ -11,6 +11,9 @@ temperature = true
 tool_call = true
 open_weights = true
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
@@ -19,8 +22,8 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 1.40
-output = 4.40
+input = 1.4
+output = 4.4
 cache_read = 0.14
 
 [limit]
@@ -31,5 +34,6 @@ output = 131_072
 input = ["text"]
 output = ["text"]
 
-[interleaved]
-field = "reasoning_content"
+[experimental.modes.priority]
+cost = { input = 1.75, output = 5.5, cache_read = 0.175 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
index 06b9cb00646..2792714f8e4 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3-flash.toml
@@ -15,17 +15,24 @@ base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 last_updated = "2026-09-07"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.15
-output = 0.50
+output = 0.5
 cache_read = 0.03
 
 [limit]
 context = 1_048_573
+
+[modalities]
+input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 0.1875, output = 0.625, cache_read = 0.0375 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
index 6ba5c186726..c805f171742 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p3.toml
@@ -19,18 +19,22 @@ base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
 last_updated = "2026-09-07"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
-input = 1.40
-output = 4.40
+input = 1.4
+output = 4.4
 cache_read = 0.26
 
 [limit]
 context = 1_048_573
 output = 262_144
+
+[experimental.modes.priority]
+cost = { input = 1.75, output = 5.5, cache_read = 0.325 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-120b.toml b/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-120b.toml
index b62f6b4eb62..3b5586b6964 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-120b.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-120b.toml
@@ -1,7 +1,5 @@
 name = "GPT OSS 120B"
 description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-# $.reasoning_effort = "low" | "medium" | "high".
-# https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
 family = "gpt-oss"
 release_date = "2025-08-05"
 last_updated = "2026-06-16"
@@ -17,7 +15,7 @@ values = ["low", "medium", "high"]
 
 [cost]
 input = 0.15
-output = 0.60
+output = 0.6
 cache_read = 0.015
 
 [limit]
@@ -27,3 +25,7 @@ output = 32_768
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[experimental.modes.priority]
+cost = { input = 0.18, output = 0.72, cache_read = 0.018 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
index 290cb4c7819..cf312414120 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
@@ -1,7 +1,7 @@
-name = "Kimi K2.6"
-description = "Kimi reasoning model for long-horizon research, planning, and tool use"
 # Fireworks documents thinking.type = "enabled", but no explicit disabled
 # value for this model. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
+name = "Kimi K2.6"
+description = "Kimi reasoning model for long-horizon research, planning, and tool use"
 family = "kimi-thinking"
 release_date = "2026-04-17"
 last_updated = "2026-04-17"
@@ -11,13 +11,16 @@ temperature = true
 tool_call = true
 open_weights = true
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-cache_read = 0.16
 input = 0.95
-output = 4.00
+output = 4
+cache_read = 0.16
 
 [limit]
 context = 262_000
@@ -27,5 +30,6 @@ output = 262_000
 input = ["text", "image"]
 output = ["text"]
 
-[interleaved]
-field = "reasoning_content"
+[experimental.modes.priority]
+cost = { input = 1.5, output = 6, cache_read = 0.22 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
index 47df91e54ee..4db378e3e5e 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
@@ -1,7 +1,7 @@
-name = "Kimi K2.7 Code"
-description = "Kimi coding model for software agents, refactors, and repository reasoning"
 # Fireworks documents thinking.type = "enabled", but no explicit disabled
 # value for this model. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
+name = "Kimi K2.7 Code"
+description = "Kimi coding model for software agents, refactors, and repository reasoning"
 family = "kimi-k2"
 release_date = "2026-06-12"
 last_updated = "2026-06-16"
@@ -11,13 +11,16 @@ temperature = true
 tool_call = true
 open_weights = true
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
-cache_read = 0.19
 input = 0.95
-output = 4.00
+output = 4
+cache_read = 0.19
 
 [limit]
 context = 262_000
@@ -27,5 +30,6 @@ output = 262_000
 input = ["text", "image"]
 output = ["text"]
 
-[interleaved]
-field = "reasoning_content"
+[experimental.modes.priority]
+cost = { input = 1.425, output = 6, cache_read = 0.285 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k3.toml
index 6780b801c9c..065b801125b 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k3.toml
@@ -2,6 +2,9 @@ base_model = "moonshotai/kimi-k3"
 release_date = "2026-07-27"
 last_updated = "2026-07-27"
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "toggle"
 
@@ -13,13 +16,14 @@ values = ["low", "medium", "high", "max"]
 type = "budget_tokens"
 min = 1_024
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
-input = 3.0
-output = 15.0
+input = 3
+output = 15
 cache_read = 0.3
 
 [modalities]
 input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 3.75, output = 18.75, cache_read = 0.375 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
index 89f9d16f043..295f4a7d2cd 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
@@ -1,23 +1,26 @@
-name = "MiniMax-M3"
-description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
 # $.reasoning_effort = "low" | "medium" | "high".
 # https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
+name = "MiniMax-M3"
+description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
 family = "minimax"
 release_date = "2026-06-12"
 last_updated = "2026-06-12"
-attachment = true
+attachment = false
 reasoning = true
 temperature = true
 tool_call = true
 open_weights = true
 
+[interleaved]
+field = "reasoning_content"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.30
-output = 1.20
+input = 0.3
+output = 1.2
 cache_read = 0.06
 
 [limit]
@@ -25,8 +28,9 @@ context = 512_000
 output = 512_000
 
 [modalities]
-input = ["text", "image", "video"]
+input = ["text"]
 output = ["text"]
 
-[interleaved]
-field = "reasoning_content"
+[experimental.modes.priority]
+cost = { input = 0.45, output = 1.8, cache_read = 0.09 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/nemotron-3-ultra-nvfp4.toml b/providers/fireworks-ai/models/accounts/fireworks/models/nemotron-3-ultra-nvfp4.toml
index 169414e233e..baa9d9b890a 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/nemotron-3-ultra-nvfp4.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/nemotron-3-ultra-nvfp4.toml
@@ -4,9 +4,9 @@ base_model = "nvidia/nemotron-3-ultra-550b-a55b"
 type = "toggle"
 
 [cost]
-input = 0.60
-output = 2.40
-cache_read = 0.119
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-max.toml b/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-max.toml
index 61607c96136..ff82f0c6561 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-max.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/qwen3p8-max.toml
@@ -1,18 +1,19 @@
 base_model = "alibaba/qwen3.8-max"
 
-attachment = false
-
 [[reasoning_options]]
 type = "toggle"
 
-[modalities]
-input = ["text"]
-output = ["text"]
-
 [cost]
-input = 2.00
-output = 6.00
+input = 2
+output = 6
 cache_read = 0.25
 
 [limit]
 context = 262_144
+
+[modalities]
+input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 3, output = 9, cache_read = 0.375 }
+provider = { body = { service_tier = "priority" } }
diff --git a/sync.md b/sync.md
index 70a06dab55b..d2c8bc3520c 100644
--- a/sync.md
+++ b/sync.md
@@ -295,6 +295,18 @@ OVHcloud AI Endpoints is implemented in `packages/core/src/sync/providers/ovhclo
 - `attachment` is derived from non-text `input_modalities`, and `open_weights` from the presence of `hugging_face_id`.
 - `release_date`/`last_updated` default to the catalog `created` timestamp but preserve any existing hand-authored dates; `knowledge`, `family`, `status`, `interleaved`, and `limit.input` are preserved when present.
 
+## Fireworks AI Notes
+
+Fireworks AI is implemented in `packages/core/src/sync/providers/fireworks-ai.ts`.
+
+- Run it with `bun models:sync fireworks-ai` or `bun fireworks:sync`.
+- Source endpoint: `https://api.fireworks.ai/v1/serverless/models`; required auth: `FIREWORKS_API_KEY`.
+- The serverless catalog exposes one row per serving mode with live per-million-token pricing, invocation identifiers, aliases, advertised context length, and input/output modalities.
+- Standard rows update base-model IDs. Fast and other alternate-resource rows update the model or router named by `usage_identifier`; their aliases are tracked as additional invocation IDs. Flag-only modes such as Priority become priced `experimental.modes` with the required request-body `service_tier` instead of duplicate catalog IDs. A priority-only model remains discoverable with that service-tier recipe on its base ID.
+- New text/vision invocation IDs are reported but not created automatically because the endpoint does not yet provide output limits, reasoning controls, tool support, or open-weight status. Embedding/reranking rows are ignored because the catalog provider entries describe generation models.
+- Pricing and modalities come from the matching serverless serving mode. Existing exact context caps smaller than the advertised API value are preserved, while a lower API ceiling is applied. Authored output limits, reasoning options, tool support, and other metadata not exposed by the endpoint are preserved.
+- Models absent from the serverless response are retained for manual lifecycle review.
+
 ## DigitalOcean Notes
 
 - DigitalOcean is implemented in `packages/core/src/sync/providers/digitalocean.ts`.

From 48efabd43e197d01e92923a7b5c2576ba79de614 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 04:30:41 +0000
Subject: [PATCH 1335/2012] chore(sync): update NanoGPT model catalog (#7117)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From 40a74bba6c25226e43c048e20df03489e1fd48f5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 04:30:46 +0000
Subject: [PATCH 1336/2012] chore(sync): update Kilo model catalog (#7121)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index a43a707c98d..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From ab427172b7fdd6250353847910b8481bbda67660 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 04:31:18 +0000
Subject: [PATCH 1337/2012] chore(sync): update OpenRouter model catalog
 (#7119)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 93d4a9d7ab9..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 2407ebe687389d4b6e9f94abcb7f9123ca8cd307 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 05:27:30 +0000
Subject: [PATCH 1338/2012] chore(sync): update Eden AI model catalog (#7141)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/deepseek/deepseek-chat.toml    |  6 +++---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml  |  9 +++------
 .../edenai/models/deepseek/deepseek-v4-flash.toml      | 10 +++++++---
 providers/edenai/models/deepseek/deepseek-v4-pro.toml  |  6 +++---
 4 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/providers/edenai/models/deepseek/deepseek-chat.toml b/providers/edenai/models/deepseek/deepseek-chat.toml
index 68423502b1b..9daa8f3a3e5 100644
--- a/providers/edenai/models/deepseek/deepseek-chat.toml
+++ b/providers/edenai/models/deepseek/deepseek-chat.toml
@@ -3,9 +3,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.28
-output = 0.42
-cache_read = 0.028
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 131_072
diff --git a/providers/edenai/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/edenai/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 54c9aff2f3e..afcc9b8045b 100644
--- a/providers/edenai/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/edenai/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,6 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
-
-[limit]
-context = 1_048_576
+input = 0.15
+output = 0.6
+cache_read = 0.003
diff --git a/providers/edenai/models/deepseek/deepseek-v4-flash.toml b/providers/edenai/models/deepseek/deepseek-v4-flash.toml
index d8f7c84db4d..f3e849c7f29 100644
--- a/providers/edenai/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/edenai/models/deepseek/deepseek-v4-flash.toml
@@ -1,10 +1,14 @@
 base_model = "deepseek/deepseek-v4-flash"
+attachment = true
 
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.15
+output = 0.6
+cache_read = 0.003
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/edenai/models/deepseek/deepseek-v4-pro.toml b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
index 6b7c04704c9..043e1bb1f41 100644
--- a/providers/edenai/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
@@ -5,6 +5,6 @@ type = "effort"
 values = ["none", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022

From d65928d53650fa54c6da8ec9ce7480148df19ec9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 05:27:32 +0000
Subject: [PATCH 1339/2012] chore(sync): update OpenRouter model catalog
 (#7140)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..ec32a69e6bd 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From 63208b610286df04694413e35a41a723c9e5e7b7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 05:27:34 +0000
Subject: [PATCH 1340/2012] chore(sync): update Kilo model catalog (#7139)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f09092fb568 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From badd763fe264f9e6787dbb772a67531d865c6d6a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 06:41:45 +0000
Subject: [PATCH 1341/2012] chore(sync): update OpenRouter model catalog
 (#7144)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 1273df92943..2331765a04d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.01
+input = 0.055
+output = 0.11
+cache_read = 0.011
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..93d4a9d7ab9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 87bb10ff937c3041584342217051e3932e337d2b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 06:41:48 +0000
Subject: [PATCH 1342/2012] chore(sync): update Kilo model catalog (#7143)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..a43a707c98d 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 9efe466efb905b66ad5fd6d8cfd15e29a2afc2a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 06:41:53 +0000
Subject: [PATCH 1343/2012] chore(sync): update NanoGPT model catalog (#7145)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From 1e0701787abcf0c29511e25df4574499691f4db2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 08:34:36 +0000
Subject: [PATCH 1344/2012] chore(sync): update OpenRouter model catalog
 (#7152)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml  | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 2331765a04d..df31c0bbb32 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.055
-output = 0.11
-cache_read = 0.011
+input = 0.06
+output = 0.12
+cache_read = 0.012
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index ec32a69e6bd..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 491097cb133febbb3bff4679b9bfa3b61e7915b2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 08:34:59 +0000
Subject: [PATCH 1345/2012] chore(sync): update Kilo model catalog (#7151)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f09092fb568..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From c6d34526484bf96586fbce79ef848bff16a8984a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 09:32:58 +0000
Subject: [PATCH 1346/2012] chore(sync): update Kilo model catalog (#7155)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 5942ecf4ef8f63490a90bc7de05a4784f2b794ad Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 09:33:02 +0000
Subject: [PATCH 1347/2012] chore(sync): update OpenRouter model catalog
 (#7156)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 6433c74de0287a1e926137b6b10c2a4e6ad9e21d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 10:28:04 +0000
Subject: [PATCH 1348/2012] chore(sync): update OpenRouter model catalog
 (#7160)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 93d4a9d7ab9..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 30d70c201e37ba86b06143046e7d2b301e6bee07 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 10:28:06 +0000
Subject: [PATCH 1349/2012] chore(sync): update Kilo model catalog (#7161)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index a43a707c98d..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 177e19ebd2e9751f871630417eecde42e4bffec1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 10:28:10 +0000
Subject: [PATCH 1350/2012] chore(sync): update NanoGPT model catalog (#7159)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 providers/nano-gpt/models/google/gemini-flash-latest.toml   | 2 +-
 providers/nano-gpt/models/google/gemini-pro-latest.toml     | 2 +-
 5 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/google/gemini-flash-latest.toml b/providers/nano-gpt/models/google/gemini-flash-latest.toml
index 1060d3dd657..3e206c4240e 100644
--- a/providers/nano-gpt/models/google/gemini-flash-latest.toml
+++ b/providers/nano-gpt/models/google/gemini-flash-latest.toml
@@ -10,7 +10,7 @@ values = ["minimal", "low", "medium", "high"]
 input = 0.75
 output = 3.75
 cache_read = 0.075
-cache_write = 0.075
+cache_write = 0.041667
 
 [limit]
 input = 1_048_576
diff --git a/providers/nano-gpt/models/google/gemini-pro-latest.toml b/providers/nano-gpt/models/google/gemini-pro-latest.toml
index 7f53052baf1..8841cf18ebd 100644
--- a/providers/nano-gpt/models/google/gemini-pro-latest.toml
+++ b/providers/nano-gpt/models/google/gemini-pro-latest.toml
@@ -26,5 +26,5 @@ input = 1_048_576
 output = 65_536
 
 [modalities]
-input = ["text", "image", "audio"]
+input = ["text", "image", "video", "audio"]
 output = ["text"]

From 27688ca055d8de30f5e196f4629739f96737cfab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 11:25:31 +0000
Subject: [PATCH 1351/2012] chore(sync): update NanoGPT model catalog (#7162)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/deepclaude.toml | 23 -----------------------
 1 file changed, 23 deletions(-)
 delete mode 100644 providers/nano-gpt/models/deepclaude.toml

diff --git a/providers/nano-gpt/models/deepclaude.toml b/providers/nano-gpt/models/deepclaude.toml
deleted file mode 100644
index 6edc6988da6..00000000000
--- a/providers/nano-gpt/models/deepclaude.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "DeepClaude"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-release_date = "2025-03-10"
-last_updated = "2025-02-01"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[limit]
-context = 128_000
-input = 128_000
-output = 8_192
-
-[modalities]
-input = ["text", "pdf"]
-output = ["text"]

From 0e04d3787eaf1c50e40117bf798f2b4241370b5d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 12:36:51 +0000
Subject: [PATCH 1352/2012] chore(sync): update Charm Hyper model catalog
 (#7163)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.3.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/hyper/models/glm-5.3.toml b/providers/hyper/models/glm-5.3.toml
index 98852073c09..bc974016adc 100644
--- a/providers/hyper/models/glm-5.3.toml
+++ b/providers/hyper/models/glm-5.3.toml
@@ -1,7 +1,6 @@
 base_model = "zhipuai/glm-5.3"
 release_date = "2026-08-28"
 last_updated = "2026-08-31"
-attachment = true
 
 [[reasoning_options]]
 type = "effort"
@@ -15,6 +14,3 @@ cache_read = 0.283088
 [limit]
 context = 1_048_576
 output = 262_144
-
-[modalities]
-input = ["text", "image"]

From 0802cf3a8c41dfd8ede74ddc200e361545021db2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 13:28:02 +0000
Subject: [PATCH 1353/2012] chore(sync): update Kilo model catalog (#7167)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml   |  3 ++-
 providers/kilo/models/z-ai/glm-5.2:free.toml    | 17 +++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml   |  8 ++++----
 3 files changed, 23 insertions(+), 5 deletions(-)
 create mode 100644 providers/kilo/models/z-ai/glm-5.2:free.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/z-ai/glm-5.2:free.toml b/providers/kilo/models/z-ai/glm-5.2:free.toml
new file mode 100644
index 00000000000..7bd0bf98a27
--- /dev/null
+++ b/providers/kilo/models/z-ai/glm-5.2:free.toml
@@ -0,0 +1,17 @@
+base_model = "zhipuai/glm-5.2"
+name = "Z.ai: GLM 5.2 (free)"
+description = "GLM 5.2 is a large-scale reasoning model from Z.ai. It supports text input and output with a 1M-token context window, and is suited for long-horizon agent workflows, project-level software engineering,..."
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "xhigh"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 32_768
+output = 29_491
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..957bc0a82fa 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 7e1cb24fd483db68961022969df296d9e5848431 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 13:28:05 +0000
Subject: [PATCH 1354/2012] chore(sync): update OpenRouter model catalog
 (#7168)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  7 +++---
 .../openrouter/models/z-ai/glm-5.2:free.toml  | 22 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  8 +++----
 4 files changed, 33 insertions(+), 10 deletions(-)
 create mode 100644 providers/openrouter/models/z-ai/glm-5.2:free.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 451e7c3012c..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9834
-output = 2.9502
-cache_read = 0.03278
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/z-ai/glm-5.2:free.toml b/providers/openrouter/models/z-ai/glm-5.2:free.toml
new file mode 100644
index 00000000000..a48b6dabee1
--- /dev/null
+++ b/providers/openrouter/models/z-ai/glm-5.2:free.toml
@@ -0,0 +1,22 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2 (free)"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 32_768
+output = 29_491
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..50d7ce57993 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 9b510ddcd01b187aa1c1f3106208dcc327bce7ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 14:27:58 +0000
Subject: [PATCH 1355/2012] chore(sync): update OpenRouter model catalog
 (#7169)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml         | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index c54d27b9d18..d5918c075a2 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index dc34ba0ccf7..98fd23a037b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 50d7ce57993..e0b3ef90d2e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
 context = 1_048_576

From addf0179de0219511c1127278fcb775b1b9b29f0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 14:28:00 +0000
Subject: [PATCH 1356/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7171)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/mistral/devstral-2.toml     | 23 -------------------
 .../models/mistral/devstral-small-2.toml      | 23 -------------------
 .../vercel/models/mistral/ministral-14b.toml  |  3 ++-
 .../models/mistral/mistral-large-3.toml       |  7 +++---
 .../models/mistral/mistral-medium-3.5.toml    |  3 ++-
 .../vercel/models/mistral/mistral-medium.toml | 23 -------------------
 .../vercel/models/mistral/mistral-nemo.toml   | 10 ++++----
 .../vercel/models/mistral/mistral-small.toml  |  7 +++---
 .../vercel/models/mistral/pixtral-12b.toml    |  1 -
 9 files changed, 17 insertions(+), 83 deletions(-)
 delete mode 100644 providers/vercel/models/mistral/devstral-2.toml
 delete mode 100644 providers/vercel/models/mistral/devstral-small-2.toml
 delete mode 100644 providers/vercel/models/mistral/mistral-medium.toml
 delete mode 120000 providers/vercel/models/mistral/pixtral-12b.toml

diff --git a/providers/vercel/models/mistral/devstral-2.toml b/providers/vercel/models/mistral/devstral-2.toml
deleted file mode 100644
index 39dcd7ae233..00000000000
--- a/providers/vercel/models/mistral/devstral-2.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Devstral 2"
-description = "Mistral coding agent model for repository tasks and software engineering workflows"
-family = "devstral"
-release_date = "2025-12-09"
-last_updated = "2025-12-09"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = false
-
-[cost]
-input = 0.4
-output = 2
-
-[limit]
-context = 256_000
-output = 256_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/vercel/models/mistral/devstral-small-2.toml b/providers/vercel/models/mistral/devstral-small-2.toml
deleted file mode 100644
index ba8cd035573..00000000000
--- a/providers/vercel/models/mistral/devstral-small-2.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Devstral Small 2"
-description = "Mistral coding agent model for repository tasks and software engineering workflows"
-family = "devstral"
-release_date = "2025-12-09"
-last_updated = "2025-05-07"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2024-10"
-open_weights = false
-
-[cost]
-input = 0.1
-output = 0.3
-
-[limit]
-context = 256_000
-output = 256_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/mistral/ministral-14b.toml b/providers/vercel/models/mistral/ministral-14b.toml
index 0e0cab05c58..4d50bcd5a30 100644
--- a/providers/vercel/models/mistral/ministral-14b.toml
+++ b/providers/vercel/models/mistral/ministral-14b.toml
@@ -13,9 +13,10 @@ open_weights = false
 [cost]
 input = 0.2
 output = 0.2
+cache_read = 0.02
 
 [limit]
-context = 256_000
+context = 262_144
 output = 256_000
 
 [modalities]
diff --git a/providers/vercel/models/mistral/mistral-large-3.toml b/providers/vercel/models/mistral/mistral-large-3.toml
index 34f5793e606..ed49c090141 100644
--- a/providers/vercel/models/mistral/mistral-large-3.toml
+++ b/providers/vercel/models/mistral/mistral-large-3.toml
@@ -7,16 +7,17 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = false
-open_weights = false
 knowledge = "2024-10"
+open_weights = false
 
 [cost]
 input = 0.5
 output = 1.5
+cache_read = 0.05
 
 [limit]
-context = 256000
-output = 256000
+context = 262_144
+output = 256_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/vercel/models/mistral/mistral-medium-3.5.toml b/providers/vercel/models/mistral/mistral-medium-3.5.toml
index 71e72a2295f..df6e48742b3 100644
--- a/providers/vercel/models/mistral/mistral-medium-3.5.toml
+++ b/providers/vercel/models/mistral/mistral-medium-3.5.toml
@@ -16,9 +16,10 @@ values = ["none", "high"]
 [cost]
 input = 1.5
 output = 7.5
+cache_read = 0.15
 
 [limit]
-context = 256_000
+context = 262_144
 output = 256_000
 
 [modalities]
diff --git a/providers/vercel/models/mistral/mistral-medium.toml b/providers/vercel/models/mistral/mistral-medium.toml
deleted file mode 100644
index 55cd916da3d..00000000000
--- a/providers/vercel/models/mistral/mistral-medium.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Mistral Medium 3.1"
-description = "Mistral model for multilingual chat, reasoning, and tool-assisted workflows"
-family = "mistral-medium"
-release_date = "2025-05-07"
-last_updated = "2025-05-07"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = false
-knowledge = "2024-10"
-
-[cost]
-input = 0.4
-output = 2
-
-[limit]
-context = 128000
-output = 64000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/mistral/mistral-nemo.toml b/providers/vercel/models/mistral/mistral-nemo.toml
index d9f4b04fa40..d3fdcb8bd53 100644
--- a/providers/vercel/models/mistral/mistral-nemo.toml
+++ b/providers/vercel/models/mistral/mistral-nemo.toml
@@ -11,13 +11,13 @@ knowledge = "2024-04"
 open_weights = false
 
 [cost]
-input = 0.15
-output = 0.15
+input = 0.04
+output = 0.17
 
 [limit]
-context = 128_000
-output = 128_000
+context = 60_288
+output = 16_000
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
 output = ["text"]
diff --git a/providers/vercel/models/mistral/mistral-small.toml b/providers/vercel/models/mistral/mistral-small.toml
index 6dfa15d7429..588d0d8c5fb 100644
--- a/providers/vercel/models/mistral/mistral-small.toml
+++ b/providers/vercel/models/mistral/mistral-small.toml
@@ -11,11 +11,12 @@ knowledge = "2025-06"
 open_weights = true
 
 [cost]
-input = 0.1
-output = 0.3
+input = 0.15
+output = 0.6
+cache_read = 0.015
 
 [limit]
-context = 32_000
+context = 262_144
 output = 4_000
 
 [modalities]
diff --git a/providers/vercel/models/mistral/pixtral-12b.toml b/providers/vercel/models/mistral/pixtral-12b.toml
deleted file mode 120000
index c1f9fff3648..00000000000
--- a/providers/vercel/models/mistral/pixtral-12b.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../../mistral/models/pixtral-12b.toml
\ No newline at end of file

From 784d02302b91f0391dc3a0ebb16b3536a308fdf3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 14:28:03 +0000
Subject: [PATCH 1357/2012] chore(sync): update Kilo model catalog (#7172)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml      | 4 ++--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 673b730eb39..f491c2853c8 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_000_000
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 957bc0a82fa..bd4537b7471 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
-context = 1_048_576
+context = 1_000_000
 output = 393_216
 
 [modalities]

From 90d6d247960c287b8aae17a46abd2de4ea39dd62 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 14:28:10 +0000
Subject: [PATCH 1358/2012] chore(sync): update Charm Hyper model catalog
 (#7170)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gemma-4-26b-a4b-it.toml              | 6 +++---
 providers/hyper/models/glm-5.1.toml                         | 4 ++--
 providers/hyper/models/glm-5.toml                           | 6 +++---
 providers/hyper/models/kimi-k2.5.toml                       | 6 +++---
 .../models/llama-4-maverick-17b-128e-instruct-fp8.toml      | 6 +++---
 providers/hyper/models/minimax-m2.7.toml                    | 6 +++---
 6 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/providers/hyper/models/gemma-4-26b-a4b-it.toml b/providers/hyper/models/gemma-4-26b-a4b-it.toml
index 2ef2f0a55c3..dd4a8093488 100644
--- a/providers/hyper/models/gemma-4-26b-a4b-it.toml
+++ b/providers/hyper/models/gemma-4-26b-a4b-it.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.1
-output = 0.374
-cache_read = 0.05
+input = 0.098
+output = 0.334
+cache_read = 0.049
 
 [limit]
 context = 256_000
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
index e102da3b5ad..ea8b1d2c372 100644
--- a/providers/hyper/models/glm-5.1.toml
+++ b/providers/hyper/models/glm-5.1.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 1.318
+input = 1.358
 output = 4.268
-cache_read = 0.659
+cache_read = 0.679
 
 [limit]
 context = 202_750
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index e92beec4670..9882a80de94 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.94
-output = 3.008
-cache_read = 0.47
+input = 0.86
+output = 2.752
+cache_read = 0.43
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
index 5a60cf903a0..59545971994 100644
--- a/providers/hyper/models/kimi-k2.5.toml
+++ b/providers/hyper/models/kimi-k2.5.toml
@@ -5,9 +5,9 @@ attachment = false
 reasoning_options = []
 
 [cost]
-input = 0.5344
-output = 2.815
-cache_read = 0.2672
+input = 0.5584
+output = 2.935
+cache_read = 0.2792
 
 [limit]
 output = 26_214
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
index 5a9ed7085e5..b8610bed8df 100644
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 attachment = false
 
 [cost]
-input = 0.255
-output = 0.8365
-cache_read = 0.1275
+input = 0.25188
+output = 0.82563
+cache_read = 0.12594
 
 [limit]
 context = 430_000
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 13b1f1bd0ed..701182a4ee1 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.462
-output = 1.728
-cache_read = 0.231
+input = 0.418
+output = 1.588
+cache_read = 0.209
 
 [limit]
 context = 262_100

From aefb23178489c993bf8ad0f0f51c1ae2c7dfe8d0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 15:25:56 +0000
Subject: [PATCH 1359/2012] chore(sync): update Kilo model catalog (#7175)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index bd4537b7471..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 393_216
 
 [modalities]

From f12d56792880ef367b9f57f490e956432e423d2d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 15:25:59 +0000
Subject: [PATCH 1360/2012] chore(sync): update OpenRouter model catalog
 (#7174)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e0b3ef90d2e..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From b448c34932725f0c98df63911f2556123982ee1a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 15:26:01 +0000
Subject: [PATCH 1361/2012] chore(sync): update Eden AI model catalog (#7176)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index 6917caeb537..c540731d87a 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34653
-output = 0.46204
+input = 0.34617
+output = 0.46156
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index cb91c32767e..d11cc1ac90a 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.750815
-output = 0.750815
+input = 0.750035
+output = 0.750035
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 06f34e160e0..26c9489ce11 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173265
-output = 0.750815
+input = 0.173085
+output = 0.750035
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 4b4d3fe1c8b..c7d9a5ffc2c 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46204
-output = 0.92408
+input = 0.46156
+output = 0.92312
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index e18f7060b40..42cb87b432b 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173265
-output = 0.69306
+input = 0.173085
+output = 0.69234
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 4b4a9a94743..f2cdbc67c92 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.03959
-output = 1.03959
+input = 1.03851
+output = 1.03851

From 91a8677cbeab4407f1e48247e4ddb479abaaa3cc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 16:28:27 +0000
Subject: [PATCH 1362/2012] chore(sync): update Kilo model catalog (#7177)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From a955cd19fe9869d2b26c6447e59c1521d4ab3623 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 16:28:31 +0000
Subject: [PATCH 1363/2012] chore(sync): update OpenRouter model catalog
 (#7178)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From ebe4c82b60b0fbc84989ccd5592e2e88734a610f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?plaf=C3=BC?= <plafue@users.noreply.github.com>
Date: Tue, 15 Sep 2026 19:01:46 +0200
Subject: [PATCH 1364/2012] Add deepseek-v4.1-flash model for TensorX provider
 (#7166)

---
 .../tensorx/models/deepseek/deepseek-v4.1-flash.toml | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 providers/tensorx/models/deepseek/deepseek-v4.1-flash.toml

diff --git a/providers/tensorx/models/deepseek/deepseek-v4.1-flash.toml b/providers/tensorx/models/deepseek/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..97d6ea672e1
--- /dev/null
+++ b/providers/tensorx/models/deepseek/deepseek-v4.1-flash.toml
@@ -0,0 +1,12 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle" # API: {"chat_template_kwargs": {"thinking": true}} (default off)
+
+[cost]
+input = 0.5
+output = 1.5
+cache_read = 0.13
+
+[limit]
+context = 1048576

From 443f4e12389fdfb930002d3af2688520cfc0ed07 Mon Sep 17 00:00:00 2001
From: henrikdukefoss <76407537+henrikdukefoss@users.noreply.github.com>
Date: Tue, 15 Sep 2026 19:01:57 +0200
Subject: [PATCH 1365/2012] fix(standardcompute): use OpenRouter SDK (#7157)

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
---
 providers/standardcompute/provider.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/standardcompute/provider.toml b/providers/standardcompute/provider.toml
index 2a9d36e18d7..47e78cf63e0 100644
--- a/providers/standardcompute/provider.toml
+++ b/providers/standardcompute/provider.toml
@@ -15,6 +15,6 @@
 # `reasoning_details` (streaming included).
 name = "Standard Compute"
 env = ["STANDARDCOMPUTE_API_KEY"]
-npm = "@ai-sdk/openai-compatible"
+npm = "@openrouter/ai-sdk-provider"
 api = "https://api.stdcmpt.com/v1"
 doc = "https://standardcompute.com/models"

From 9739db9c77d19515d830625999b631e36322297e Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 12:02:16 -0500
Subject: [PATCH 1366/2012] fix: Add deepseek-v4.1-flash to the international
 Alibaba Token Plan provider (#7147)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4.1-flash.toml           | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml

diff --git a/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml b/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..f32c843d2e8
--- /dev/null
+++ b/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,22 @@
+# Reasoning HTTP format (accessed 2026-09-15):
+# Toggle: enable_thinking = true|false. Chat Completions `reasoning_effort`
+# accepts low, high (default), and max; medium and xhigh map to high.
+# Reasoning streams on `reasoning_content`.
+# Model availability: Token Plan Personal and Team editions (Singapore).
+# Sources:
+# https://docs.qwencloud.com/api-reference/chat/openai-chat
+# https://www.alibabacloud.com/help/en/model-studio/deepseek-api
+# https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-overview
+# https://www.alibabacloud.com/help/en/model-studio/token-plan-team-overview
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 7cb01c5e99e1b5a4e8c16f9522c609d35682aaf0 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 12:05:00 -0500
Subject: [PATCH 1367/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/models/deepseek-v4-pro (#7129)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/models/deepseek-v4-pro.toml     | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..ca93dd0ae06
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
@@ -0,0 +1,20 @@
+# Fireworks lists DeepSeek-V4-Pro (preview) as on-demand only — Serverless is
+# "Not supported", so there is no per-token serverless price to record.
+# Model page: https://fireworks.ai/models/fireworks/deepseek-v4-pro (accessed 2026-09-15)
+# Serverless catalog omits this id (only DeepSeek V4 Pro 0813 is priced):
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-15)
+# Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
+# Effort: reasoning_effort = high|max; low/medium promote to high, xhigh to max
+# Reasoning API: https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-09-15)
+# Peer surface: providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
+base_model = "deepseek/deepseek-v4-pro"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]

From 973aefe19c582a03160f092b2883c9611f49a3bb Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 12:10:25 -0500
Subject: [PATCH 1368/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/minimax-latest (#7138)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/minimax-latest.toml     | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml
new file mode 100644
index 00000000000..f9d41fb264a
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml
@@ -0,0 +1,42 @@
+# Version-tracking router for the latest MiniMax model on Fireworks Serverless.
+# Currently resolves to accounts/fireworks/models/minimax-m3 (MiniMax M3):
+# - https://docs.fireworks.ai/ecosystem/fireconnect/models (minimax-latest)
+# - FireConnect ROUTER_SPEC_ALIASES: "minimax-latest" -> "minimax-m3"
+#   https://github.com/fw-ai/fireconnect/blob/0ff8a6c5/packages/setup-cli/lib/fireworks/model-specs.mjs
+# Pricing matches MiniMax M3 standard serverless (input / cached input / output):
+# $0.30 / $0.06 / $1.20; Priority $0.45 / $0.09 / $1.80
+# - https://docs.fireworks.ai/serverless/pricing
+# - https://fireworks.ai/models/fireworks/minimax-m3
+# Context/output ceilings and text-only modalities match the pinned minimax-m3
+# sibling on this host (Fireworks serverless context 512k).
+# Effort: $.reasoning_effort = "low" | "medium" | "high" (same surface as
+# accounts/fireworks/models/minimax-m3).
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "minimax/MiniMax-M3"
+name = "MiniMax Latest"
+attachment = false
+release_date = "2026-06-12"
+last_updated = "2026-09-15"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+context = 512_000
+output = 512_000
+
+[modalities]
+input = ["text"]
+
+[experimental.modes.priority]
+cost = { input = 0.45, output = 1.8, cache_read = 0.09 }
+provider = { body = { service_tier = "priority" } }

From 56669785604e0da6bab792b1d49e0b626f14e5c5 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:09 +0000
Subject: [PATCH 1369/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/kimi-latest (#7130)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/kimi-latest.toml        | 41 +++++++++++++++++++
 1 file changed, 41 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-latest.toml
new file mode 100644
index 00000000000..53861f148c0
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-latest.toml
@@ -0,0 +1,41 @@
+# Version-tracking Standard-path router for the current Kimi flagship
+# (currently Kimi K3). Short ID `kimi-latest` expands to
+# accounts/fireworks/routers/kimi-latest.
+# https://docs.fireworks.ai/ecosystem/fireconnect/models
+# Pricing matches Kimi K3 Standard / Priority (USD per 1M tokens):
+# Standard $3.00 / $0.30 / $15.00, Priority $3.75 / $0.375 / $18.75
+# https://docs.fireworks.ai/serverless/pricing
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|medium|high|max
+# Budget: thinking.budget_tokens (min 1024; conflicts with reasoning_effort)
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "moonshotai/kimi-k3"
+name = "Kimi Latest"
+release_date = "2026-07-27"
+last_updated = "2026-09-15"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1_024
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+
+[modalities]
+input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 3.75, output = 18.75, cache_read = 0.375 }
+provider = { body = { service_tier = "priority" } }

From 059375644142466189505aa6ff6908f09244fb8a Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:13 +0000
Subject: [PATCH 1370/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/deepseek-flash-latest (#7131)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../routers/deepseek-flash-latest.toml        | 36 +++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-flash-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-flash-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-flash-latest.toml
new file mode 100644
index 00000000000..a0062c45fe7
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-flash-latest.toml
@@ -0,0 +1,36 @@
+# Version-tracking standard-path router for current DeepSeek Flash
+# (tracks the live Flash SKU; currently DeepSeek V4.1 Flash).
+# Short ID: deepseek-flash-latest
+# Full model: accounts/fireworks/routers/deepseek-flash-latest
+# FireConnect: https://docs.fireworks.ai/ecosystem/fireconnect/models
+#   (latest vs fast; pinned deepseek-v4-flash deprecated → deepseek-flash-latest)
+# Underlying serverless model: accounts/fireworks/models/deepseek-v4p1-flash
+# Pricing (Standard / Priority, input / cached / output per 1M):
+#   https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-15)
+#   https://fireworks.ai/models/deepseek-ai/deepseek-v4p1-flash (accessed 2026-09-15)
+#   Standard $0.22 / $0.007 / $0.66; Priority $0.275 / $0.00875 / $0.825
+# Toggle: thinking.type = enabled|disabled (reasoning_effort = "none" or false also disables)
+# Effort: reasoning_effort = low|medium|high|xhigh|max
+#   Same surface as accounts/fireworks/models/deepseek-v4p1-flash (verified peer 2026-09-10).
+# Reasoning API: https://docs.fireworks.ai/api-reference/post-chatcompletions
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek Flash Latest"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.007
+
+[experimental.modes.priority]
+cost = { input = 0.275, output = 0.825, cache_read = 0.00875 }
+provider = { body = { service_tier = "priority" } }

From 53d288d302975c655203b5385e29c57ea3ccd204 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:17 +0000
Subject: [PATCH 1371/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/kimi-fast-latest (#7132)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/kimi-fast-latest.toml   | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-fast-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-fast-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-fast-latest.toml
new file mode 100644
index 00000000000..97505dbfb22
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-fast-latest.toml
@@ -0,0 +1,39 @@
+# Version-tracking Fast serving-path router for the current Kimi flagship.
+# Same quality as kimi-latest / pinned kimi-k3; Fast path aims for 100+ tok/s
+# at a higher $/token. Currently resolves to Kimi K3 Fast.
+# Docs: https://docs.fireworks.ai/ecosystem/fireconnect/models
+# Serving paths: https://docs.fireworks.ai/serverless/serving-paths
+# Pricing (Kimi K3 Fast row, accessed 2026-09-15):
+# https://docs.fireworks.ai/serverless/pricing — $4.50 / $0.45 / $22.50
+# per 1M (input / cached input / output). Pinned peer:
+# accounts/fireworks/routers/kimi-k3-fast.
+# Toggle: thinking.type = enabled|disabled (or thinking object)
+# Effort: reasoning_effort = low|medium|high|max
+# Budget: thinking.budget_tokens (N >= 1024); conflicts with reasoning_effort
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "moonshotai/kimi-k3"
+name = "Kimi Fast Latest"
+release_date = "2026-07-27"
+last_updated = "2026-09-15"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1_024
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 4.5
+output = 22.5
+cache_read = 0.45
+
+[modalities]
+input = ["text", "image"]

From a56824e91663a01ecc5cac8cffbd816f5c16b319 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:25 +0000
Subject: [PATCH 1372/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/models/minimax-m2p7 (#7133)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/models/minimax-m2p7.toml        | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
new file mode 100644
index 00000000000..2a241ff93f0
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
@@ -0,0 +1,22 @@
+# https://fireworks.ai/models/fireworks/minimax-m2p7 (accessed 2026-09-15)
+# Serverless: $0.30 / $0.059 / $1.20 per 1M (input / cached input / output).
+# Context Length 196k on Fireworks (lab base is 204_800); HF MiniMaxAI/MiniMax-M2.7.
+# Not listed on docs.fireworks.ai/serverless/pricing headline table (no Priority tier published).
+# $.reasoning_effort = "low" | "medium" | "high" (same Fireworks surface as MiniMax M3).
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "minimax/MiniMax-M2.7"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.059
+
+[limit]
+context = 196_608

From a395cd6e36d900e07a1cd881116b1b561c1c29f5 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:30 +0000
Subject: [PATCH 1373/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/glm-latest (#7134)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/glm-latest.toml         | 37 +++++++++++++++++++
 1 file changed, 37 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-latest.toml
new file mode 100644
index 00000000000..ff486ff3f73
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-latest.toml
@@ -0,0 +1,37 @@
+# Version-tracking standard-path router for the current flagship GLM.
+# FireConnect documents glm-latest as a Latest (standard path) alias that
+# moves to the current model automatically:
+# https://docs.fireworks.ai/ecosystem/fireconnect/models
+# Current target is GLM 5.3 (accounts/fireworks/models/glm-5p3). FireConnect
+# main maps ROUTER_SPEC_ALIASES["glm-latest"] = "glm-5p3" and resolves
+# candidates ["glm-5p3", "glm-5p2"] preferring the first present in catalog:
+# https://github.com/fw-ai/fireconnect/blob/main/packages/setup-cli/lib/fireworks/model-specs.mjs
+# Pricing matches GLM 5.3 Standard on the serverless pricing page (accessed
+# 2026-09-15): $1.40 / $0.26 / $4.40 per 1M (input / cached input / output).
+# Priority tier: $1.75 / $0.325 / $5.50.
+# https://docs.fireworks.ai/serverless/pricing
+# Limits, reasoning, and interleaved field match the sibling
+# accounts/fireworks/models/glm-5p3 entry (thinking-only; effort low/high/max).
+base_model = "zhipuai/glm-5.3"
+name = "GLM Latest"
+last_updated = "2026-09-15"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_573
+output = 262_144
+
+[experimental.modes.priority]
+cost = { input = 1.75, output = 5.5, cache_read = 0.325 }
+provider = { body = { service_tier = "priority" } }

From ba437a6c8b190ce346c1cf1cce700ac42f91f56c Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:36 +0000
Subject: [PATCH 1374/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/glm-fast-latest (#7135)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/glm-fast-latest.toml    | 38 +++++++++++++++++++
 1 file changed, 38 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-fast-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-fast-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-fast-latest.toml
new file mode 100644
index 00000000000..f7d8ed7d2f5
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-fast-latest.toml
@@ -0,0 +1,38 @@
+# Version-tracking Fast serving-path alias for the current flagship GLM Fast
+# router. Same weights/quality as the pinned Fast path; higher $/token than
+# Standard; aims for 100+ tok/s (Fireworks serving paths).
+#
+# Current target (2026-09-15): GLM 5.3 Fast
+# (accounts/fireworks/routers/glm-5p3-fast). FireConnect prefers
+# glm-5p3-fast over glm-5p2-fast when the newer Fast router is in the
+# serverless catalog:
+#   https://github.com/fw-ai/fireconnect/blob/main/packages/setup-cli/lib/fireworks/model-specs.mjs
+#   (GLM_FAST_LATEST_BASE_CANDIDATES = ["glm-5p3-fast", "glm-5p2-fast"])
+# Docs:
+#   https://docs.fireworks.ai/ecosystem/fireconnect/models
+#     (glm-fast-latest: Fast text-only agent loops, 1M context)
+#   https://docs.fireworks.ai/serverless/serving-paths
+#   https://docs.fireworks.ai/serverless/pricing
+#     (GLM 5.3 Fast: $2.10 / $0.39 / $6.60 per 1M input / cached / output)
+# Peer limits/reasoning match authored glm-5p3-fast (thinking-only effort
+# low|high|max; context 1_048_572; output 262_144).
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 Fast (Latest)"
+release_date = "2026-08-28"
+last_updated = "2026-09-15"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.10
+output = 6.60
+cache_read = 0.39
+
+[limit]
+context = 1_048_572
+output = 262_144

From f266e00c4f6f98ff14dba3e3823614dc000fd1c1 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:40 +0000
Subject: [PATCH 1375/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/glm-flash-latest (#7136)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../fireworks/routers/glm-flash-latest.toml   | 36 +++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml
new file mode 100644
index 00000000000..9e47aa19fea
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml
@@ -0,0 +1,36 @@
+# Version-tracking router alias for the current GLM Flash model on Fireworks
+# Serverless (currently GLM 5.3 Flash / accounts/fireworks/models/glm-5p3-flash).
+# Short ID glm-flash-latest; canonical path accounts/fireworks/routers/glm-flash-latest.
+# Sources (accessed 2026-09-15):
+# - Alias target: FireConnect ROUTER_SPEC_ALIASES maps glm-flash-latest → glm-5p3-flash
+#   https://github.com/fw-ai/fireconnect/blob/main/packages/setup-cli/lib/fireworks/model-specs.mjs
+# - FireConnect models doc: glm-flash-latest is vision-capable GLM for image tasks
+#   https://docs.fireworks.ai/ecosystem/fireconnect/models
+# - Pricing / base model page (GLM 5.3 Flash Standard): $0.15 / $0.03 / $0.50 per 1M
+#   (input / cached input / output)
+#   https://docs.fireworks.ai/serverless/pricing
+#   https://fireworks.ai/models/fireworks/glm-5p3-flash
+# Limits, reasoning controls, and modalities match the sibling pinned
+# accounts/fireworks/models/glm-5p3-flash entry (thinking-only effort low/high/max;
+# Fireworks text+image serverless surface).
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM Flash Latest (GLM 5.3 Flash)"
+last_updated = "2026-09-15"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_573
+
+[modalities]
+input = ["text", "image"]

From 47157880bbc4f0dbb1cfde886345d0bebed76b85 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:11:45 +0000
Subject: [PATCH 1376/2012] fix: [missing-model] fireworks-ai:
 accounts/fireworks/routers/deepseek-pro-latest (#7137)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../routers/deepseek-pro-latest.toml          | 34 +++++++++++++++++++
 1 file changed, 34 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-pro-latest.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-pro-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-pro-latest.toml
new file mode 100644
index 00000000000..d15bb087698
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-pro-latest.toml
@@ -0,0 +1,34 @@
+# Version-tracking "latest" router for DeepSeek Pro on Fireworks Serverless
+# (standard path — not Fast). FireConnect documents the short ID
+# deepseek-pro-latest and expands it to accounts/fireworks/routers/deepseek-pro-latest.
+# Static + catalog resolution currently targets deepseek-v4-pro-0813:
+# https://docs.fireworks.ai/ecosystem/fireconnect/models
+# https://github.com/fw-ai/fireconnect/blob/main/packages/setup-cli/lib/fireworks/model-specs.mjs
+# (ROUTER_SPEC_ALIASES + DEEPSEEK_PRO_LATEST_BASE_CANDIDATES; accessed 2026-09-15)
+# Pricing matches DeepSeek V4 Pro (0813) standard serverless rates:
+# https://docs.fireworks.ai/serverless/pricing
+# https://fireworks.ai/models/deepseek-ai/deepseek-v4-pro-0813
+# Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
+# Effort: reasoning_effort = high|max; low/medium promote to high, xhigh to max
+# Reasoning: https://docs.fireworks.ai/api-reference/post-chatcompletions
+base_model = "deepseek/deepseek-v4-pro-0813"
+name = "DeepSeek Pro Latest"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 1.32
+output = 3.96
+cache_read = 0.044
+
+[experimental.modes.priority]
+cost = { input = 1.65, output = 4.95, cache_read = 0.055 }
+provider = { body = { service_tier = "priority" } }

From 10520c43e8bc5d2e28124897e6329153b8196227 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:23:30 +0000
Subject: [PATCH 1377/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7180)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-1-flash.toml           | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/llmgateway/models/deepseek-v4-1-flash.toml

diff --git a/providers/llmgateway/models/deepseek-v4-1-flash.toml b/providers/llmgateway/models/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..9b9286b5771
--- /dev/null
+++ b/providers/llmgateway/models/deepseek-v4-1-flash.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek-V4.1-Flash"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+family = "deepseek"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.01
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text"]
+output = ["text"]

From b52f1fddbeb5f35c742b35bb5adeb8a26798e834 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:23:41 +0000
Subject: [PATCH 1378/2012] chore(sync): update LLM Gateway model catalog
 (#7181)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-v4.1-flash.toml                  |  5 +++-
 .../models/runware/deepseek-v4-1-flash.toml   | 28 +++++++++++++++++++
 2 files changed, 32 insertions(+), 1 deletion(-)
 create mode 100644 providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
index d946756de4a..ad9191cc116 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
@@ -1,10 +1,13 @@
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4.1 Flash (Consensus Protocol)"
 attachment = false
-reasoning = false
 tool_call = false
 structured_output = false
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
 [cost]
 input = 0.2
 output = 0.6
diff --git a/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml b/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..1d71ee3bb77
--- /dev/null
+++ b/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml
@@ -0,0 +1,28 @@
+name = "DeepSeek-V4.1-Flash (Runware)"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+family = "deepseek"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.01
+
+[limit]
+context = 1_048_576
+output = 1_048_576
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 596c04f19824e4dc197f21c59b16b73c22756fb0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 17:23:44 +0000
Subject: [PATCH 1379/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7182)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../gemini-3.8-live-extended-thinking.toml    | 21 +++++++++++++++++++
 .../vercel/models/google/gemini-3.8-live.toml | 21 +++++++++++++++++++
 2 files changed, 42 insertions(+)
 create mode 100644 providers/vercel/models/google/gemini-3.8-live-extended-thinking.toml
 create mode 100644 providers/vercel/models/google/gemini-3.8-live.toml

diff --git a/providers/vercel/models/google/gemini-3.8-live-extended-thinking.toml b/providers/vercel/models/google/gemini-3.8-live-extended-thinking.toml
new file mode 100644
index 00000000000..e87d4aa6da0
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.8-live-extended-thinking.toml
@@ -0,0 +1,21 @@
+name = "Gemini 3.8 Live Extended Thinking"
+description = "Speech generation model for controllable voice, narration, and audio delivery"
+family = "gemini"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.75
+output = 4.5
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "audio"]
+output = ["text", "audio"]
diff --git a/providers/vercel/models/google/gemini-3.8-live.toml b/providers/vercel/models/google/gemini-3.8-live.toml
new file mode 100644
index 00000000000..2e5224a264a
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.8-live.toml
@@ -0,0 +1,21 @@
+name = "Gemini 3.8 Live"
+description = "Speech generation model for controllable voice, narration, and audio delivery"
+family = "gemini"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.75
+output = 4.5
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "audio"]
+output = ["text", "audio"]

From 2faff287964571bb43d8a24009d58b4c6d0c3113 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 18:28:58 +0000
Subject: [PATCH 1380/2012] chore(sync): update Charm Hyper model catalog
 (#7185)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/hyper/models/glm-5.3.toml b/providers/hyper/models/glm-5.3.toml
index bc974016adc..bdd46e1420e 100644
--- a/providers/hyper/models/glm-5.3.toml
+++ b/providers/hyper/models/glm-5.3.toml
@@ -13,4 +13,4 @@ cache_read = 0.283088
 
 [limit]
 context = 1_048_576
-output = 262_144
+output = 128_000

From 3fdd5fab6a4ee09e583dc47f217c1cf92b153022 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 18:29:03 +0000
Subject: [PATCH 1381/2012] chore(sync): update Tinfoil model catalog (#7186)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../tinfoil/models/deepseek-v4-flash.toml     | 19 -------------------
 1 file changed, 19 deletions(-)
 delete mode 100644 providers/tinfoil/models/deepseek-v4-flash.toml

diff --git a/providers/tinfoil/models/deepseek-v4-flash.toml b/providers/tinfoil/models/deepseek-v4-flash.toml
deleted file mode 100644
index b2b050f3391..00000000000
--- a/providers/tinfoil/models/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-# Sources (accessed 2026-08-26):
-# - https://inference.tinfoil.sh/v1/models
-# - https://docs.tinfoil.sh/models/chat.md (weights deepseek-ai/DeepSeek-V4-Flash-0731)
-# - https://docs.tinfoil.sh/guides/reasoning.md
-# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max on Tinfoil.
-# Off is reasoning_effort=none; no separate toggle.
-base_model = "deepseek/deepseek-v4-flash-0731"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.3
-output = 0.7
-cache_read = 0.06
-
-[limit]
-context = 1_048_576

From 4f785c1e3a44a436a12f1530cc127d1592d99227 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 18:29:09 +0000
Subject: [PATCH 1382/2012] chore(sync): update OpenRouter model catalog
 (#7184)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576

From 3aa938856bc7aa121fc9b4c1ede1dadf65816ba4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 19:23:26 +0000
Subject: [PATCH 1383/2012] chore(sync): update OpenRouter model catalog
 (#7188)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576

From fd516b3de9f8bf97285474ed53b8039afb4ea84a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 19:23:35 +0000
Subject: [PATCH 1384/2012] chore(sync): update Charm Hyper model catalog
 (#7187)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4.1-flash.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/hyper/models/deepseek-v4.1-flash.toml b/providers/hyper/models/deepseek-v4.1-flash.toml
index c67cd095316..e9180c12434 100644
--- a/providers/hyper/models/deepseek-v4.1-flash.toml
+++ b/providers/hyper/models/deepseek-v4.1-flash.toml
@@ -10,4 +10,4 @@ output = 1.30656
 cache_read = 0.032664
 
 [limit]
-output = 13_107
+output = 32_768

From 57b814ad51e0ddbc3e7aacd132f2e3b5ce6fb04a Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Tue, 15 Sep 2026 14:51:19 -0500
Subject: [PATCH 1385/2012] fix(amazon-bedrock): correct Anthropic catalog
 (#6790)

* fix(amazon-bedrock): correct Anthropic catalog

* fix(amazon-bedrock): clarify legacy Claude pricing

* fix(anthropic): preserve dated Opus snapshot metadata

* fix(amazon-bedrock): defer retired Haiku profiles

---------

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 26 ++++----------
 ...thropic.claude-opus-4-1-20250805-v1:0.toml |  6 +++-
 ...thropic.claude-opus-4-5-20251101-v1:0.toml | 27 ++++----------
 .../models/anthropic.claude-opus-4-6-v1.toml  | 17 ++++++---
 .../models/anthropic.claude-opus-4-7.toml     | 26 ++++----------
 .../models/anthropic.claude-opus-4-8.toml     |  6 +++-
 .../models/anthropic.claude-opus-5.toml       |  6 +++-
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 26 ++++----------
 .../models/anthropic.claude-sonnet-4-6.toml   | 21 ++++++++---
 .../models/anthropic.claude-sonnet-5.toml     |  8 ++++-
 ...thropic.claude-sonnet-4-20250514-v1:0.toml | 21 +++++++++++
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 16 ++++++---
 .../au.anthropic.claude-opus-4-6-v1.toml      | 35 +++++++-----------
 .../models/au.anthropic.claude-opus-4-7.toml  | 11 +++---
 .../models/au.anthropic.claude-opus-4-8.toml  | 16 ++++++---
 .../models/au.anthropic.claude-opus-5.toml    | 16 ++++++---
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 18 ++++++----
 .../au.anthropic.claude-sonnet-4-6.toml       | 24 +++++--------
 .../models/au.anthropic.claude-sonnet-5.toml  | 18 +++++++---
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 25 ++++---------
 ...thropic.claude-opus-4-5-20251101-v1:0.toml | 26 ++++----------
 .../eu.anthropic.claude-opus-4-6-v1.toml      |  6 +++-
 .../models/eu.anthropic.claude-opus-4-7.toml  | 25 ++++---------
 .../models/eu.anthropic.claude-opus-4-8.toml  |  6 +++-
 .../models/eu.anthropic.claude-opus-5.toml    |  6 +++-
 ...thropic.claude-sonnet-4-20250514-v1:0.toml | 21 +++++++++++
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 25 ++++---------
 .../eu.anthropic.claude-sonnet-4-6.toml       | 10 +++++-
 .../models/eu.anthropic.claude-sonnet-5.toml  |  8 ++++-
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 25 ++++---------
 ...thropic.claude-opus-4-5-20251101-v1:0.toml | 26 ++++----------
 .../global.anthropic.claude-opus-4-6-v1.toml  |  6 +++-
 .../global.anthropic.claude-opus-4-7.toml     | 25 ++++---------
 .../global.anthropic.claude-opus-4-8.toml     |  6 +++-
 .../global.anthropic.claude-opus-5.toml       |  6 +++-
 ...thropic.claude-sonnet-4-20250514-v1:0.toml | 20 +++++++++++
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 25 ++++---------
 .../global.anthropic.claude-sonnet-4-6.toml   | 10 +++++-
 .../global.anthropic.claude-sonnet-5.toml     |  8 ++++-
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 16 ++++++---
 .../models/jp.anthropic.claude-opus-4-7.toml  | 16 ++++++---
 .../models/jp.anthropic.claude-opus-4-8.toml  | 16 ++++++---
 .../models/jp.anthropic.claude-opus-5.toml    | 15 +++++---
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 18 ++++++----
 .../jp.anthropic.claude-sonnet-4-6.toml       | 20 ++++++++---
 .../models/jp.anthropic.claude-sonnet-5.toml  | 17 ++++++---
 .../models/us.anthropic.claude-fable-5.toml   | 11 +++---
 ...hropic.claude-haiku-4-5-20251001-v1:0.toml | 35 +++++++-----------
 ...thropic.claude-opus-4-1-20250805-v1:0.toml |  8 ++++-
 ...thropic.claude-opus-4-5-20251101-v1:0.toml | 36 +++++++------------
 .../us.anthropic.claude-opus-4-6-v1.toml      | 16 ++++++---
 .../models/us.anthropic.claude-opus-4-7.toml  | 35 +++++++-----------
 .../models/us.anthropic.claude-opus-4-8.toml  | 16 ++++++---
 .../models/us.anthropic.claude-opus-5.toml    | 16 ++++++---
 ...thropic.claude-sonnet-4-20250514-v1:0.toml | 21 +++++++++++
 ...ropic.claude-sonnet-4-5-20250929-v1:0.toml | 35 +++++++-----------
 .../us.anthropic.claude-sonnet-4-6.toml       | 20 ++++++++---
 .../models/us.anthropic.claude-sonnet-5.toml  | 18 +++++++---
 58 files changed, 561 insertions(+), 483 deletions(-)
 create mode 100644 providers/amazon-bedrock/models/apac.anthropic.claude-sonnet-4-20250514-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-20250514-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-20250514-v1:0.toml
 create mode 100644 providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-20250514-v1:0.toml

diff --git a/providers/amazon-bedrock/models/anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/anthropic.claude-haiku-4-5-20251001-v1:0.toml
index bce63a8a894..12ca63865e1 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,27 +1,13 @@
-name = "Claude Haiku 4.5"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2025-10-15"
-last_updated = "2025-10-15"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-haiku-4-5-20251001"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-02-28"
-open_weights = false
 
 [cost]
 input = 1.00
 output = 5.00
 cache_read = 0.10
 cache_write = 1.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-4-1-20250805-v1:0.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-4-1-20250805-v1:0.toml
index 89ead55375f..372bc75261d 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-4-1-20250805-v1:0.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-4-1-20250805-v1:0.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-1.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
 base_model = "anthropic/claude-opus-4-1-20250805"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 status = "deprecated"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-4-5-20251101-v1:0.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-4-5-20251101-v1:0.toml
index b2b3b8936e3..dfd31d0c75e 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-4-5-20251101-v1:0.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-4-5-20251101-v1:0.toml
@@ -1,27 +1,14 @@
-name = "Claude Opus 4.5"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-11-24"
-last_updated = "2025-08-01"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-5.html
+# https://platform.claude.com/docs/en/models/opus-4-5/overview
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Effort/budget: additionalModelRequestFields.output_config.effort / thinking.budget_tokens (Converse); output_config.effort / thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-opus-4-5-20251101"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-03-31"
-open_weights = false
 
 [cost]
 input = 5.00
 output = 25.00
 cache_read = 0.50
 cache_write = 6.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-4-6-v1.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-4-6-v1.toml
index 8d8f04e574d..4f732a8f7ac 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-4-6-v1.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-4-6-v1.toml
@@ -1,9 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-6.html
+# Live (2026-09-09): bare ID succeeds in eu-west-2; us-east-1 rejects it as profile-only.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/eu-west-2/index.json
+# Runtime in-region availability: eu-west-2; regional on-demand USD/MTok, 5-minute cache writes.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-4-7.toml
index 7efb04ef356..c1b7e0d345b 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-4-7.toml
@@ -1,26 +1,12 @@
-name = "Claude Opus 4.7"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-04-16"
-last_updated = "2026-04-16"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-temperature = false
-tool_call = true
-knowledge = "2026-01-31"
-open_weights = false
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-4-7"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5.00
 output = 25.00
 cache_read = 0.50
 cache_write = 6.25
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-4-8.toml
index 281926139d8..a3812d1bb20 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-4-8.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-5.toml
index d39c1cca5a6..0051dd2351e 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-5.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5
diff --git a/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index 31e3db99d51..ded65bfb0ac 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,27 +1,13 @@
-name = "Claude Sonnet 4.5"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-09-29"
-last_updated = "2025-09-29"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-07-31"
-open_weights = false
 
 [cost]
 input = 3.00
 output = 15.00
 cache_read = 0.30
 cache_write = 3.75
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-6.toml
index b45c15d979a..65803c0fd9e 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-sonnet-4-6.toml
@@ -1,9 +1,20 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# Live (2026-09-09): bare ID succeeds in eu-west-2; us-east-1 rejects it as profile-only.
+# Live eu-west-2 Converse (2026-09-09) accepts maxTokens=128000; Bedrock output override.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/eu-west-2/index.json
+# Runtime in-region availability: eu-west-2; regional on-demand USD/MTok, 5-minute cache writes.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
+
+[limit]
+output = 128_000
diff --git a/providers/amazon-bedrock/models/anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/anthropic.claude-sonnet-5.toml
index 2cf922c12c0..50b43c95c2a 100644
--- a/providers/amazon-bedrock/models/anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/anthropic.claude-sonnet-5.toml
@@ -1,6 +1,12 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-structured_output = true
+structured_output = false
 
 [cost]
 input = 2
diff --git a/providers/amazon-bedrock/models/apac.anthropic.claude-sonnet-4-20250514-v1:0.toml b/providers/amazon-bedrock/models/apac.anthropic.claude-sonnet-4-20250514-v1:0.toml
new file mode 100644
index 00000000000..2a6fded47f3
--- /dev/null
+++ b/providers/amazon-bedrock/models/apac.anthropic.claude-sonnet-4-20250514-v1:0.toml
@@ -0,0 +1,21 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4.html
+# https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListInferenceProfiles.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# Exact legacy Standard dimensions remain 1.0x ($3/$15), unlike newer Claude geo SKUs.
+# Profile verified from ap-northeast-1, ap-southeast-1, and ap-southeast-2; set AWS_REGION to a supported source.
+# Uses the native Bedrock provider's bedrock-runtime Converse/InvokeModel surface.
+# On-demand USD/MTok from ap-northeast-1; 5-minute cache writes. Legacy, EOL October 14, 2026.
+# Live US-family probe, 2026-09-09: Legacy access denied (404) without active use in the last 30 days.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-20250514"
+name = "Claude Sonnet 4 (APAC)"
+status = "deprecated"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/au.anthropic.claude-haiku-4-5-20251001-v1:0.toml
index 6e4d8fcacef..03c205019d3 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,10 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
 base_model = "anthropic/claude-haiku-4-5-20251001"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Haiku 4.5 (AU)"
 structured_output = true
 
 [cost]
-input = 1
-output = 5
-cache_read = 0.1
-cache_write = 1.25
+input = 1.10
+output = 5.50
+cache_read = 0.11
+cache_write = 1.375
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-6-v1.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-6-v1.toml
index 6e8a1028470..8d3e36cbb89 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-6-v1.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-6-v1.toml
@@ -1,27 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-6.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
+base_model = "anthropic/claude-opus-4-6"
 name = "AU Anthropic Claude Opus 4.6"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-02-05"
-last_updated = "2026-02-05"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-05"
-open_weights = false
 
 [cost]
-input = 16.50
-output = 82.50
-cache_read = 1.65
-cache_write = 20.625
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
index 6f693e67bde..840fa8d887d 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-7.toml
@@ -1,8 +1,11 @@
-# AU cross-Region inference profile, verified via ListInferenceProfiles from ap-southeast-2.
-# AU geo pricing (Price List ap-southeast-2 standard): input/output/cache-read/cache-write USD per 1M tokens.
-# Note: sibling au.* entries are priced at 1.0x; the Price List bills AU-geo calls at the 1.1x standard rate.
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-7"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.7 (AU)"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-8.toml
index b6fa9726b02..4fd408bba1d 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-4-8.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.8 (AU)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-5.toml
index 38706b90db2..4d56662ce47 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-5.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 5 (AU)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index a7d968dea03..e61e67a9216 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,10 +1,16 @@
-base_model = "anthropic/claude-sonnet-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.5 (AU)"
 structured_output = true
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-6.toml
index 5ff9a6de33c..f5c45b3a6ee 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-4-6.toml
@@ -1,16 +1,13 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live ap-southeast-2 Converse, 2026-09-09: maxTokens=128000 succeeds; 128001 is rejected with model limit 128000.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
+base_model = "anthropic/claude-sonnet-4-6"
 name = "AU Anthropic Claude Sonnet 4.6"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2026-02-17"
-last_updated = "2026-02-17"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-08"
-open_weights = false
 
 [cost]
 input = 3.30
@@ -19,9 +16,4 @@ cache_read = 0.33
 cache_write = 4.125
 
 [limit]
-context = 1_000_000
 output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-5.toml
index 9981d78e7f7..e73016bc1af 100644
--- a/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-sonnet-5.toml
@@ -1,10 +1,18 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-southeast-2/index.json
+# AU source ap-southeast-2: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Sonnet 5 (AU)"
-structured_output = true
+structured_output = false
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
+input = 2.20
+output = 11
+cache_read = 0.22
+cache_write = 2.75
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-haiku-4-5-20251001-v1:0.toml
index 406040a8176..8e2dd57e7a0 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,27 +1,14 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-haiku-4-5-20251001"
 name = "Claude Haiku 4.5 (EU)"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2025-10-15"
-last_updated = "2025-10-15"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-02-28"
-open_weights = false
 
 [cost]
 input = 1.10
 output = 5.50
 cache_read = 0.11
 cache_write = 1.375
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-5-20251101-v1:0.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-5-20251101-v1:0.toml
index 3690b2fb3e7..8a6f9b6e945 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-5-20251101-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-5-20251101-v1:0.toml
@@ -1,27 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-5.html
+# https://platform.claude.com/docs/en/models/opus-4-5/overview
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Effort/budget: additionalModelRequestFields.output_config.effort / thinking.budget_tokens (Converse); output_config.effort / thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-opus-4-5-20251101"
 name = "Claude Opus 4.5 (EU)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-11-24"
-last_updated = "2025-08-01"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-03-31"
-open_weights = false
 
 [cost]
 input = 5.50
 output = 27.50
 cache_read = 0.55
 cache_write = 6.875
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-6-v1.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-6-v1.toml
index e546467818d..13b62307847 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-6-v1.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-6-v1.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-6.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Opus 4.6 (EU)"
 structured_output = true
 
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-7.toml
index 3f5e8613af8..abb2e658a07 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-7.toml
@@ -1,26 +1,13 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-4-7"
 name = "Claude Opus 4.7 (EU)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-04-16"
-last_updated = "2026-04-16"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-temperature = false
-tool_call = true
-knowledge = "2026-01-31"
-open_weights = false
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5.50
 output = 27.50
 cache_read = 0.55
 cache_write = 6.875
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-8.toml
index 148eb6abef2..7a5df9c39c0 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-8.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.8 (EU)"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5.toml
index fc39e7a818f..eb26d63d65b 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 5 (EU)"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-20250514-v1:0.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-20250514-v1:0.toml
new file mode 100644
index 00000000000..59cbbb4b369
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-20250514-v1:0.toml
@@ -0,0 +1,21 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4.html
+# https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListInferenceProfiles.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/eu-west-1/index.json
+# Exact legacy Standard dimensions remain 1.0x ($3/$15), unlike newer Claude geo SKUs.
+# Profile verified from eu-west-1; set AWS_REGION to a supported source, such as eu-west-1 or eu-central-1.
+# Uses the native Bedrock provider's bedrock-runtime Converse/InvokeModel surface.
+# On-demand USD/MTok from eu-west-1; 5-minute cache writes. Legacy, EOL October 14, 2026.
+# Live US-family probe, 2026-09-09: Legacy access denied (404) without active use in the last 30 days.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-20250514"
+name = "Claude Sonnet 4 (EU)"
+status = "deprecated"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index 516bb5407d3..1d940d0cea0 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,27 +1,14 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
 name = "Claude Sonnet 4.5 (EU)"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-09-29"
-last_updated = "2025-09-29"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-07-31"
-open_weights = false
 
 [cost]
 input = 3.30
 output = 16.50
 cache_read = 0.33
 cache_write = 4.125
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-6.toml
index e3073490f10..d3b7de6a40b 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-6.toml
@@ -1,5 +1,10 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# Live eu-west-1 Converse (2026-09-09) accepts maxTokens=128000; Bedrock output override.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.6 (EU)"
 structured_output = true
 
@@ -8,3 +13,6 @@ input = 3.3
 output = 16.5
 cache_read = 0.33
 cache_write = 4.125
+
+[limit]
+output = 128_000
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-5.toml
index d854eebc78d..fd7029779c5 100644
--- a/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-5.toml
@@ -1,7 +1,13 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Sonnet 5 (EU)"
-structured_output = true
+structured_output = false
 
 [cost]
 input = 2.2
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/global.anthropic.claude-haiku-4-5-20251001-v1:0.toml
index 16f48ddcce4..daeb3fdbe73 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,27 +1,14 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-haiku-4-5-20251001"
 name = "Claude Haiku 4.5 (Global)"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2025-10-15"
-last_updated = "2025-10-15"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-02-28"
-open_weights = false
 
 [cost]
 input = 1.00
 output = 5.00
 cache_read = 0.10
 cache_write = 1.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-5-20251101-v1:0.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-5-20251101-v1:0.toml
index b30635e142c..2ae0b78fe56 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-5-20251101-v1:0.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-5-20251101-v1:0.toml
@@ -1,27 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-5.html
+# https://platform.claude.com/docs/en/models/opus-4-5/overview
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Effort/budget: additionalModelRequestFields.output_config.effort / thinking.budget_tokens (Converse); output_config.effort / thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-opus-4-5-20251101"
 name = "Claude Opus 4.5 (Global)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-11-24"
-last_updated = "2025-08-01"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-03-31"
-open_weights = false
 
 [cost]
 input = 5.00
 output = 25.00
 cache_read = 0.50
 cache_write = 6.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-6-v1.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-6-v1.toml
index 7dfcb74c73f..f6a0bced8d1 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-6-v1.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-6-v1.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-6.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Opus 4.6 (Global)"
 structured_output = true
 
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-7.toml
index 2a8e102806d..593db1ad85c 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-7.toml
@@ -1,26 +1,13 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-4-7"
 name = "Claude Opus 4.7 (Global)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-04-16"
-last_updated = "2026-04-16"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-temperature = false
-tool_call = true
-knowledge = "2026-01-31"
-open_weights = false
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5.00
 output = 25.00
 cache_read = 0.50
 cache_write = 6.25
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-8.toml
index 153b7ebe8fc..a561ed7c5eb 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-4-8.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.8 (Global)"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-5.toml
index 0efb900f316..87242170f4a 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-5.toml
@@ -1,5 +1,9 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 5 (Global)"
 
 [cost]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-20250514-v1:0.toml b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-20250514-v1:0.toml
new file mode 100644
index 00000000000..c8373367e7b
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-20250514-v1:0.toml
@@ -0,0 +1,20 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4.html
+# https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListInferenceProfiles.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# Supported source AWS_REGION values: us-east-1, us-east-2, us-west-2, eu-west-1, ap-northeast-1.
+# Profile verified from us-east-1, eu-west-1, and ap-northeast-1; native bedrock-runtime Converse/InvokeModel surface.
+# Global on-demand USD/MTok from us-east-1; 5-minute cache writes. Legacy, EOL October 14, 2026.
+# Live US-family probe, 2026-09-09: Legacy access denied (404) without active use in the last 30 days.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-20250514"
+name = "Claude Sonnet 4 (Global)"
+status = "deprecated"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index 32ddbb933f0..69ed7512da7 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,27 +1,14 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
 name = "Claude Sonnet 4.5 (Global)"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-09-29"
-last_updated = "2025-09-29"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-07-31"
-open_weights = false
 
 [cost]
 input = 3.00
 output = 15.00
 cache_read = 0.30
 cache_write = 3.75
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-6.toml
index 7b9186bd75e..2c9c143472b 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-6.toml
@@ -1,5 +1,10 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# Live us-east-1 Converse (2026-09-09) accepts maxTokens=128000; Bedrock output override.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.6 (Global)"
 structured_output = true
 
@@ -8,3 +13,6 @@ input = 3
 output = 15
 cache_read = 0.3
 cache_write = 3.75
+
+[limit]
+output = 128_000
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-5.toml
index f6be5729b4f..2c26e3d1413 100644
--- a/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-sonnet-5.toml
@@ -1,7 +1,13 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Sonnet 5 (Global)"
-structured_output = true
+structured_output = false
 
 [cost]
 input = 2
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-haiku-4-5-20251001-v1:0.toml
index 28a924e5a59..66bbb15f3c6 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,10 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# JP source ap-northeast-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
 base_model = "anthropic/claude-haiku-4-5-20251001"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Haiku 4.5 (JP)"
 structured_output = true
 
 [cost]
-input = 1
-output = 5
-cache_read = 0.1
-cache_write = 1.25
+input = 1.10
+output = 5.50
+cache_read = 0.11
+cache_write = 1.375
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-7.toml
index 658c4398650..b0970c1f6fd 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-7.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# JP source ap-northeast-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-7"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.7 (JP)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-8.toml
index dca27a26015..86851ef7341 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-4-8.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# JP source ap-northeast-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.8 (JP)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5.toml
index f83fb2052f4..55a91488eb1 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5.toml
@@ -1,9 +1,14 @@
+# Model controls: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# Pricing: Tokyo Standard dimensions, https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 5 (JP)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index 84bd54d57ce..4256653910e 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,10 +1,16 @@
-base_model = "anthropic/claude-sonnet-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# JP source ap-northeast-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.5 (JP)"
 structured_output = true
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-6.toml
index fb3ac1ec4a3..37a8894f9e0 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-4-6.toml
@@ -1,10 +1,20 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# Live ap-northeast-1 Converse (2026-09-09) accepts maxTokens=128000; Bedrock output override.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# JP source ap-northeast-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.6 (JP)"
 structured_output = true
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
+
+[limit]
+output = 128_000
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-5.toml
index 9da9ba0a8f5..78b1466e25f 100644
--- a/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-sonnet-5.toml
@@ -1,10 +1,17 @@
+# Model controls: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# Pricing: Tokyo Standard dimensions, https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/ap-northeast-1/index.json
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Sonnet 5 (JP)"
-structured_output = true
+structured_output = false
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
+input = 2.2
+output = 11
+cache_read = 0.22
+cache_write = 2.75
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-fable-5.toml b/providers/amazon-bedrock/models/us.anthropic.claude-fable-5.toml
index cd65098e8fa..2a06181f5d9 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-fable-5.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-fable-5.toml
@@ -1,9 +1,12 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-fable-5.html
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: standard regional USD/MTok; cache_write is the 5-minute rate.
 base_model = "anthropic/claude-fable-5"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Fable 5 (US)"
 
 [cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
+input = 11
+output = 55
+cache_read = 1.1
+cache_write = 13.75
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-haiku-4-5-20251001-v1:0.toml b/providers/amazon-bedrock/models/us.anthropic.claude-haiku-4-5-20251001-v1:0.toml
index 812dfc952dc..581497ca08a 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-haiku-4-5-20251001-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-haiku-4-5-20251001-v1:0.toml
@@ -1,27 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-haiku-4-5-20251001"
 name = "Claude Haiku 4.5 (US)"
-description = "Fast Claude model for responsive assistance, classification, and lightweight agents"
-family = "claude-haiku"
-release_date = "2025-10-15"
-last_updated = "2025-10-15"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-02-28"
-open_weights = false
 
 [cost]
-input = 1.00
-output = 5.00
-cache_read = 0.10
-cache_write = 1.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
+input = 1.10
+output = 5.50
+cache_read = 0.11
+cache_write = 1.375
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-1-20250805-v1:0.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-1-20250805-v1:0.toml
index f91effbdd45..c09142b2825 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-1-20250805-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-1-20250805-v1:0.toml
@@ -1,5 +1,11 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-1.html
+# Exact us-east-1 Standard dimensions are $15/$75 with $1.50/$18.75 cache rates; this legacy SKU is not in the newer 1.1x band.
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
 base_model = "anthropic/claude-opus-4-1-20250805"
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 status = "deprecated"
 name = "Claude Opus 4.1 (US)"
 
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-5-20251101-v1:0.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-5-20251101-v1:0.toml
index b35bdc03dc6..6666804ef8c 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-5-20251101-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-5-20251101-v1:0.toml
@@ -1,27 +1,17 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-5.html
+# https://platform.claude.com/docs/en/models/opus-4-5/overview
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Effort/budget: additionalModelRequestFields.output_config.effort / thinking.budget_tokens (Converse); output_config.effort / thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-opus-4-5-20251101"
 name = "Claude Opus 4.5 (US)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-11-24"
-last_updated = "2025-08-01"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-03-31"
-open_weights = false
 
 [cost]
-input = 5.00
-output = 25.00
-cache_read = 0.50
-cache_write = 6.25
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-6-v1.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-6-v1.toml
index 58b7cefede8..e2db7b054c7 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-6-v1.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-6-v1.toml
@@ -1,10 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-6.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Opus 4.6 (US)"
 structured_output = true
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-7.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-7.toml
index ae9e774662d..3433dbded02 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-7.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-7.toml
@@ -1,26 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-7.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-4-7"
 name = "Claude Opus 4.7 (US)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2026-04-16"
-last_updated = "2026-04-16"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-temperature = false
-tool_call = true
-knowledge = "2026-01-31"
-open_weights = false
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
-input = 5.00
-output = 25.00
-cache_read = 0.50
-cache_write = 6.25
-
-[limit]
-context = 1_000_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-8.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-8.toml
index 284fbe711a5..7f0036c92db 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-8.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-4-8.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-4-8.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 4.8 (US)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-5.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-5.toml
index f9ab43be58b..601aa76cbba 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-opus-5.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-5.toml
@@ -1,9 +1,15 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-opus-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel). Disabled thinking requires low/medium/high.
 base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Opus 5 (US)"
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 5.50
+output = 27.50
+cache_read = 0.55
+cache_write = 6.875
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-20250514-v1:0.toml b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-20250514-v1:0.toml
new file mode 100644
index 00000000000..59dc9cc90ad
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-20250514-v1:0.toml
@@ -0,0 +1,21 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4.html
+# https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListInferenceProfiles.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# Exact legacy Standard dimensions remain 1.0x ($3/$15), unlike newer Claude geo SKUs.
+# Profile verified from us-east-1; set AWS_REGION to a supported source (us-east-1, us-east-2, us-west-1, us-west-2).
+# Uses the native Bedrock provider's bedrock-runtime Converse/InvokeModel surface.
+# On-demand USD/MTok from us-east-1; 5-minute cache writes. Legacy, EOL October 14, 2026.
+# Live US-family probe, 2026-09-09: Legacy access denied (404) without active use in the last 30 days.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-20250514"
+name = "Claude Sonnet 4 (US)"
+status = "deprecated"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-5-20250929-v1:0.toml b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
index 0aab63954a5..75e0f1219c4 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-5-20250929-v1:0.toml
@@ -1,27 +1,16 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = enabled|disabled (Converse); thinking.type (InvokeModel).
+# Budget: additionalModelRequestFields.thinking.budget_tokens (Converse); thinking.budget_tokens (InvokeModel).
+base_model = "anthropic/claude-sonnet-4-5-20250929"
 name = "Claude Sonnet 4.5 (US)"
-description = "Balanced Claude model for coding, analysis, agent workflows, and cost control"
-family = "claude-sonnet"
-release_date = "2025-09-29"
-last_updated = "2025-09-29"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1_024 }]
-temperature = true
-tool_call = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024 }]
 structured_output = true
-knowledge = "2025-07-31"
-open_weights = false
 
 [cost]
-input = 3.00
-output = 15.00
-cache_read = 0.30
-cache_write = 3.75
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-6.toml b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-6.toml
index 0ca79f293f2..f3767fec0e6 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-6.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-6.toml
@@ -1,10 +1,20 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-4-6.html
+# Live us-east-1 Converse (2026-09-09): maxTokens=128000 succeeds; 128001 is rejected with model limit 128000.
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: regional on-demand USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Manual thinking.type = enabled still accepts thinking.budget_tokens; effort is output_config.effort, both inside additionalModelRequestFields on Converse.
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
 name = "Claude Sonnet 4.6 (US)"
 structured_output = true
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
+
+[limit]
+output = 128_000
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-5.toml b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-5.toml
index 0f3cef71d84..31c8e20733d 100644
--- a/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-5.toml
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-sonnet-5.toml
@@ -1,10 +1,18 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-anthropic-claude-sonnet-5.html
+# https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting
+# API: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
+# Live US/global Sonnet 5 schema probes, 2026-09-09: outputConfig.textFormat rejected (output_config.format: Extra inputs are not permitted).
+# Pricing: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json
+# US source us-east-1: standard regional USD/MTok; cache_write is the 5-minute rate.
+# Toggle: additionalModelRequestFields.thinking.type = adaptive|disabled (Converse); thinking.type (InvokeModel).
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
 base_model = "anthropic/claude-sonnet-5"
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 name = "Claude Sonnet 5 (US)"
-structured_output = true
+structured_output = false
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
+input = 2.20
+output = 11
+cache_read = 0.22
+cache_write = 2.75

From fdc824cc7ba4b594fb75d057c1353a8070817eeb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 20:25:47 +0000
Subject: [PATCH 1386/2012] chore(sync): update OpenRouter model catalog
 (#7189)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576

From 8b35e067ad9ffe175dfdecb00b617313ea88da5d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 20:25:57 +0000
Subject: [PATCH 1387/2012] chore(sync): update Merge Gateway model catalog
 (#7190)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 0264c6bc733..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 0cf8b2722718a41be9408f1d9cf694f052adfb0e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 21:23:39 +0000
Subject: [PATCH 1388/2012] chore(sync): update LLM Gateway model catalog
 (#7191)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/glm-5.3.toml                   | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/alibaba/glm-5.3.toml

diff --git a/providers/llmgateway-providers/models/alibaba/glm-5.3.toml b/providers/llmgateway-providers/models/alibaba/glm-5.3.toml
new file mode 100644
index 00000000000..1cde92a3889
--- /dev/null
+++ b/providers/llmgateway-providers/models/alibaba/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (Alibaba Cloud)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.28
+
+[limit]
+context = 1_048_576

From 676bb7c41b36e1a93ecf995c519aed92b77b6c92 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 21:23:41 +0000
Subject: [PATCH 1389/2012] chore(sync): update OpenRouter model catalog
 (#7192)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576

From 92bca2ea6f8c5d47e6e034d3e60bfe540a4c1bc3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 22:24:55 +0000
Subject: [PATCH 1390/2012] chore(sync): update OpenRouter model catalog
 (#7193)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..f7dc249d032 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.0763
+output = 0.1526
+cache_read = 0.01526
 
 [limit]
 context = 1_048_576

From 12660c29fdf7a48665f904e2da56a46d703bb9a6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 23:23:49 +0000
Subject: [PATCH 1391/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7194)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../inference-net/schematron-v2-small.toml    | 22 +++++++++++++++++++
 .../inference-net/schematron-v2-turbo.toml    | 22 +++++++++++++++++++
 .../vercel/models/openai/gpt-live-1.toml      | 17 ++++++++++++++
 3 files changed, 61 insertions(+)
 create mode 100644 providers/vercel/models/inference-net/schematron-v2-small.toml
 create mode 100644 providers/vercel/models/inference-net/schematron-v2-turbo.toml
 create mode 100644 providers/vercel/models/openai/gpt-live-1.toml

diff --git a/providers/vercel/models/inference-net/schematron-v2-small.toml b/providers/vercel/models/inference-net/schematron-v2-small.toml
new file mode 100644
index 00000000000..441cd688f11
--- /dev/null
+++ b/providers/vercel/models/inference-net/schematron-v2-small.toml
@@ -0,0 +1,22 @@
+name = "Schematron V2 Small"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "o"
+release_date = "2026-04-16"
+last_updated = "2026-04-16"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.05
+output = 0.23
+cache_read = 0.05
+
+[limit]
+context = 128_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/inference-net/schematron-v2-turbo.toml b/providers/vercel/models/inference-net/schematron-v2-turbo.toml
new file mode 100644
index 00000000000..5ac1c1c9009
--- /dev/null
+++ b/providers/vercel/models/inference-net/schematron-v2-turbo.toml
@@ -0,0 +1,22 @@
+name = "Schematron V2 Turbo"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "o"
+release_date = "2026-04-16"
+last_updated = "2026-04-16"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.03
+output = 0.15
+cache_read = 0.03
+
+[limit]
+context = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/vercel/models/openai/gpt-live-1.toml b/providers/vercel/models/openai/gpt-live-1.toml
new file mode 100644
index 00000000000..afdef373ba0
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-live-1.toml
@@ -0,0 +1,17 @@
+name = "GPT-Live 1"
+description = "Speech generation model for controllable voice, narration, and audio delivery"
+family = "gpt"
+release_date = "2026-09-10"
+last_updated = "2026-09-10"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "audio"]
+output = ["text", "audio"]

From 43bde61a201e2c5f257c4af2bf0fe7731edefe48 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 15 Sep 2026 23:24:08 +0000
Subject: [PATCH 1392/2012] chore(sync): update OpenRouter model catalog
 (#7195)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml  | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index df31c0bbb32..2331765a04d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.12
-cache_read = 0.012
+input = 0.055
+output = 0.11
+cache_read = 0.011
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f7dc249d032..d6a823fa222 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0763
-output = 0.1526
-cache_read = 0.01526
+input = 0.07476
+output = 0.14952
+cache_read = 0.014952
 
 [limit]
 context = 1_048_576

From e388d7d404088c08b57b7432069e765cc81e99d3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 00:59:47 +0000
Subject: [PATCH 1393/2012] chore(sync): update NanoGPT model catalog (#7197)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/gemma-4-12b-it.toml |  4 +--
 providers/nano-gpt/models/gemma-4-e2b-it.toml | 15 -----------
 providers/nano-gpt/models/gemma-4-e4b-it.toml | 15 -----------
 .../nano-gpt/models/qwen/qwen3.5-0.8b.toml    | 25 -------------------
 .../nano-gpt/models/qwen/qwen3.5-2b.toml      | 25 -------------------
 .../nano-gpt/models/qwen/qwen3.5-4b.toml      | 25 -------------------
 6 files changed, 2 insertions(+), 107 deletions(-)
 delete mode 100644 providers/nano-gpt/models/gemma-4-e2b-it.toml
 delete mode 100644 providers/nano-gpt/models/gemma-4-e4b-it.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.5-2b.toml
 delete mode 100644 providers/nano-gpt/models/qwen/qwen3.5-4b.toml

diff --git a/providers/nano-gpt/models/gemma-4-12b-it.toml b/providers/nano-gpt/models/gemma-4-12b-it.toml
index 7c2890a9c37..6959fb985f1 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it.toml
@@ -15,8 +15,8 @@ output = 0.25
 cache_read = 0.025
 
 [limit]
-context = 262_144
-input = 262_144
+context = 131_072
+input = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/gemma-4-e2b-it.toml b/providers/nano-gpt/models/gemma-4-e2b-it.toml
deleted file mode 100644
index b6becef6c3b..00000000000
--- a/providers/nano-gpt/models/gemma-4-e2b-it.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "google/gemma-4-E2B-it"
-name = "Gemma 4 E2B Instruct"
-reasoning = false
-
-[cost]
-input = 0.02
-output = 0.1
-cache_read = 0.01
-
-[limit]
-input = 131_072
-output = 16_384
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/gemma-4-e4b-it.toml b/providers/nano-gpt/models/gemma-4-e4b-it.toml
deleted file mode 100644
index 931ddf4bf29..00000000000
--- a/providers/nano-gpt/models/gemma-4-e4b-it.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "google/gemma-4-E4B-it"
-name = "Gemma 4 E4B Instruct"
-reasoning = false
-
-[cost]
-input = 0.04
-output = 0.2
-cache_read = 0.02
-
-[limit]
-input = 131_072
-output = 16_384
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml b/providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml
deleted file mode 100644
index 4e2cecccc6e..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.5-0.8b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen3.5 0.8B"
-description = "Qwen3.5 0.8B is a lightweight open-weight multimodal model from Alibaba for fast reasoning, visual understanding, tool use, and JSON output."
-family = "qwen3.5"
-release_date = "2026-08-16"
-last_updated = "2026-08-16"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.06
-output = 0.12
-cache_read = 0.03
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-2b.toml b/providers/nano-gpt/models/qwen/qwen3.5-2b.toml
deleted file mode 100644
index cf562f087e6..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.5-2b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen3.5 2B"
-description = "Qwen3.5 2B is a small open-weight multimodal model from Alibaba for efficient reasoning, coding, visual understanding, tool use, and JSON output."
-family = "qwen3.5"
-release_date = "2026-08-16"
-last_updated = "2026-08-16"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.08
-output = 0.16
-cache_read = 0.04
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.5-4b.toml b/providers/nano-gpt/models/qwen/qwen3.5-4b.toml
deleted file mode 100644
index 0538b8a5be8..00000000000
--- a/providers/nano-gpt/models/qwen/qwen3.5-4b.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen3.5 4B"
-description = "Qwen3.5 4B is a compact open-weight multimodal model from Alibaba for reasoning, coding, visual understanding, tool use, and structured output."
-family = "qwen3.5"
-release_date = "2026-08-16"
-last_updated = "2026-08-16"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.1
-output = 0.2
-cache_read = 0.05
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 013fa51555f73720de7b2e8b600b450578826010 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 00:59:58 +0000
Subject: [PATCH 1394/2012] chore(sync): update Kilo model catalog (#7198)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 .../mistralai/mistral-small-3.2-24b-instruct.toml      |  6 +++---
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml |  4 ++--
 providers/kilo/models/qwen/qwen3.7-max.toml            |  8 ++++----
 providers/kilo/models/qwen/qwen3.7-plus.toml           |  2 +-
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml   |  4 ++--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     |  8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 9 files changed, 25 insertions(+), 26 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml b/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
index 68645a3ea09..acb807ef2ce 100644
--- a/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
+++ b/providers/kilo/models/mistralai/mistral-small-3.2-24b-instruct.toml
@@ -11,11 +11,11 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.075
-output = 0.2
+input = 0.09375
+output = 0.25
 
 [limit]
-context = 128_000
+context = 256_000
 output = 16_384
 
 [modalities]
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 25c72a0b0fb..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 202_800
-output = 182_520
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/qwen/qwen3.7-max.toml b/providers/kilo/models/qwen/qwen3.7-max.toml
index 9a5aff06afe..a57091e05fe 100644
--- a/providers/kilo/models/qwen/qwen3.7-max.toml
+++ b/providers/kilo/models/qwen/qwen3.7-max.toml
@@ -7,10 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 1.25
-output = 3.75
-cache_read = 0.125
-cache_write = 1.5625
+input = 1.475
+output = 4.425
+cache_read = 0.295
+cache_write = 1.84375
 
 [limit]
 output = 131_072
diff --git a/providers/kilo/models/qwen/qwen3.7-plus.toml b/providers/kilo/models/qwen/qwen3.7-plus.toml
index 61512d88eb0..9ec531e90c6 100644
--- a/providers/kilo/models/qwen/qwen3.7-plus.toml
+++ b/providers/kilo/models/qwen/qwen3.7-plus.toml
@@ -9,7 +9,7 @@ values = ["none", "high"]
 [cost]
 input = 0.32
 output = 1.28
-cache_read = 0.032
+cache_read = 0.064
 cache_write = 0.4
 
 [limit]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index f491c2853c8..673b730eb39 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_000_000
-output = 393_216
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 38aa9fed430..d603d7bc671 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.85
+output = 2.8985
+cache_read = 0.17
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From d87ec357c1139e98ec8ae653efd7937ba0b718e9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 01:00:01 +0000
Subject: [PATCH 1395/2012] chore(sync): update OpenRouter model catalog
 (#7196)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/minimax/minimax-m1.toml       | 2 +-
 .../models/mistralai/mistral-small-3.2-24b-instruct.toml  | 4 ++--
 .../openrouter/models/moonshotai/kimi-k2.7-code.toml      | 6 +++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml       | 6 +++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 11 files changed, 31 insertions(+), 32 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d6a823fa222..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07476
-output = 0.14952
-cache_read = 0.014952
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..451e7c3012c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.9834
+output = 2.9502
+cache_read = 0.03278
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/minimax/minimax-m1.toml b/providers/openrouter/models/minimax/minimax-m1.toml
index 178789497a8..063613f9d24 100644
--- a/providers/openrouter/models/minimax/minimax-m1.toml
+++ b/providers/openrouter/models/minimax/minimax-m1.toml
@@ -17,7 +17,7 @@ open_weights = false
 type = "toggle"
 
 [cost]
-input = 0.55
+input = 0.4
 output = 2.2
 
 [limit]
diff --git a/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml b/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
index a05f6600220..809cb9dd124 100644
--- a/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
+++ b/providers/openrouter/models/mistralai/mistral-small-3.2-24b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2023-10-31"
 open_weights = true
 
 [cost]
-input = 0.075
-output = 0.2
+input = 0.09375
+output = 0.25
 
 [limit]
 context = 256_000
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
index 0006cbd4e10..307383582d9 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
@@ -3,9 +3,9 @@ temperature = true
 reasoning_options = []
 
 [cost]
-input = 0.71
-output = 3.5
-cache_read = 0.15
+input = 0.7062
+output = 3.21
+cache_read = 0.18
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 5dcd117448e..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 182_520
+output = 32_768
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index d5918c075a2..96c5bbb7f77 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.1
+output = 0.3333
+cache_read = 0.02
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 98fd23a037b..dc34ba0ccf7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3e3ce108967..14b0cdd0d13 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.85
+output = 2.8985
+cache_read = 0.17
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From f0ba113021484a4b93ff9c3b7a54e3666970e85d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 01:35:20 +0000
Subject: [PATCH 1396/2012] chore(sync): update OpenRouter model catalog
 (#7201)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..93d4a9d7ab9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 14b0cdd0d13..3e3ce108967 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.85
-output = 2.8985
-cache_read = 0.17
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From 63c629f7d72beb7623729737d08447bd6057a7d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 01:35:23 +0000
Subject: [PATCH 1397/2012] chore(sync): update Kilo model catalog (#7200)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     |  8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..a43a707c98d 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index d603d7bc671..38aa9fed430 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.85
-output = 2.8985
-cache_read = 0.17
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From a734b83ae19d5e3b6ca7d3d5f3bc9abb589892cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 01:35:27 +0000
Subject: [PATCH 1398/2012] chore(sync): update NanoGPT model catalog (#7202)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From e4d18bc4d7ebfb1394acd20789c95b70be3762e1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 04:30:42 +0000
Subject: [PATCH 1399/2012] chore(sync): update Kilo model catalog (#7205)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 4 ++--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 83d4e34b47c..dee6c4b128f 100644
--- a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 262_144
-output = 16_384
+context = 131_072
+output = 32_768
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index a43a707c98d..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 803a222b95c022016c5a5e4c39b35242af9429d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 04:30:47 +0000
Subject: [PATCH 1400/2012] chore(sync): update NanoGPT model catalog (#7207)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From d35a1ed9a7e92659644baf4584e6d47e12d2317e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 04:30:54 +0000
Subject: [PATCH 1401/2012] chore(sync): update OpenRouter model catalog
 (#7206)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 2331765a04d..df31c0bbb32 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.055
-output = 0.11
-cache_read = 0.011
+input = 0.06
+output = 0.12
+cache_read = 0.012
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 3f78c99ff08..d857fa53dab 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,12 +12,12 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.15
-output = 0.6
+input = 0.13
+output = 0.52
 
 [limit]
 context = 262_144
-output = 16_384
+output = 32_768
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 93d4a9d7ab9..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 23e22d292e9f282b23eb155a76c64ae8ef27d44b Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Wed, 16 Sep 2026 00:04:54 -0500
Subject: [PATCH 1402/2012] fix(zhipuai): move GLM flash models to the
 glm-flash family (#7208)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 models/zhipuai/glm-4.6v-flash.toml                              | 2 +-
 models/zhipuai/glm-5.3-flash.toml                               | 2 +-
 providers/cline-pass/models/cline-pass/glm-5.3-flash.toml       | 2 +-
 .../cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml | 2 +-
 providers/huggingface/models/zai-org/GLM-4.7-Flash.toml         | 2 +-
 providers/jiekou/models/zai-org/glm-4.7-flash.toml              | 2 +-
 providers/llmgateway-providers/models/zai/glm-4.6v-flashx.toml  | 2 +-
 providers/llmgateway/models/glm-4.6v-flashx.toml                | 2 +-
 providers/nano-gpt/models/glm-4.1v-thinking-flash.toml          | 2 +-
 providers/nano-gpt/models/glm-4.1v-thinking-flashx.toml         | 2 +-
 providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml      | 2 +-
 .../nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml   | 2 +-
 providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml    | 2 +-
 providers/novita-ai/models/zai-org/glm-4.7-flash.toml           | 2 +-
 providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml | 2 +-
 providers/venice/models/zai-org-glm-4.7-flash.toml              | 1 -
 16 files changed, 15 insertions(+), 16 deletions(-)

diff --git a/models/zhipuai/glm-4.6v-flash.toml b/models/zhipuai/glm-4.6v-flash.toml
index c9a3789715b..12d051076bb 100644
--- a/models/zhipuai/glm-4.6v-flash.toml
+++ b/models/zhipuai/glm-4.6v-flash.toml
@@ -3,7 +3,7 @@
 # - https://huggingface.co/zai-org/GLM-4.6V-Flash
 name = "GLM-4.6V-Flash"
 description = "Lightweight GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
 attachment = true
diff --git a/models/zhipuai/glm-5.3-flash.toml b/models/zhipuai/glm-5.3-flash.toml
index d080a79f86f..12091c3b2be 100644
--- a/models/zhipuai/glm-5.3-flash.toml
+++ b/models/zhipuai/glm-5.3-flash.toml
@@ -1,7 +1,7 @@
 # Open weights: https://huggingface.co/zai-org/GLM-5.3-Flash (MIT, safetensors)
 name = "GLM-5.3-Flash"
 description = "Native multimodal GLM model for efficient coding and long-horizon agent tasks"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-08-26"
 last_updated = "2026-08-26"
 attachment = true
diff --git a/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml b/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
index 7f9e9f9ec7d..3c24c9833c2 100644
--- a/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
+++ b/providers/cline-pass/models/cline-pass/glm-5.3-flash.toml
@@ -5,7 +5,7 @@
 # Z.ai's GLM-5.3-Flash (https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash).
 name = "cline-pass/glm-5.3-flash"
 description = "Latest natively multimodal model in the GLM-5 series"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-08-26"
 last_updated = "2026-08-26"
 attachment = true
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
index 5a0a6c35b26..000ebf8a49a 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -3,7 +3,7 @@
 # https://developers.cloudflare.com/workers-ai/models/glm-5.3-flash/
 name = "Glm 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-08-26"
 last_updated = "2026-08-26"
 attachment = true
diff --git a/providers/huggingface/models/zai-org/GLM-4.7-Flash.toml b/providers/huggingface/models/zai-org/GLM-4.7-Flash.toml
index b0524072f5c..aa24fe9b12f 100644
--- a/providers/huggingface/models/zai-org/GLM-4.7-Flash.toml
+++ b/providers/huggingface/models/zai-org/GLM-4.7-Flash.toml
@@ -1,6 +1,6 @@
 name = "GLM-4.7-Flash"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-08-08"
 last_updated = "2025-08-08"
 attachment = false
diff --git a/providers/jiekou/models/zai-org/glm-4.7-flash.toml b/providers/jiekou/models/zai-org/glm-4.7-flash.toml
index 81dc436cc68..6d3d8dda497 100644
--- a/providers/jiekou/models/zai-org/glm-4.7-flash.toml
+++ b/providers/jiekou/models/zai-org/glm-4.7-flash.toml
@@ -1,6 +1,6 @@
 name = "GLM-4.7-Flash"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-01"
 last_updated = "2026-01"
 attachment = false
diff --git a/providers/llmgateway-providers/models/zai/glm-4.6v-flashx.toml b/providers/llmgateway-providers/models/zai/glm-4.6v-flashx.toml
index be8d5892ddd..423aaa1934e 100644
--- a/providers/llmgateway-providers/models/zai/glm-4.6v-flashx.toml
+++ b/providers/llmgateway-providers/models/zai/glm-4.6v-flashx.toml
@@ -4,7 +4,7 @@
 # https://docs.llmgateway.io/features/reasoning
 name = "GLM-4.6V FlashX (Z AI)"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
 attachment = true
diff --git a/providers/llmgateway/models/glm-4.6v-flashx.toml b/providers/llmgateway/models/glm-4.6v-flashx.toml
index 20ba7000522..c0f3104470b 100644
--- a/providers/llmgateway/models/glm-4.6v-flashx.toml
+++ b/providers/llmgateway/models/glm-4.6v-flashx.toml
@@ -1,6 +1,6 @@
 name = "GLM-4.6V FlashX"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
 attachment = true
diff --git a/providers/nano-gpt/models/glm-4.1v-thinking-flash.toml b/providers/nano-gpt/models/glm-4.1v-thinking-flash.toml
index 29e59994e81..07a69b18656 100644
--- a/providers/nano-gpt/models/glm-4.1v-thinking-flash.toml
+++ b/providers/nano-gpt/models/glm-4.1v-thinking-flash.toml
@@ -1,6 +1,6 @@
 name = "GLM 4.1V Thinking Flash"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-07-09"
 last_updated = "2025-07-09"
 attachment = true
diff --git a/providers/nano-gpt/models/glm-4.1v-thinking-flashx.toml b/providers/nano-gpt/models/glm-4.1v-thinking-flashx.toml
index 730a47adcf9..0d366cbcc97 100644
--- a/providers/nano-gpt/models/glm-4.1v-thinking-flashx.toml
+++ b/providers/nano-gpt/models/glm-4.1v-thinking-flashx.toml
@@ -1,6 +1,6 @@
 name = "GLM 4.1V Thinking FlashX"
 description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "glm"
+family = "glm-flash"
 release_date = "2025-07-09"
 last_updated = "2025-07-09"
 attachment = true
diff --git a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
index 4ff6c178da5..3491643442e 100644
--- a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
@@ -1,6 +1,6 @@
 name = "GLM 4.7 Flash Original"
 description = "GLM-4.7-Flash is a lightweight 30B model optimized for coding and agentic tasks. Balances high performance with efficiency, perfect for local deployment."
-family = "glm"
+family = "glm-flash"
 release_date = "2026-01-19"
 last_updated = "2026-01-19"
 attachment = false
diff --git a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
index 60fa9006fc9..65480647fea 100644
--- a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
+++ b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
@@ -1,6 +1,6 @@
 name = "GLM 4.7 Flash Original Thinking"
 description = "GLM-4.7-Flash with extended thinking capabilities for complex reasoning. Lightweight 30B model optimized for coding and agentic tasks."
-family = "glm"
+family = "glm-flash"
 release_date = "2026-01-19"
 last_updated = "2026-01-19"
 attachment = false
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
index 8136288a6f3..f0280c43a62 100644
--- a/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-uncensored.toml
@@ -1,6 +1,6 @@
 name = "GLM 5.3 Flash Uncensored"
 description = "GLM 5.3 Flash Uncensored is an uncensored fine-tune of the efficient 320B mixture-of-experts reasoning model, built for unrestricted chat, creative writing, coding, agentic work, tool use, and long-context tasks."
-family = "glm"
+family = "glm-flash"
 release_date = "2026-07-29"
 last_updated = "2026-08-27"
 attachment = true
diff --git a/providers/novita-ai/models/zai-org/glm-4.7-flash.toml b/providers/novita-ai/models/zai-org/glm-4.7-flash.toml
index b5e02bf1ccb..c0079c3decf 100644
--- a/providers/novita-ai/models/zai-org/glm-4.7-flash.toml
+++ b/providers/novita-ai/models/zai-org/glm-4.7-flash.toml
@@ -1,6 +1,6 @@
 name = "GLM-4.7-Flash"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-01-19"
 last_updated = "2026-01-19"
 attachment = false
diff --git a/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml b/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
index b3fbe1e97a2..6134bf75d0c 100644
--- a/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
+++ b/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
@@ -2,7 +2,7 @@
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
 name = "GLM 4.7 Flash Heretic"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm"
+family = "glm-flash"
 release_date = "2026-02-04"
 last_updated = "2026-06-11"
 attachment = false
diff --git a/providers/venice/models/zai-org-glm-4.7-flash.toml b/providers/venice/models/zai-org-glm-4.7-flash.toml
index 126400b50d5..854e1fd3c20 100644
--- a/providers/venice/models/zai-org-glm-4.7-flash.toml
+++ b/providers/venice/models/zai-org-glm-4.7-flash.toml
@@ -2,7 +2,6 @@
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
 base_model = "zhipuai/glm-4.7-flash"
 name = "GLM 4.7 Flash"
-family = "glm"
 release_date = "2026-01-29"
 last_updated = "2026-06-11"
 structured_output = true

From 68d4f1dd978d63eb21dd1c928712f40ff8a94222 Mon Sep 17 00:00:00 2001
From: MickeyXin <49945543+fx247562340@users.noreply.github.com>
Date: Wed, 16 Sep 2026 13:15:08 +0800
Subject: [PATCH 1403/2012] fix(vancine): update DeepSeek V4.1 Flash model ID
 (#7164)

Co-authored-by: fx247562340 <fx247562340@users.noreply.github.com>
---
 .../{deepseek-flash.toml => deepseek-v4.1-flash.toml}       | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
 rename providers/vancine/models/{deepseek-flash.toml => deepseek-v4.1-flash.toml} (80%)

diff --git a/providers/vancine/models/deepseek-flash.toml b/providers/vancine/models/deepseek-v4.1-flash.toml
similarity index 80%
rename from providers/vancine/models/deepseek-flash.toml
rename to providers/vancine/models/deepseek-v4.1-flash.toml
index 022e839c94b..3b03b84f862 100644
--- a/providers/vancine/models/deepseek-flash.toml
+++ b/providers/vancine/models/deepseek-v4.1-flash.toml
@@ -1,9 +1,11 @@
-# Sources (accessed 2026-09-11):
+# DeepSeek official sources and lab model (accessed 2026-09-11):
 # https://api-docs.deepseek.com/quick_start/pricing/
 # https://api-docs.deepseek.com/guides/thinking_mode/
 # models/deepseek/deepseek-v4.1-flash.toml
 # providers/deepseek/models/deepseek-flash.toml
-# Vancine serves DeepSeek V4.1 Flash as the platform model id deepseek-flash.
+# Vancine current platform model ID deepseek-v4.1-flash and public pricing
+# re-verified from https://vancine.com/api/pricing at 2026-09-15T07:51:35Z.
+# base_model remains deepseek/deepseek-v4.1-flash.
 # Reasoning HTTP format: thinking.type = "enabled" | "disabled";
 # reasoning_effort = "low" | "high" | "max". Flash maps low -> low
 # (unlike Pro). xhigh maps to high. Reasoning tokens are billed at the

From 88c4dd7c416457e5d4cff960ca7406fd8590482f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=BC=A0=E5=9C=A3=E5=A5=87?=
 <zhangshengqi20@mails.ucas.ac.cn>
Date: Wed, 16 Sep 2026 13:15:17 +0800
Subject: [PATCH 1404/2012] feat(alibaba-cn): add glm-5.3 (#7183)

---
 providers/alibaba-cn/models/glm-5.3.toml | 25 ++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/alibaba-cn/models/glm-5.3.toml

diff --git a/providers/alibaba-cn/models/glm-5.3.toml b/providers/alibaba-cn/models/glm-5.3.toml
new file mode 100644
index 00000000000..f6d0f6e67e3
--- /dev/null
+++ b/providers/alibaba-cn/models/glm-5.3.toml
@@ -0,0 +1,25 @@
+# GLM-5.3 on Alibaba Bailian (Model Studio, Beijing): always-on thinking with
+# reasoning_effort low|high|max (default max). Thinking cannot be turned off
+# (enable_thinking only accepts true; passing false errors) and there is no
+# thinking_budget — so a single effort option, no toggle, no budget.
+# reasoning_content is streamed in deltas.
+# Pricing: Beijing CNY 8 / 28 / 2 (cache hit) per 1M tokens, same list price
+# as glm-5.2 on the same host (see providers/alibaba-cn/models/glm-5.2.toml).
+# Limits and modalities are identical to the zhipuai lab entry
+# (1M context, 131_072 output, text-only), so they are inherited and not restated.
+# https://bailian.console.aliyun.com/cn-beijing?tab=model#/model-market/detail/glm-5.3?serviceSite=asia-pacific-china
+# https://docs.bigmodel.cn/cn/guide/models/text/glm-5.3
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.1
+output = 3.851
+cache_read = 0.275
+cache_write = 0

From a20debfd6113785af87b0811c0017bfa32b92da0 Mon Sep 17 00:00:00 2001
From: Oliver Mee <102673257+oliver-mee@users.noreply.github.com>
Date: Wed, 16 Sep 2026 13:15:27 +0800
Subject: [PATCH 1405/2012] fix(alibaba-token-plan): correct
 deepseek-v4.1-flash structured_output for the plan gateway (#7203)

---
 .../models/deepseek-v4.1-flash.toml           | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml b/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml
index f32c843d2e8..6856b6ad35b 100644
--- a/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml
+++ b/providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml
@@ -1,8 +1,10 @@
 # Reasoning HTTP format (accessed 2026-09-15):
-# Toggle: enable_thinking = true|false. Chat Completions `reasoning_effort`
-# accepts low, high (default), and max; medium and xhigh map to high.
-# Reasoning streams on `reasoning_content`.
-# Model availability: Token Plan Personal and Team editions (Singapore).
+# Toggle: enable_thinking = true|false. Thinking content streams on
+# `reasoning_content`. Chat Completions `reasoning_effort` accepts the string
+# ladder low/high (default)/max on the Token Plan gateway (probed 2026-09-15);
+# medium and xhigh are accepted as aliases per family precedent. The vendor
+# OpenAI-chat reference publishes integer effort (1-100); the plan gateway
+# rejects integers with a 400 for this id (probed 2026-09-15).
 # Sources:
 # https://docs.qwencloud.com/api-reference/chat/openai-chat
 # https://www.alibabacloud.com/help/en/model-studio/deepseek-api
@@ -10,6 +12,15 @@
 # https://www.alibabacloud.com/help/en/model-studio/token-plan-team-overview
 base_model = "deepseek/deepseek-v4.1-flash"
 
+# Structured output: PARTIAL on this gateway despite the lab row's
+# structured_output = true. response_format json_object works, but strict
+# json_schema is rejected with 400 "This response_format type is unavailable
+# now" in both thinking modes (probed 2026-09-15 on the international Token
+# Plan gateway, Personal and Team editions). Omitting this field would
+# inherit the lab value and advertise an unsupported capability, so the
+# override is explicit.
+structured_output = false
+
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]

From 95963b54c8879da564d23949038dddab3337e17f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 05:27:25 +0000
Subject: [PATCH 1406/2012] chore(sync): update Weights & Biases model catalog
 (#7210)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/wandb/models/zai-org/GLM-5.3-Flash.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
index 91d7e42d00a..678fccd1812 100644
--- a/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
+++ b/providers/wandb/models/zai-org/GLM-5.3-Flash.toml
@@ -3,6 +3,7 @@
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 Flash"
 description = "GLM-5.3-Flash is a natively multimodal model with 320B total parameters and 18B active parameters."
+family = "glm"
 reasoning_options = []
 
 [cost]

From 9576187b9b0ee21e57eb8268547f3b570904271f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 05:27:30 +0000
Subject: [PATCH 1407/2012] chore(sync): update Venice model catalog (#7209)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml b/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
index 6134bf75d0c..b3fbe1e97a2 100644
--- a/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
+++ b/providers/venice/models/olafangensan-glm-4.7-flash-heretic.toml
@@ -2,7 +2,7 @@
 # https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
 name = "GLM 4.7 Flash Heretic"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm-flash"
+family = "glm"
 release_date = "2026-02-04"
 last_updated = "2026-06-11"
 attachment = false

From 3c4992e5b2478d69a41db974569d34a8c0e030e6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 01:17:50 -0500
Subject: [PATCH 1408/2012] fix(sync): repair Fireworks catalog updates (#7212)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/fireworks-ai.ts   | 28 +++++++++++++------
 packages/core/test/fireworks-ai-sync.test.ts  |  8 ++++++
 .../fireworks/models/deepseek-v4-pro.toml     | 17 +++++++----
 .../fireworks/models/minimax-m2p7.toml        | 14 ++++++----
 .../accounts/fireworks/models/minimax-m3.toml |  2 +-
 .../fireworks/routers/glm-flash-latest.toml   |  4 +++
 .../fireworks/routers/minimax-latest.toml     |  3 +-
 .../fireworks/routers/qwen-max-latest.toml    | 26 +++++++++++++++++
 sync.md                                       |  4 +--
 9 files changed, 82 insertions(+), 24 deletions(-)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/qwen-max-latest.toml

diff --git a/packages/core/src/sync/providers/fireworks-ai.ts b/packages/core/src/sync/providers/fireworks-ai.ts
index aa1d366ca9e..e6fc5cf6bc5 100644
--- a/packages/core/src/sync/providers/fireworks-ai.ts
+++ b/packages/core/src/sync/providers/fireworks-ai.ts
@@ -113,7 +113,9 @@ export function expandFireworksModels(models: FireworksModel[]): FireworksCatalo
 
     for (const model of rows) {
       if (model.usage_identifier !== undefined) add(model.usage_identifier, model, []);
-      for (const alias of model.aliases ?? []) add(alias, model, []);
+      for (const alias of model.aliases ?? []) {
+        add(alias, model, model === defaultRow ? flagModes : []);
+      }
     }
   }
   return [...expanded.values()];
@@ -136,13 +138,23 @@ function catalogModalities(values: string[], fallback: Modality[]): Modality[] {
   return modalities.length === 0 ? fallback : modalities;
 }
 
-function pricing(model: Pick<FireworksModel, "pricing">, existing: NonNullable<SyncedFullModel["cost"]>) {
+function pricing(
+  model: Pick<FireworksModel, "id" | "pricing" | "serverless_mode">,
+  existing?: NonNullable<SyncedFullModel["cost"]>,
+): NonNullable<SyncedFullModel["cost"]> {
   const bySku = new Map(model.pricing.map((price) => [price.sku, Number(price.amount)]));
+  const input = bySku.get("LLM input tokens (uncached)") ?? existing?.input;
+  const output = bySku.get("LLM output tokens") ?? existing?.output;
+  if (input === undefined || output === undefined) {
+    throw new Error(
+      `Fireworks AI model ${model.id} ${model.serverless_mode} mode has incomplete token pricing`,
+    );
+  }
   return {
     ...existing,
-    input: bySku.get("LLM input tokens (uncached)") ?? existing.input,
-    cache_read: bySku.get("LLM input tokens (cached)") ?? existing.cache_read,
-    output: bySku.get("LLM output tokens") ?? existing.output,
+    input,
+    cache_read: bySku.get("LLM input tokens (cached)") ?? existing?.cache_read,
+    output,
   };
 }
 
@@ -218,11 +230,11 @@ export function buildFireworksModel(
     || limit.context === undefined
     || limit.output === undefined
     || modalities === undefined
-    || cost === undefined
   ) {
     throw new Error(`Fireworks AI model ${model.catalogId} has incomplete local TOML metadata required for sync`);
   }
 
+  const modelCost = pricing(model, cost);
   const input = catalogModalities(model.input_modalities, modalities.input);
   const outputModalities = catalogModalities(model.output_modalities, modalities.output);
   // Fireworks reports the advertised context window, while some deployments
@@ -248,7 +260,7 @@ export function buildFireworksModel(
     open_weights: openWeights,
     status: existing.status,
     interleaved: existing.interleaved,
-    cost: pricing(model, cost),
+    cost: modelCost,
     limit: {
       context,
       input: limit.input,
@@ -259,7 +271,7 @@ export function buildFireworksModel(
       output: outputModalities,
     },
     provider: provider(model, existing.provider),
-    experimental: experimental(model, cost, existing.experimental),
+    experimental: experimental(model, modelCost, existing.experimental),
   } satisfies SyncedFullModel;
 
   return existing.base_model === undefined
diff --git a/packages/core/test/fireworks-ai-sync.test.ts b/packages/core/test/fireworks-ai-sync.test.ts
index 54c531336f4..4f352794ab7 100644
--- a/packages/core/test/fireworks-ai-sync.test.ts
+++ b/packages/core/test/fireworks-ai-sync.test.ts
@@ -62,6 +62,7 @@ test("expands usage identifiers and aliases and attaches flag-only modes", () =>
     serverless_mode: "priority",
     service_tier: "priority",
   });
+  expect(models[1]?.flagModes).toHaveLength(1);
 });
 
 test("updates Fireworks pricing and modalities while preserving authored facts", () => {
@@ -88,6 +89,13 @@ test("updates Fireworks pricing and modalities while preserving authored facts",
   });
 });
 
+test("derives cost from Fireworks when the local model has no cost", () => {
+  const { cost: _, ...existing } = existingModel();
+  const model = buildFireworksModel(catalogModel(), existing);
+
+  expect(model.cost).toEqual({ input: 1.4, output: 4.4, cache_read: 0.14 });
+});
+
 test("uses the service-tier recipe for a priority-only model", () => {
   const priority = fireworksModel({ serverless_mode: "priority", service_tier: "priority" });
   const [model] = expandFireworksModels([priority]);
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
index ca93dd0ae06..84e0b5d334e 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
@@ -1,8 +1,6 @@
-# Fireworks lists DeepSeek-V4-Pro (preview) as on-demand only — Serverless is
-# "Not supported", so there is no per-token serverless price to record.
-# Model page: https://fireworks.ai/models/fireworks/deepseek-v4-pro (accessed 2026-09-15)
-# Serverless catalog omits this id (only DeepSeek V4 Pro 0813 is priced):
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-15)
+# GET /v1/serverless/models lists DeepSeek V4 Pro on Standard and Priority
+# at $1.20 / $0.60 / $1.20 per 1M (input / cached input / output) on
+# 2026-09-16. Both serving modes currently have the same price.
 # Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
 # Effort: reasoning_effort = high|max; low/medium promote to high, xhigh to max
 # Reasoning API: https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-09-15)
@@ -18,3 +16,12 @@ type = "toggle"
 [[reasoning_options]]
 type = "effort"
 values = ["high", "max"]
+
+[cost]
+input = 1.2
+output = 1.2
+cache_read = 0.6
+
+[experimental.modes.priority]
+cost = { input = 1.2, output = 1.2, cache_read = 0.6 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
index 2a241ff93f0..b64465a927f 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
@@ -1,7 +1,6 @@
-# https://fireworks.ai/models/fireworks/minimax-m2p7 (accessed 2026-09-15)
-# Serverless: $0.30 / $0.059 / $1.20 per 1M (input / cached input / output).
-# Context Length 196k on Fireworks (lab base is 204_800); HF MiniMaxAI/MiniMax-M2.7.
-# Not listed on docs.fireworks.ai/serverless/pricing headline table (no Priority tier published).
+# GET /v1/serverless/models lists MiniMax M2.7 as Priority-only at
+# $1.20 / $0.60 / $1.20 per 1M (input / cached input / output) with a
+# 196,608-token context on 2026-09-16.
 # $.reasoning_effort = "low" | "medium" | "high" (same Fireworks surface as MiniMax M3).
 # https://docs.fireworks.ai/guides/reasoning
 base_model = "minimax/MiniMax-M2.7"
@@ -14,9 +13,12 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.3
+input = 1.2
 output = 1.2
-cache_read = 0.059
+cache_read = 0.6
 
 [limit]
 context = 196_608
+
+[provider]
+body = { service_tier = "priority" }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
index 295f4a7d2cd..e87e3cd0cce 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m3.toml
@@ -1,7 +1,7 @@
 # $.reasoning_effort = "low" | "medium" | "high".
 # https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
 name = "MiniMax-M3"
-description = "MiniMax multimodal coding model for long-context reasoning and agent tasks"
+description = "Fireworks text-only MiniMax coding model for long-context reasoning and agent tasks"
 family = "minimax"
 release_date = "2026-06-12"
 last_updated = "2026-06-12"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml
index 9e47aa19fea..72dad231910 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-flash-latest.toml
@@ -34,3 +34,7 @@ context = 1_048_573
 
 [modalities]
 input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 0.1875, output = 0.625, cache_read = 0.0375 }
+provider = { body = { service_tier = "priority" } }
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml
index f9d41fb264a..3a4afb34e7f 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/minimax-latest.toml
@@ -14,9 +14,9 @@
 # https://docs.fireworks.ai/guides/reasoning
 base_model = "minimax/MiniMax-M3"
 name = "MiniMax Latest"
-attachment = false
 release_date = "2026-06-12"
 last_updated = "2026-09-15"
+attachment = false
 
 [interleaved]
 field = "reasoning_content"
@@ -32,7 +32,6 @@ cache_read = 0.06
 
 [limit]
 context = 512_000
-output = 512_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/qwen-max-latest.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/qwen-max-latest.toml
new file mode 100644
index 00000000000..f0ec43be08c
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/qwen-max-latest.toml
@@ -0,0 +1,26 @@
+# Version-tracking standard-path alias for Qwen 3.8 Max, advertised by
+# https://api.fireworks.ai/v1/serverless/models on 2026-09-16 as
+# accounts/fireworks/routers/qwen-max-latest. The same response supplies
+# standard and Priority pricing, text/image input, and the Priority recipe.
+# Toggle: thinking.type = enabled|disabled (matches the pinned Qwen 3.8 Max route).
+base_model = "alibaba/qwen3.8-max"
+name = "Qwen Max Latest (Qwen3.8 Max)"
+last_updated = "2026-09-16"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 262_144
+
+[modalities]
+input = ["text", "image"]
+
+[experimental.modes.priority]
+cost = { input = 3, output = 9, cache_read = 0.375 }
+provider = { body = { service_tier = "priority" } }
diff --git a/sync.md b/sync.md
index d2c8bc3520c..6bb7e47bb22 100644
--- a/sync.md
+++ b/sync.md
@@ -302,9 +302,9 @@ Fireworks AI is implemented in `packages/core/src/sync/providers/fireworks-ai.ts
 - Run it with `bun models:sync fireworks-ai` or `bun fireworks:sync`.
 - Source endpoint: `https://api.fireworks.ai/v1/serverless/models`; required auth: `FIREWORKS_API_KEY`.
 - The serverless catalog exposes one row per serving mode with live per-million-token pricing, invocation identifiers, aliases, advertised context length, and input/output modalities.
-- Standard rows update base-model IDs. Fast and other alternate-resource rows update the model or router named by `usage_identifier`; their aliases are tracked as additional invocation IDs. Flag-only modes such as Priority become priced `experimental.modes` with the required request-body `service_tier` instead of duplicate catalog IDs. A priority-only model remains discoverable with that service-tier recipe on its base ID.
+- Standard rows update base-model IDs. Fast and other alternate-resource rows update the model or router named by `usage_identifier`; their aliases are tracked as additional invocation IDs. Standard-path aliases inherit the base model's flag modes. Flag-only modes such as Priority become priced `experimental.modes` with the required request-body `service_tier` instead of duplicate catalog IDs. A priority-only model remains discoverable with that service-tier recipe on its base ID.
 - New text/vision invocation IDs are reported but not created automatically because the endpoint does not yet provide output limits, reasoning controls, tool support, or open-weight status. Embedding/reranking rows are ignored because the catalog provider entries describe generation models.
-- Pricing and modalities come from the matching serverless serving mode. Existing exact context caps smaller than the advertised API value are preserved, while a lower API ceiling is applied. Authored output limits, reasoning options, tool support, and other metadata not exposed by the endpoint are preserved.
+- Pricing and modalities come from the matching serverless serving mode; complete API pricing can seed an existing model whose authored TOML has no cost. Existing exact context caps smaller than the advertised API value are preserved, while a lower API ceiling is applied. Authored output limits, reasoning options, tool support, and other metadata not exposed by the endpoint are preserved.
 - Models absent from the serverless response are retained for manual lifecycle review.
 
 ## DigitalOcean Notes

From 6450610f61f156e67e6526a92f812406946a2647 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 06:41:38 +0000
Subject: [PATCH 1409/2012] chore(sync): update Kilo model catalog (#7216)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 providers/kilo/models/~moonshotai/kimi-latest.toml       | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..a43a707c98d 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 7c318d1ad2dd9c74789d3a199271ea4589882b2f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 06:41:41 +0000
Subject: [PATCH 1410/2012] chore(sync): update NanoGPT model catalog (#7218)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From a090d9528e40fb83f56e11c258f92630a6f0b638 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 06:41:44 +0000
Subject: [PATCH 1411/2012] chore(sync): update Requesty model catalog (#7215)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/deepseek-v4-pro-0813@eu.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/requesty/models/deepseek-v4-pro-0813@eu.toml b/providers/requesty/models/deepseek-v4-pro-0813@eu.toml
index 52382440332..d8e6ab21e5d 100644
--- a/providers/requesty/models/deepseek-v4-pro-0813@eu.toml
+++ b/providers/requesty/models/deepseek-v4-pro-0813@eu.toml
@@ -15,4 +15,4 @@ cache_read = 0.44
 
 [limit]
 context = 1_048_576
-output = 1_048_576
+output = 65_536

From 11c16e1999a65ccf3d36bbe961af12ae40db3322 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 06:41:47 +0000
Subject: [PATCH 1412/2012] chore(sync): update Venice model catalog (#7217)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-52-codex.toml    | 24 -------------------
 1 file changed, 24 deletions(-)
 delete mode 100644 providers/venice/models/openai-gpt-52-codex.toml

diff --git a/providers/venice/models/openai-gpt-52-codex.toml b/providers/venice/models/openai-gpt-52-codex.toml
deleted file mode 100644
index ff416e3bdaf..00000000000
--- a/providers/venice/models/openai-gpt-52-codex.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-# Live /models (2026-06-25): none|minimal|low|medium|high; no dedicated reasoning budget.
-# https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
-base_model = "openai/gpt-5.2-codex"
-family = "gpt"
-release_date = "2025-01-15"
-last_updated = "2026-06-11"
-knowledge = "2025-08"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
-
-[cost]
-input = 2.19
-output = 17.5
-cache_read = 0.219
-
-[limit]
-context = 256_000
-output = 65_536
-
-[modalities]
-input = ["text", "image"]

From 655a680e17b5dcff2d777d6d6dbbb62c220d043d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 06:41:51 +0000
Subject: [PATCH 1413/2012] chore(sync): update OpenRouter model catalog
 (#7219)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..93d4a9d7ab9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.96
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From d798e0ff3116ca928bc994fc73f733d16d5c4c79 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 07:30:27 +0000
Subject: [PATCH 1414/2012] chore(sync): update OpenRouter model catalog
 (#7221)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index d857fa53dab..3f78c99ff08 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,12 +12,12 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.13
-output = 0.52
+input = 0.15
+output = 0.6
 
 [limit]
 context = 262_144
-output = 32_768
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3e3ce108967..e9834460354 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.8392
+output = 3.3824
+cache_read = 0.1802
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From a93ece5eb559f0dc00892aa7d3386868e53e62d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 07:30:31 +0000
Subject: [PATCH 1415/2012] chore(sync): update Kilo model catalog (#7220)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml    |  4 ++--
 providers/kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index dee6c4b128f..83d4e34b47c 100644
--- a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 131_072
-output = 32_768
+context = 262_144
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 38aa9fed430..2bff4154345 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.8392
+output = 3.3824
+cache_read = 0.1802
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From e6856eff9aad176de55c53cb8e34167ed21772c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 08:34:55 +0000
Subject: [PATCH 1416/2012] chore(sync): update OpenRouter model catalog
 (#7224)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml        | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index e9834460354..3e3ce108967 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8392
-output = 3.3824
-cache_read = 0.1802
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From 86c6810dbea50bb46838accfb284567376f89054 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 08:34:59 +0000
Subject: [PATCH 1417/2012] chore(sync): update Cortecs model catalog (#7225)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/llama-3.3-70b-instruct.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/cortecs/models/llama-3.3-70b-instruct.toml b/providers/cortecs/models/llama-3.3-70b-instruct.toml
index 4b2b5f7f4d9..204e0ed29b9 100644
--- a/providers/cortecs/models/llama-3.3-70b-instruct.toml
+++ b/providers/cortecs/models/llama-3.3-70b-instruct.toml
@@ -2,8 +2,8 @@ base_model = "meta/llama-3.3-70b-instruct"
 structured_output = true
 
 [cost]
-input = 0.129
-output = 0.399
+input = 0.724
+output = 0.724
 
 [limit]
 context = 131_000

From 6f781d61a90d3a8c725ecea16a05092534d98a20 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 08:35:03 +0000
Subject: [PATCH 1418/2012] chore(sync): update Kilo model catalog (#7223)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 10 +++++-----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2bff4154345..38aa9fed430 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8392
-output = 3.3824
-cache_read = 0.1802
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 6d0de56f95ae30afe76dafd779241650ffcffcea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 09:28:51 +0000
Subject: [PATCH 1419/2012] chore(sync): update Kilo model catalog (#7229)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..d7bc1aba5f7 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 1.775
+output = 9.94
+cache_read = 0.2059
 
 [limit]
 context = 1_048_576

From 2e0ce128179958ba6adc09d8126db01f0c7aaf32 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 09:28:55 +0000
Subject: [PATCH 1420/2012] chore(sync): update OpenRouter model catalog
 (#7228)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..2f86131e1a2 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 1.775
+output = 9.94
+cache_read = 0.2059
 
 [limit]
 context = 1_048_576

From d8e1873e0a8ba8f7526dda234a7707aec995502a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 09:29:00 +0000
Subject: [PATCH 1421/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7230)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/Qwen3.8-27B.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway/models/Qwen3.8-27B.toml b/providers/llmgateway/models/Qwen3.8-27B.toml
index b3afd48e7ed..b25199e56d5 100644
--- a/providers/llmgateway/models/Qwen3.8-27B.toml
+++ b/providers/llmgateway/models/Qwen3.8-27B.toml
@@ -14,8 +14,8 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.2
-output = 2
+input = 0.08
+output = 0.35
 cache_read = 0.05
 
 [limit]

From f65073a13de7de21092a9ad4522dfcda9001f914 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 09:29:03 +0000
Subject: [PATCH 1422/2012] chore(sync): update LLM Gateway model catalog
 (#7227)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/Qwen3.8-27B.toml                 | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
index bbe9b9c35d1..c02ad50d9d8 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/Qwen3.8-27B.toml
@@ -14,8 +14,8 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.2
-output = 2
+input = 0.08
+output = 0.35
 cache_read = 0.05
 
 [limit]

From 8c24a9ac5e833a319136a4013e2d08fe096eee10 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 10:27:37 +0000
Subject: [PATCH 1423/2012] chore(sync): update OpenRouter model catalog
 (#7231)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 93d4a9d7ab9..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 2f86131e1a2..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.775
-output = 9.94
-cache_read = 0.2059
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From daa22fcd778207d1f7a143e0a9b84099e0b92e7c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 10:27:51 +0000
Subject: [PATCH 1424/2012] chore(sync): update Kilo model catalog (#7233)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 providers/kilo/models/~moonshotai/kimi-latest.toml       | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index a43a707c98d..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.96
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index d7bc1aba5f7..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.775
-output = 9.94
-cache_read = 0.2059
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 4f77a59abbacdcfa41cc2f4d06b3ed0323ab64c4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 10:27:53 +0000
Subject: [PATCH 1425/2012] chore(sync): update NanoGPT model catalog (#7232)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From c44504662756c747182a5f97d73f8e259ce58ebb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 11:25:35 +0000
Subject: [PATCH 1426/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7237)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-v4-1-flash.toml           | 28 -------------------
 1 file changed, 28 deletions(-)
 delete mode 100644 providers/llmgateway/models/deepseek-v4-1-flash.toml

diff --git a/providers/llmgateway/models/deepseek-v4-1-flash.toml b/providers/llmgateway/models/deepseek-v4-1-flash.toml
deleted file mode 100644
index 9b9286b5771..00000000000
--- a/providers/llmgateway/models/deepseek-v4-1-flash.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "DeepSeek-V4.1-Flash"
-description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-family = "deepseek"
-release_date = "2026-09-15"
-last_updated = "2026-09-15"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.01
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 5bb5a4ed326f2d266e782aae2092b03a4a78664b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 11:25:39 +0000
Subject: [PATCH 1427/2012] chore(sync): update Kilo model catalog (#7238)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From fc016496b410cedb6fd87d2b8a5c2fe0a0be6956 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 11:25:45 +0000
Subject: [PATCH 1428/2012] chore(sync): update OpenRouter model catalog
 (#7240)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml      | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index 96c5bbb7f77..f82829f96ea 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.3333
-cache_read = 0.02
+input = 0.09
+output = 0.3
+cache_read = 0.018
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 495f9992234ce24ebb61cf6dc65dda4bd5a6114a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 11:25:52 +0000
Subject: [PATCH 1429/2012] chore(sync): update LLM Gateway model catalog
 (#7239)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/runware/deepseek-v4-1-flash.toml   | 28 -------------------
 1 file changed, 28 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml

diff --git a/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml b/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml
deleted file mode 100644
index 1d71ee3bb77..00000000000
--- a/providers/llmgateway-providers/models/runware/deepseek-v4-1-flash.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-name = "DeepSeek-V4.1-Flash (Runware)"
-description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-family = "deepseek"
-release_date = "2026-09-15"
-last_updated = "2026-09-15"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.01
-
-[limit]
-context = 1_048_576
-output = 1_048_576
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 1f2ea60debac196e812d475243821e1f47d61eca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 12:37:32 +0000
Subject: [PATCH 1430/2012] chore(sync): update Kilo model catalog (#7241)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/gryphe/mythomax-l2-13b.toml    | 4 ++--
 providers/kilo/models/thedrummer/unslopnemo-12b.toml | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/gryphe/mythomax-l2-13b.toml b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
index a7751171cf6..9cb22a7518d 100644
--- a/providers/kilo/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/kilo/models/gryphe/mythomax-l2-13b.toml
@@ -10,8 +10,8 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.06
-output = 0.06
+input = 0.08
+output = 0.11
 
 [limit]
 context = 4_096
diff --git a/providers/kilo/models/thedrummer/unslopnemo-12b.toml b/providers/kilo/models/thedrummer/unslopnemo-12b.toml
index f1b04078f03..8348cb4c6d4 100644
--- a/providers/kilo/models/thedrummer/unslopnemo-12b.toml
+++ b/providers/kilo/models/thedrummer/unslopnemo-12b.toml
@@ -5,7 +5,7 @@ last_updated = "2024-11-08"
 attachment = false
 reasoning = false
 temperature = true
-tool_call = true
+tool_call = false
 structured_output = true
 open_weights = false
 

From 4f8cdcd4ea5a988f5ed9be083441be8bfa5a0be4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 12:37:39 +0000
Subject: [PATCH 1431/2012] chore(sync): update OpenRouter model catalog
 (#7242)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/gryphe/mythomax-l2-13b.toml    | 4 ++--
 providers/openrouter/models/thedrummer/unslopnemo-12b.toml | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
index 0ddba1dbce0..2028e3c85b6 100644
--- a/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
+++ b/providers/openrouter/models/gryphe/mythomax-l2-13b.toml
@@ -11,8 +11,8 @@ knowledge = "2023-06-30"
 open_weights = true
 
 [cost]
-input = 0.06
-output = 0.06
+input = 0.08
+output = 0.11
 
 [limit]
 context = 8_192
diff --git a/providers/openrouter/models/thedrummer/unslopnemo-12b.toml b/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
index 2fee039a839..2b733ccb0cb 100644
--- a/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
+++ b/providers/openrouter/models/thedrummer/unslopnemo-12b.toml
@@ -5,7 +5,7 @@ last_updated = "2024-11-08"
 attachment = false
 reasoning = false
 temperature = true
-tool_call = true
+tool_call = false
 structured_output = true
 knowledge = "2024-04-30"
 open_weights = true

From 57691ec9f571d14ce5cc9bdd4c80589898c59b88 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 13:27:16 +0000
Subject: [PATCH 1432/2012] chore(sync): update OpenRouter model catalog
 (#7245)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml   | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 3f78c99ff08..d857fa53dab 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,12 +12,12 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.15
-output = 0.6
+input = 0.13
+output = 0.52
 
 [limit]
 context = 262_144
-output = 16_384
+output = 32_768
 
 [modalities]
 input = ["text", "image"]

From c6580f62577b2897c8ba90ac6437d8247e1e7944 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 13:27:18 +0000
Subject: [PATCH 1433/2012] chore(sync): update Chutes model catalog (#7244)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/chutes/models/moonshotai/Kimi-K2.6-TEE.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/chutes/models/moonshotai/Kimi-K2.6-TEE.toml b/providers/chutes/models/moonshotai/Kimi-K2.6-TEE.toml
index 5ed8b82d4a2..5ffbf637959 100644
--- a/providers/chutes/models/moonshotai/Kimi-K2.6-TEE.toml
+++ b/providers/chutes/models/moonshotai/Kimi-K2.6-TEE.toml
@@ -12,9 +12,9 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 0.58
-output = 3.4
-cache_read = 0.05799999999999998
+input = 0.5
+output = 2.85
+cache_read = 0.04999999999999999
 
 [limit]
 output = 65_535

From c4ee85f0b2ea43f68c8ab1112f5de155368aee45 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 13:27:22 +0000
Subject: [PATCH 1434/2012] chore(sync): update Kilo model catalog (#7246)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 83d4e34b47c..dee6c4b128f 100644
--- a/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/kilo/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -15,8 +15,8 @@ input = 0.13
 output = 0.52
 
 [limit]
-context = 262_144
-output = 16_384
+context = 131_072
+output = 32_768
 
 [modalities]
 input = ["text", "image"]

From fb8c1eb9b9f7e329ea0ac247a076eca52a3b75a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 14:28:45 +0000
Subject: [PATCH 1435/2012] chore(sync): update Cortecs model catalog (#7248)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/deepseek-v4-flash-0731.toml | 6 +++---
 providers/cortecs/models/glm-5.2.toml                | 6 +++---
 providers/cortecs/models/kimi-k2.6.toml              | 6 +++---
 providers/cortecs/models/kimi-k2.7-code.toml         | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/cortecs/models/deepseek-v4-flash-0731.toml b/providers/cortecs/models/deepseek-v4-flash-0731.toml
index a6037a9e2fe..c6b4f299c71 100644
--- a/providers/cortecs/models/deepseek-v4-flash-0731.toml
+++ b/providers/cortecs/models/deepseek-v4-flash-0731.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.09
-output = 0.17
-cache_read = 0.014
+input = 0.055
+output = 0.174
+cache_read = 0.009
 
 [limit]
 context = 1_048_576
diff --git a/providers/cortecs/models/glm-5.2.toml b/providers/cortecs/models/glm-5.2.toml
index 984c3e96d79..c18a497ae1e 100644
--- a/providers/cortecs/models/glm-5.2.toml
+++ b/providers/cortecs/models/glm-5.2.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 1.2
-output = 4.2
-cache_read = 0.26
+input = 1.1
+output = 2.99
+cache_read = 0.18
 
 [limit]
 context = 1_048_576
diff --git a/providers/cortecs/models/kimi-k2.6.toml b/providers/cortecs/models/kimi-k2.6.toml
index 6d220564edf..559856afd3a 100644
--- a/providers/cortecs/models/kimi-k2.6.toml
+++ b/providers/cortecs/models/kimi-k2.6.toml
@@ -8,9 +8,9 @@ field = "reasoning_content"
 type = "toggle"
 
 [cost]
-input = 0.773
-output = 3.38
-cache_read = 0.193
+input = 0.516
+output = 2.869
+cache_read = 0.115
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/cortecs/models/kimi-k2.7-code.toml b/providers/cortecs/models/kimi-k2.7-code.toml
index 30590475cd6..7d104ad2503 100644
--- a/providers/cortecs/models/kimi-k2.7-code.toml
+++ b/providers/cortecs/models/kimi-k2.7-code.toml
@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.75
-output = 3.5
-cache_read = 0.201
+input = 0.706
+output = 3.208
+cache_read = 0.18
 
 [modalities]
 input = ["text", "image", "pdf"]

From fe61deb412030455b6983ad2d982d004fb730b7a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 14:28:56 +0000
Subject: [PATCH 1436/2012] chore(sync): update Chutes model catalog (#7247)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml b/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
index ad2277c0b87..f44b33e43ce 100644
--- a/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
+++ b/providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml
@@ -6,9 +6,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.32
-output = 2.5
-cache_read = 0.031999999999999994
+input = 0.24
+output = 2.2
+cache_read = 0.023999999999999994
 
 [limit]
 output = 65_536

From 9a02962417cc505e96845c3ec750919155f559ec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 14:29:01 +0000
Subject: [PATCH 1437/2012] chore(sync): update Kilo model catalog (#7250)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 ++-
 .../kilo/models/~deepseek/deepseek-flash-latest.toml   |  4 ++--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 673b730eb39..f491c2853c8 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_000_000
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..bd4537b7471 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
-context = 1_048_576
-output = 384_000
+context = 1_000_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From cfc1c187d86d1fd1dada1bf13c21fa82cae9edfb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 14:29:29 +0000
Subject: [PATCH 1438/2012] chore(sync): update OpenRouter model catalog
 (#7249)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 4 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 451e7c3012c..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9834
-output = 2.9502
-cache_read = 0.03278
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index dc34ba0ccf7..98fd23a037b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..e0b3ef90d2e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From e3626b49388e2dbea382b91b81e9d4237c117402 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Wed, 16 Sep 2026 22:44:00 +0800
Subject: [PATCH 1439/2012] feat(opencode): add Union Alpha

---
 providers/opencode-go/models/union-alpha.toml | 25 +++++++++++++++++++
 providers/opencode/models/union-alpha.toml    | 25 +++++++++++++++++++
 2 files changed, 50 insertions(+)
 create mode 100644 providers/opencode-go/models/union-alpha.toml
 create mode 100644 providers/opencode/models/union-alpha.toml

diff --git a/providers/opencode-go/models/union-alpha.toml b/providers/opencode-go/models/union-alpha.toml
new file mode 100644
index 00000000000..19374530637
--- /dev/null
+++ b/providers/opencode-go/models/union-alpha.toml
@@ -0,0 +1,25 @@
+name = "Union Alpha"
+description = "Stealth model built for agentic coding"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = true
+reasoning_options = []
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file
diff --git a/providers/opencode/models/union-alpha.toml b/providers/opencode/models/union-alpha.toml
new file mode 100644
index 00000000000..19374530637
--- /dev/null
+++ b/providers/opencode/models/union-alpha.toml
@@ -0,0 +1,25 @@
+name = "Union Alpha"
+description = "Stealth model built for agentic coding"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = true
+reasoning_options = []
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file

From 75518776b9fb7e51d0de48b83469f6a6b4159415 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Wed, 16 Sep 2026 23:08:55 +0800
Subject: [PATCH 1440/2012] fix(opencode): label Union Alpha as free

---
 providers/opencode-go/models/union-alpha.toml | 2 +-
 providers/opencode/models/union-alpha.toml    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/opencode-go/models/union-alpha.toml b/providers/opencode-go/models/union-alpha.toml
index 19374530637..25794670176 100644
--- a/providers/opencode-go/models/union-alpha.toml
+++ b/providers/opencode-go/models/union-alpha.toml
@@ -1,4 +1,4 @@
-name = "Union Alpha"
+name = "Union Alpha Free"
 description = "Stealth model built for agentic coding"
 release_date = "2026-09-16"
 last_updated = "2026-09-16"
diff --git a/providers/opencode/models/union-alpha.toml b/providers/opencode/models/union-alpha.toml
index 19374530637..25794670176 100644
--- a/providers/opencode/models/union-alpha.toml
+++ b/providers/opencode/models/union-alpha.toml
@@ -1,4 +1,4 @@
-name = "Union Alpha"
+name = "Union Alpha Free"
 description = "Stealth model built for agentic coding"
 release_date = "2026-09-16"
 last_updated = "2026-09-16"

From 2e1b8171f0a93d81f40941185b25baebfd6799d0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 15:25:33 +0000
Subject: [PATCH 1441/2012] chore(sync): update OpenRouter model catalog
 (#7251)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++---
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    |  8 +++----
 .../models/stealth/union-alpha.toml           | 23 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++---
 4 files changed, 33 insertions(+), 10 deletions(-)
 create mode 100644 providers/openrouter/models/stealth/union-alpha.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520
diff --git a/providers/openrouter/models/stealth/union-alpha.toml b/providers/openrouter/models/stealth/union-alpha.toml
new file mode 100644
index 00000000000..fed182b70f6
--- /dev/null
+++ b/providers/openrouter/models/stealth/union-alpha.toml
@@ -0,0 +1,23 @@
+name = "Union Alpha"
+description = "Multimodal model for analyzing text, images, documents, and rich media"
+family = "alpha"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e0b3ef90d2e..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From be29e6beae3cfd57fcb1688ec46ce5cc22c6ade4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 15:25:40 +0000
Subject: [PATCH 1442/2012] chore(sync): update Eden AI model catalog (#7252)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index c540731d87a..13d977a7fd0 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34617
-output = 0.46156
+input = 0.34611
+output = 0.46148
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index d11cc1ac90a..4032a872079 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.750035
-output = 0.750035
+input = 0.749905
+output = 0.749905
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 26c9489ce11..2d3e6ad6c1a 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173085
-output = 0.750035
+input = 0.173055
+output = 0.749905
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index c7d9a5ffc2c..9b9acde4f4d 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46156
-output = 0.92312
+input = 0.46148
+output = 0.92296
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 42cb87b432b..6e9e91ebf7e 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173085
-output = 0.69234
+input = 0.173055
+output = 0.69222
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index f2cdbc67c92..6baa7372a71 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.03851
-output = 1.03851
+input = 1.038331
+output = 1.038331

From 11c86c5ef907f20e4923efb4b76d40137de6637b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 15:25:43 +0000
Subject: [PATCH 1443/2012] chore(sync): update Kilo model catalog (#7253)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nvidia/nemotron-3-ultra-550b-a55b.toml    |  4 ++--
 .../kilo/models/stealth/union-alpha.toml      | 23 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  8 +++----
 3 files changed, 29 insertions(+), 6 deletions(-)
 create mode 100644 providers/kilo/models/stealth/union-alpha.toml

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520
diff --git a/providers/kilo/models/stealth/union-alpha.toml b/providers/kilo/models/stealth/union-alpha.toml
new file mode 100644
index 00000000000..cc5ad3f5672
--- /dev/null
+++ b/providers/kilo/models/stealth/union-alpha.toml
@@ -0,0 +1,23 @@
+name = "Union Alpha"
+description = "Union Alpha is a multimodal model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks. Union Alpha is a stealth model...."
+family = "alpha"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index bd4537b7471..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 393_216
 
 [modalities]

From 76571c179465c0dd3e3adafb7f41e1a5b33598e3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 15:26:34 +0000
Subject: [PATCH 1444/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7254)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/kwaipilot/kat-coder-air-v2.5.toml  | 24 ------------------
 .../models/kwaipilot/kat-coder-pro-v1.toml    | 25 -------------------
 .../models/kwaipilot/kat-coder-pro-v2.5.toml  | 24 ------------------
 .../models/kwaipilot/kat-coder-pro-v2.toml    | 24 ------------------
 .../vercel/models/moonshotai/kimi-k2.5.toml   |  5 ++--
 .../vercel/models/openai/gpt-oss-20b.toml     |  5 ++--
 6 files changed, 4 insertions(+), 103 deletions(-)
 delete mode 100644 providers/vercel/models/kwaipilot/kat-coder-air-v2.5.toml
 delete mode 100644 providers/vercel/models/kwaipilot/kat-coder-pro-v1.toml
 delete mode 100644 providers/vercel/models/kwaipilot/kat-coder-pro-v2.5.toml
 delete mode 100644 providers/vercel/models/kwaipilot/kat-coder-pro-v2.toml

diff --git a/providers/vercel/models/kwaipilot/kat-coder-air-v2.5.toml b/providers/vercel/models/kwaipilot/kat-coder-air-v2.5.toml
deleted file mode 100644
index 0ea0c30f8c9..00000000000
--- a/providers/vercel/models/kwaipilot/kat-coder-air-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kat Coder Air V2.5"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-07-10"
-last_updated = "2026-07-10"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.15
-output = 0.6
-cache_read = 0.03
-
-[limit]
-context = 256_000
-output = 80_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/kwaipilot/kat-coder-pro-v1.toml b/providers/vercel/models/kwaipilot/kat-coder-pro-v1.toml
deleted file mode 100644
index 103b91326ad..00000000000
--- a/providers/vercel/models/kwaipilot/kat-coder-pro-v1.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "KAT-Coder-Pro V1"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2025-11-09"
-last_updated = "2025-10-24"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = false
-knowledge = "2024-10"
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.06
-
-[limit]
-context = 256_000
-output = 32_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/vercel/models/kwaipilot/kat-coder-pro-v2.5.toml b/providers/vercel/models/kwaipilot/kat-coder-pro-v2.5.toml
deleted file mode 100644
index f3a7b4d82b5..00000000000
--- a/providers/vercel/models/kwaipilot/kat-coder-pro-v2.5.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kat Coder Pro V2.5"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-07-10"
-last_updated = "2026-07-10"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = false
-reasoning_options = []
-
-[cost]
-input = 0.74
-output = 2.96
-cache_read = 0.15
-
-[limit]
-context = 256_000
-output = 80_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/kwaipilot/kat-coder-pro-v2.toml b/providers/vercel/models/kwaipilot/kat-coder-pro-v2.toml
deleted file mode 100644
index 7c1813d40df..00000000000
--- a/providers/vercel/models/kwaipilot/kat-coder-pro-v2.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kat Coder Pro V2"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-attachment = false
-reasoning = true
-reasoning_options = []
-tool_call = true
-temperature = true
-release_date = "2026-03-27"
-last_updated = "2026-03-30"
-open_weights = false
-
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.06
-
-[limit]
-context = 256_000
-output = 256_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/vercel/models/moonshotai/kimi-k2.5.toml b/providers/vercel/models/moonshotai/kimi-k2.5.toml
index e58211e60c2..3cceed8535b 100644
--- a/providers/vercel/models/moonshotai/kimi-k2.5.toml
+++ b/providers/vercel/models/moonshotai/kimi-k2.5.toml
@@ -9,11 +9,10 @@ type = "toggle"
 [cost]
 input = 0.6
 output = 3
-cache_read = 0.1
 
 [limit]
-context = 262_114
-output = 262_114
+context = 256_000
+output = 256_000
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/vercel/models/openai/gpt-oss-20b.toml b/providers/vercel/models/openai/gpt-oss-20b.toml
index a0c15d2d037..df18b0270e5 100644
--- a/providers/vercel/models/openai/gpt-oss-20b.toml
+++ b/providers/vercel/models/openai/gpt-oss-20b.toml
@@ -1,14 +1,13 @@
 base_model = "openai/gpt-oss-20b"
 description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-knowledge = "2024-10"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.05
-output = 0.2
+input = 0.03
+output = 0.14
 
 [limit]
 input = 122_880

From ee33f656357a9567ec38cb425814cd9da929df38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 16:28:04 +0000
Subject: [PATCH 1445/2012] chore(sync): update LLM Gateway model catalog
 (#7257)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/consensusprotocol/deepseek-v4.1-flash.toml      | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
index ad9191cc116..3ebe60d0b7b 100644
--- a/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
+++ b/providers/llmgateway-providers/models/consensusprotocol/deepseek-v4.1-flash.toml
@@ -1,12 +1,10 @@
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4.1 Flash (Consensus Protocol)"
-attachment = false
-tool_call = false
 structured_output = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high", "max"]
+values = ["low", "high", "xhigh", "max"]
 
 [cost]
 input = 0.2
@@ -15,6 +13,3 @@ cache_read = 0.005
 
 [limit]
 context = 1_048_576
-
-[modalities]
-input = ["text"]

From ba5433dcd4387f3be4cbd02867948eb628a85dea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 16:28:08 +0000
Subject: [PATCH 1446/2012] chore(sync): update Kilo model catalog (#7256)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml      | 4 ++--
 providers/kilo/models/tencent/hy3.toml                      | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 25c72a0b0fb..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 202_800
-output = 182_520
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 0448853c4e3a7cf4cf42dac5b9205cf8c6e13170 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 16:28:26 +0000
Subject: [PATCH 1447/2012] chore(sync): update OpenRouter model catalog
 (#7258)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 5dcd117448e..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 182_520
+output = 32_768
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 0b50f749bae37b92a8a25fc04740c37ef105c311 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 17:23:13 +0000
Subject: [PATCH 1448/2012] chore(sync): update Charm Hyper model catalog
 (#7260)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/glm-5.3.toml      | 1 -
 providers/hyper/models/glm-5.toml        | 6 +++---
 providers/hyper/models/minimax-m2.7.toml | 6 +++---
 3 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/hyper/models/glm-5.3.toml b/providers/hyper/models/glm-5.3.toml
index bdd46e1420e..271d0fbd971 100644
--- a/providers/hyper/models/glm-5.3.toml
+++ b/providers/hyper/models/glm-5.3.toml
@@ -12,5 +12,4 @@ output = 4.79072
 cache_read = 0.283088
 
 [limit]
-context = 1_048_576
 output = 128_000
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
index 9882a80de94..8706b519485 100644
--- a/providers/hyper/models/glm-5.toml
+++ b/providers/hyper/models/glm-5.toml
@@ -4,9 +4,9 @@ last_updated = "2026-08-16"
 reasoning_options = []
 
 [cost]
-input = 0.86
-output = 2.752
-cache_read = 0.43
+input = 0.93
+output = 2.878
+cache_read = 0.465
 
 [limit]
 context = 202_752
diff --git a/providers/hyper/models/minimax-m2.7.toml b/providers/hyper/models/minimax-m2.7.toml
index 701182a4ee1..13ca6a02401 100644
--- a/providers/hyper/models/minimax-m2.7.toml
+++ b/providers/hyper/models/minimax-m2.7.toml
@@ -4,9 +4,9 @@ last_updated = "2026-07-22"
 reasoning_options = []
 
 [cost]
-input = 0.418
-output = 1.588
-cache_read = 0.209
+input = 0.484
+output = 1.852
+cache_read = 0.242
 
 [limit]
 context = 262_100

From ce7efe2403a00e996f5080361fa0ae9aab9cf4d8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 17:23:34 +0000
Subject: [PATCH 1449/2012] chore(sync): update OpenRouter model catalog
 (#7261)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml               | 1 -
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index cf2eefb64f7..557dd11171f 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -19,4 +19,3 @@ cache_read = 0.14
 
 [limit]
 context = 1_048_576
-output = 128_000

From d4a669e22800406fd525b402931ccbfbd9e4f79c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 17:23:38 +0000
Subject: [PATCH 1450/2012] chore(sync): update Kilo model catalog (#7259)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.2.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index 80b000693ec..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -12,4 +12,3 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 128_000

From 2566eb0c6c71820e46d64aaba6da65ddee16f60c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 18:30:32 +0000
Subject: [PATCH 1451/2012] chore(sync): update OpenRouter model catalog
 (#7264)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..c738646e281 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.084
+output = 0.168
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 3a0c7f47171..c11eb6c3dae 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.648138
-output = 13.282724
-cache_read = 0.302644
+input = 3
+output = 15
+cache_read = 0.3
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 2a5d7af688de8c6c5675e77e4605b33f1df7196f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 18:30:34 +0000
Subject: [PATCH 1452/2012] chore(sync): update Kilo model catalog (#7263)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 6123b85d00d2f7f22cd7d1dd68a56a2de840bcdf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 18:30:37 +0000
Subject: [PATCH 1453/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7266)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/alibaba/qwen3-next-80b-a3b-instruct.toml    |  4 ++++
 .../models/alibaba/qwen3-next-80b-a3b-thinking.toml    |  4 ++++
 providers/vercel/models/alibaba/qwen3.8-max.toml       |  2 +-
 .../vercel/models/deepseek/deepseek-v4.1-flash.toml    | 10 +++++++---
 4 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/providers/vercel/models/alibaba/qwen3-next-80b-a3b-instruct.toml b/providers/vercel/models/alibaba/qwen3-next-80b-a3b-instruct.toml
index 30e2796a494..06998ee46a3 100644
--- a/providers/vercel/models/alibaba/qwen3-next-80b-a3b-instruct.toml
+++ b/providers/vercel/models/alibaba/qwen3-next-80b-a3b-instruct.toml
@@ -4,3 +4,7 @@ name = "Qwen3 Next 80B A3B Instruct"
 [cost]
 input = 0.15
 output = 1.2
+
+[limit]
+context = 262_114
+output = 262_114
diff --git a/providers/vercel/models/alibaba/qwen3-next-80b-a3b-thinking.toml b/providers/vercel/models/alibaba/qwen3-next-80b-a3b-thinking.toml
index f5c85876808..f01e76acd81 100644
--- a/providers/vercel/models/alibaba/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/vercel/models/alibaba/qwen3-next-80b-a3b-thinking.toml
@@ -8,3 +8,7 @@ min = 1
 [cost]
 input = 0.15
 output = 1.2
+
+[limit]
+context = 262_144
+output = 262_144
diff --git a/providers/vercel/models/alibaba/qwen3.8-max.toml b/providers/vercel/models/alibaba/qwen3.8-max.toml
index 4e6f5eac650..d73681007ac 100644
--- a/providers/vercel/models/alibaba/qwen3.8-max.toml
+++ b/providers/vercel/models/alibaba/qwen3.8-max.toml
@@ -14,9 +14,9 @@ max = 262_144
 input = 2
 output = 6
 cache_read = 0.25
-cache_write = 2.5
 
 [limit]
+context = 262_144
 output = 128_000
 
 [modalities]
diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
index c2b9c23b86c..c185e36af68 100644
--- a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
@@ -10,6 +10,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.3
+output = 1.2
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 32_768

From 552dfaa9ff525354c51f0dd702f5cbdbc3e2b91b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 18:30:41 +0000
Subject: [PATCH 1454/2012] chore(sync): update DigitalOcean model catalog
 (#7265)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/deepseek-v4.1-flash.toml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/providers/digitalocean/models/deepseek-v4.1-flash.toml b/providers/digitalocean/models/deepseek-v4.1-flash.toml
index 736d7e55f2d..31780863b57 100644
--- a/providers/digitalocean/models/deepseek-v4.1-flash.toml
+++ b/providers/digitalocean/models/deepseek-v4.1-flash.toml
@@ -1,5 +1,4 @@
 base_model = "deepseek/deepseek-v4.1-flash"
-attachment = false
 
 [[reasoning_options]]
 type = "effort"
@@ -13,6 +12,3 @@ cache_read = 0.006
 [limit]
 context = 1_048_576
 output = 1_048_576
-
-[modalities]
-input = ["text"]

From a475c0c61ddb73b60e9a005a4d5009638d139e88 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:23:34 +0000
Subject: [PATCH 1455/2012] chore(sync): update LLM Gateway model catalog
 (#7269)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/runware/deepseek-v4.1-flash.toml       | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/runware/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/runware/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/runware/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..98d7ceb2faf
--- /dev/null
+++ b/providers/llmgateway-providers/models/runware/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Runware)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.01
+
+[limit]
+context = 1_048_576

From c02a73cab26a579e0e313e2888afdfc930de125c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:23:36 +0000
Subject: [PATCH 1456/2012] chore(sync): update Kilo model catalog (#7268)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From d57e907ff8bba7f25e542f47a6a64ea1189543cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:23:46 +0000
Subject: [PATCH 1457/2012] chore(sync): update OpenRouter model catalog
 (#7271)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c738646e281..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.084
-output = 0.168
-cache_read = 0.0168
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 9081cc2f85899a6de0cbfa4bf1ddc13c75103a20 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 20:26:56 +0000
Subject: [PATCH 1458/2012] chore(sync): update Kilo model catalog (#7274)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520

From 47d7e7f8d97cf3212c81858061bdf0d9a6c1ed92 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 20:27:15 +0000
Subject: [PATCH 1459/2012] chore(sync): update OpenRouter model catalog
 (#7273)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4a73a16375..3568f900316 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08246
-output = 0.16492
-cache_read = 0.016492
+input = 0.08092
+output = 0.16184
+cache_read = 0.016184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520

From 7c7d8f5649be213481ee7a7a844ed076d2fd3a3b Mon Sep 17 00:00:00 2001
From: h3nnes <henrik+github@hnrk.io>
Date: Wed, 16 Sep 2026 23:16:27 +0200
Subject: [PATCH 1460/2012] feat(synthetic): add DeepSeek V4.1 Flash (#7272)

* feat(synthetic): add DeepSeek V4.1 Flash

* fix(synthetic): drop inherited modalities from DeepSeek V4.1 Flash
---
 .../hf:deepseek-ai/DeepSeek-V4.1-Flash.toml   | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/synthetic/models/hf:deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/synthetic/models/hf:deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/synthetic/models/hf:deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..1a610b722d0
--- /dev/null
+++ b/providers/synthetic/models/hf:deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,19 @@
+# GET https://api.synthetic.new/openai/v1/models (accessed 2026-09-16) lists
+# hf:deepseek-ai/DeepSeek-V4.1-Flash with reasoning_parameters.efforts
+# none|low|high|xhigh|max, text+image input, fp8 quantization.
+# Live probes: `reasoning_effort = "none"` disables reasoning; `"xhigh"` streams it.
+# Reasoning is returned in the `reasoning` message field.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "xhigh", "max"] }]
+
+interleaved = true
+
+[cost]
+input = 0.6
+output = 1.2
+cache_read = 0.03
+
+[limit]
+context = 524_288
+output = 65_536

From dc8ab4b8804861b8e974121dcf85a12b60d55f26 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 21:25:01 +0000
Subject: [PATCH 1461/2012] chore(sync): update Kilo model catalog (#7276)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 38aa9fed430..2fc8e4f5ea7 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.9
+output = 3
+cache_read = 0.15
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_000_000
+output = 131_072
 
 [modalities]
 input = ["text"]

From e6a2c00a3270e3defb10cef56f666905f00090e1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 21:25:04 +0000
Subject: [PATCH 1462/2012] chore(sync): update OpenRouter model catalog
 (#7277)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3568f900316..6cab31726d4 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08092
-output = 0.16184
-cache_read = 0.016184
+input = 0.07938
+output = 0.15876
+cache_read = 0.015876
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3e3ce108967..0a7b3acfd67 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.9
+output = 3
+cache_read = 0.15
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 131_072
 
 [modalities]
 input = ["text"]

From 1855190cc0ad6f8343bacaad99e2eae46330177f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 22:24:32 +0000
Subject: [PATCH 1463/2012] chore(sync): update OpenRouter model catalog
 (#7278)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6cab31726d4..1a4b0569196 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07938
-output = 0.15876
-cache_read = 0.015876
+input = 0.07784
+output = 0.15568
+cache_read = 0.015568
 
 [limit]
 context = 1_048_576

From 003be8312c0e34372b21e9d7e741448add430c4b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 23:24:10 +0000
Subject: [PATCH 1464/2012] chore(sync): update OpenRouter model catalog
 (#7279)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml       | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 5 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 1a4b0569196..f7dc249d032 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07784
-output = 0.15568
-cache_read = 0.015568
+input = 0.0763
+output = 0.1526
+cache_read = 0.01526
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index f82829f96ea..4dac1f2fc80 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.09
-output = 0.3
-cache_read = 0.018
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index c32fca67d4b..644122b25af 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0a7b3acfd67..9d68ba72961 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.833
+output = 2.618
+cache_read = 0.13685
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 2bf0defa9cb7643694adc594e9782ce791e89920 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 23:24:15 +0000
Subject: [PATCH 1465/2012] chore(sync): update Kilo model catalog (#7280)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml |  6 +++---
 providers/kilo/models/~z-ai/glm-flash-latest.toml  |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 10 +++++-----
 3 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index e4ce4e5dfe6..5d5fb4fce66 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.07
+output = 0.2333
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2fc8e4f5ea7..7effc6dceef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.833
+output = 2.618
+cache_read = 0.13685
 
 [limit]
-context = 1_000_000
-output = 131_072
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From f81c6e1cf70a84eb20eb8550a70bea02359af12a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:07:11 -0500
Subject: [PATCH 1466/2012] chore(sync): update Charm Hyper model catalog
 (#7270)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../hyper/models/deepseek-v4.1-flash.toml     |  6 +++---
 providers/hyper/models/glm-5.1.toml           | 13 -------------
 providers/hyper/models/glm-5.toml             | 13 -------------
 providers/hyper/models/kimi-k2.5.toml         | 16 ----------------
 providers/hyper/models/kimi-k2.6.toml         | 19 -------------------
 .../hyper/models/llama-3.3-70b-instruct.toml  | 11 -----------
 ...lama-4-maverick-17b-128e-instruct-fp8.toml | 16 ----------------
 ...oder-480b-a35b-instruct-int4-mixed-ar.toml | 12 ------------
 .../models/qwen3-next-80b-a3b-instruct.toml   | 12 ------------
 providers/hyper/models/qwen3.6-flash.toml     | 19 -------------------
 providers/hyper/models/qwen3.6-max.toml       | 17 -----------------
 providers/hyper/models/qwen3.6-plus.toml      | 19 -------------------
 12 files changed, 3 insertions(+), 170 deletions(-)
 delete mode 100644 providers/hyper/models/glm-5.1.toml
 delete mode 100644 providers/hyper/models/glm-5.toml
 delete mode 100644 providers/hyper/models/kimi-k2.5.toml
 delete mode 100644 providers/hyper/models/kimi-k2.6.toml
 delete mode 100644 providers/hyper/models/llama-3.3-70b-instruct.toml
 delete mode 100644 providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
 delete mode 100644 providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
 delete mode 100644 providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
 delete mode 100644 providers/hyper/models/qwen3.6-flash.toml
 delete mode 100644 providers/hyper/models/qwen3.6-max.toml
 delete mode 100644 providers/hyper/models/qwen3.6-plus.toml

diff --git a/providers/hyper/models/deepseek-v4.1-flash.toml b/providers/hyper/models/deepseek-v4.1-flash.toml
index e9180c12434..c703c4fcf0c 100644
--- a/providers/hyper/models/deepseek-v4.1-flash.toml
+++ b/providers/hyper/models/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.32664
-output = 1.30656
-cache_read = 0.032664
+input = 0.3
+output = 1.2
+cache_read = 0.03
 
 [limit]
 output = 32_768
diff --git a/providers/hyper/models/glm-5.1.toml b/providers/hyper/models/glm-5.1.toml
deleted file mode 100644
index ea8b1d2c372..00000000000
--- a/providers/hyper/models/glm-5.1.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "zhipuai/glm-5.1"
-release_date = "2026-06-04"
-last_updated = "2026-07-22"
-reasoning_options = []
-
-[cost]
-input = 1.358
-output = 4.268
-cache_read = 0.679
-
-[limit]
-context = 202_750
-output = 3_276
diff --git a/providers/hyper/models/glm-5.toml b/providers/hyper/models/glm-5.toml
deleted file mode 100644
index 8706b519485..00000000000
--- a/providers/hyper/models/glm-5.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "zhipuai/glm-5"
-release_date = "2026-04-13"
-last_updated = "2026-08-16"
-reasoning_options = []
-
-[cost]
-input = 0.93
-output = 2.878
-cache_read = 0.465
-
-[limit]
-context = 202_752
-output = 20_275
diff --git a/providers/hyper/models/kimi-k2.5.toml b/providers/hyper/models/kimi-k2.5.toml
deleted file mode 100644
index 59545971994..00000000000
--- a/providers/hyper/models/kimi-k2.5.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "moonshotai/kimi-k2.5"
-release_date = "2026-04-13"
-last_updated = "2026-07-22"
-attachment = false
-reasoning_options = []
-
-[cost]
-input = 0.5584
-output = 2.935
-cache_read = 0.2792
-
-[limit]
-output = 26_214
-
-[modalities]
-input = ["text"]
diff --git a/providers/hyper/models/kimi-k2.6.toml b/providers/hyper/models/kimi-k2.6.toml
deleted file mode 100644
index c51aa337b1d..00000000000
--- a/providers/hyper/models/kimi-k2.6.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "moonshotai/kimi-k2.6"
-release_date = "2026-07-03"
-last_updated = "2026-07-22"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 1.03436
-output = 4.3552
-cache_read = 0.174208
-
-[limit]
-context = 262_000
-output = 26_214
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/hyper/models/llama-3.3-70b-instruct.toml b/providers/hyper/models/llama-3.3-70b-instruct.toml
deleted file mode 100644
index 96943440fab..00000000000
--- a/providers/hyper/models/llama-3.3-70b-instruct.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "meta/llama-3.3-70b-instruct"
-release_date = "2026-04-30"
-last_updated = "2026-07-22"
-
-[cost]
-input = 0.6066
-output = 1.0386
-cache_read = 0.3033
-
-[limit]
-output = 12_800
diff --git a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
deleted file mode 100644
index b8610bed8df..00000000000
--- a/providers/hyper/models/llama-4-maverick-17b-128e-instruct-fp8.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "meta/llama-4-maverick-17b-instruct"
-release_date = "2026-04-30"
-last_updated = "2026-07-22"
-attachment = false
-
-[cost]
-input = 0.25188
-output = 0.82563
-cache_read = 0.12594
-
-[limit]
-context = 430_000
-output = 43_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml b/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
deleted file mode 100644
index 12c7d4b7413..00000000000
--- a/providers/hyper/models/qwen3-coder-480b-a35b-instruct-int4-mixed-ar.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "alibaba/qwen3-coder-480b-a35b-instruct"
-release_date = "2026-04-30"
-last_updated = "2026-07-22"
-
-[cost]
-input = 0.445
-output = 2.145
-cache_read = 0.2225
-
-[limit]
-context = 106_000
-output = 10_600
diff --git a/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml b/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
deleted file mode 100644
index 5cfa8ce4d56..00000000000
--- a/providers/hyper/models/qwen3-next-80b-a3b-instruct.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "alibaba/qwen3-next-80b-a3b-instruct"
-release_date = "2026-04-30"
-last_updated = "2026-07-22"
-
-[cost]
-input = 0.1175
-output = 1.136
-cache_read = 0.05875
-
-[limit]
-context = 262_144
-output = 26_214
diff --git a/providers/hyper/models/qwen3.6-flash.toml b/providers/hyper/models/qwen3.6-flash.toml
deleted file mode 100644
index 0554b788c42..00000000000
--- a/providers/hyper/models/qwen3.6-flash.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "alibaba/qwen3.6-flash"
-release_date = "2026-05-20"
-last_updated = "2026-07-22"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
-
-[cost]
-input = 1
-output = 4
-cache_read = 0.1
-cache_write = 1.25
-
-[limit]
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/hyper/models/qwen3.6-max.toml b/providers/hyper/models/qwen3.6-max.toml
deleted file mode 100644
index f954b161903..00000000000
--- a/providers/hyper/models/qwen3.6-max.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "alibaba/qwen3.6-max-preview"
-release_date = "2026-05-20"
-last_updated = "2026-07-22"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
-
-[cost]
-input = 2
-output = 12
-cache_read = 0.2
-cache_write = 2.5
-
-[limit]
-context = 256_000
-output = 64_000
diff --git a/providers/hyper/models/qwen3.6-plus.toml b/providers/hyper/models/qwen3.6-plus.toml
deleted file mode 100644
index b5db258f7f9..00000000000
--- a/providers/hyper/models/qwen3.6-plus.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "alibaba/qwen3.6-plus"
-release_date = "2026-05-20"
-last_updated = "2026-07-22"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high"]
-
-[cost]
-input = 2
-output = 6
-cache_read = 0.2
-cache_write = 2.5
-
-[limit]
-output = 64_000
-
-[modalities]
-input = ["text", "image"]

From e9e1535a1f1c8617d3e06608ffa3a16f7c1624ea Mon Sep 17 00:00:00 2001
From: Lauri <65590793+Mrflatt@users.noreply.github.com>
Date: Thu, 17 Sep 2026 03:08:48 +0300
Subject: [PATCH 1467/2012] feat(google-vertex): add GLM 5.2 (#7267)

---
 .../models/zai-org/glm-5.2-maas.toml          | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/google-vertex/models/zai-org/glm-5.2-maas.toml

diff --git a/providers/google-vertex/models/zai-org/glm-5.2-maas.toml b/providers/google-vertex/models/zai-org/glm-5.2-maas.toml
new file mode 100644
index 00000000000..f8680d8075c
--- /dev/null
+++ b/providers/google-vertex/models/zai-org/glm-5.2-maas.toml
@@ -0,0 +1,24 @@
+# POST .../endpoints/openapi/chat/completions; toggle with
+# $.chat_template_kwargs.enable_thinking = true | false.
+# Sources:
+# - https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/maas/zaiorg/glm-52
+# - https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "toggle" }]
+status = "beta"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.14
+
+[limit]
+context = 1_000_000
+output = 64_000
+
+[provider]
+npm = "@ai-sdk/openai-compatible"
+api = "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"

From 4cff3a15bd90eb12f55b37d35e47ac61554c5cd9 Mon Sep 17 00:00:00 2001
From: charles <crosenblum@coreweave.com>
Date: Wed, 16 Sep 2026 20:09:17 -0400
Subject: [PATCH 1468/2012] fix(wandb): display CoreWeave provider branding
 (#7255)

---
 packages/core/src/sync/providers/wandb.ts | 4 ++--
 providers/wandb/provider.toml             | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/packages/core/src/sync/providers/wandb.ts b/packages/core/src/sync/providers/wandb.ts
index a63bff674e0..0adb6b2fe65 100644
--- a/packages/core/src/sync/providers/wandb.ts
+++ b/packages/core/src/sync/providers/wandb.ts
@@ -102,7 +102,7 @@ const modalityMap: Record<string, SupportedModality | undefined> = {
 
 export const wandb = {
   id: "wandb",
-  name: "Weights & Biases",
+  name: "CoreWeave",
   modelsDir: "providers/wandb/models",
   deleteMissing: true,
   sourceID(model) {
@@ -111,7 +111,7 @@ export const wandb = {
   async fetchModels() {
     const response = await fetch(API_ENDPOINT);
     if (!response.ok) {
-      throw new Error(`W&B Inference request failed: ${response.status} ${response.statusText}`);
+      throw new Error(`CoreWeave Inference request failed: ${response.status} ${response.statusText}`);
     }
     return response.json();
   },
diff --git a/providers/wandb/provider.toml b/providers/wandb/provider.toml
index a1b023378a3..83867ad26e6 100644
--- a/providers/wandb/provider.toml
+++ b/providers/wandb/provider.toml
@@ -1,4 +1,4 @@
-name = "Weights & Biases"
+name = "CoreWeave"
 # POST /v1/chat/completions toggles eligible models with
 # chat_template_kwargs.enable_thinking=true|false; output is choices[].message.reasoning.
 # DeepSeek V4 and Gemma 4 default off; Nemotron 3, Qwen3.5/3.6, and GLM-5.1/5.2
@@ -8,4 +8,4 @@ name = "Weights & Biases"
 env = ["WANDB_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
 api = "https://api.inference.wandb.ai/v1"
-doc = "https://docs.wandb.ai/guides/integrations/inference/"
+doc = "https://docs.wandb.ai/inference"

From 3c3f0fef578009f109b7a0ced97b35f3dc47fab2 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:10:05 -0500
Subject: [PATCH 1469/2012] fix: [privatemode] sync supported models (#7236)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../privatemode-ai/models/glm-5.3-flash.toml  | 25 ++++++++++++++++++
 .../models/glm-flash-latest.toml              | 26 +++++++++++++++++++
 .../privatemode-ai/models/kimi-k2.6.toml      |  4 ++-
 .../privatemode-ai/models/kimi-latest.toml    |  4 ++-
 4 files changed, 57 insertions(+), 2 deletions(-)
 create mode 100644 providers/privatemode-ai/models/glm-5.3-flash.toml
 create mode 100644 providers/privatemode-ai/models/glm-flash-latest.toml

diff --git a/providers/privatemode-ai/models/glm-5.3-flash.toml b/providers/privatemode-ai/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..1dfd3467c63
--- /dev/null
+++ b/providers/privatemode-ai/models/glm-5.3-flash.toml
@@ -0,0 +1,25 @@
+# Sources (accessed 2026-09-16):
+# Model overview (preview): https://docs.privatemode.ai/models/overview/#glm-53-flash
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Added in v1.56.0: https://docs.privatemode.ai/release/#v1560
+# Upstream: https://huggingface.co/zai-org/GLM-5.3-Flash
+# Effort: reasoning_effort = low|high|max (default max; always on).
+# Any other value including none is silently mapped to max; reasoning cannot be off.
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10),
+# matching other Privatemode catalog entries.
+# EUR/MTok: input 0.77 / output 3.87 / cached input 0.08
+# Host exposes text+image only (docs); lab base also lists video/pdf.
+base_model = "zhipuai/glm-5.3-flash"
+status = "beta"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[modalities]
+input = ["text", "image"]
+
+[cost]
+input = 0.8897
+output = 4.4718
+cache_read = 0.0924
+
+[limit]
+context = 256_000
diff --git a/providers/privatemode-ai/models/glm-flash-latest.toml b/providers/privatemode-ai/models/glm-flash-latest.toml
new file mode 100644
index 00000000000..50cc63ce680
--- /dev/null
+++ b/providers/privatemode-ai/models/glm-flash-latest.toml
@@ -0,0 +1,26 @@
+# Sources (accessed 2026-09-16):
+# Alias routes to the latest available GLM Flash model (currently GLM-5.3-Flash):
+# https://docs.privatemode.ai/models/overview/#glm-53-flash
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Release: https://docs.privatemode.ai/release/#v1560
+# Effort: reasoning_effort = low|high|max (default max; always on).
+# Any other value including none is silently mapped to max; reasoning cannot be off.
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10),
+# matching other Privatemode catalog entries.
+# EUR/MTok: input 0.77 / output 3.87 / cached input 0.08
+# Host exposes text+image only (docs); lab base also lists video/pdf.
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM Flash (latest)"
+status = "beta"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[modalities]
+input = ["text", "image"]
+
+[cost]
+input = 0.8897
+output = 4.4718
+cache_read = 0.0924
+
+[limit]
+context = 256_000
diff --git a/providers/privatemode-ai/models/kimi-k2.6.toml b/providers/privatemode-ai/models/kimi-k2.6.toml
index 2ad5fece2e4..dc0dd54460b 100644
--- a/providers/privatemode-ai/models/kimi-k2.6.toml
+++ b/providers/privatemode-ai/models/kimi-k2.6.toml
@@ -1,12 +1,14 @@
-# Sources (accessed 2026-08-10):
+# Sources (accessed 2026-09-16):
 # Model overview & reasoning toggle: https://docs.privatemode.ai/models/overview/#kimi-k26
 # Pricing (EUR): https://docs.privatemode.ai/pricing/
 # Added in v1.40.0 (default model): https://docs.privatemode.ai/release/#v1400
+# Deprecated in v1.56.0 (migrate to GLM-5.3 or GLM-5.3-Flash): https://docs.privatemode.ai/release/#v1560
 # Upstream: https://huggingface.co/moonshotai/Kimi-K2.6
 # Toggle: chat_template_kwargs.thinking = true|false (default true / reasoning on).
 # Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
 # EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
 base_model = "moonshotai/kimi-k2.6"
+status = "deprecated"
 reasoning_options = [{ type = "toggle" }]
 
 [modalities]
diff --git a/providers/privatemode-ai/models/kimi-latest.toml b/providers/privatemode-ai/models/kimi-latest.toml
index ba3eec7dd66..990b62aef86 100644
--- a/providers/privatemode-ai/models/kimi-latest.toml
+++ b/providers/privatemode-ai/models/kimi-latest.toml
@@ -1,12 +1,14 @@
-# Sources (accessed 2026-08-10):
+# Sources (accessed 2026-09-16):
 # Alias routes to the latest available Kimi model (currently Kimi K2.6):
 # https://docs.privatemode.ai/models/overview/#kimi-k26
 # Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Deprecated in v1.56.0 (migrate to GLM-5.3 or GLM-5.3-Flash): https://docs.privatemode.ai/release/#v1560
 # Toggle: chat_template_kwargs.thinking = true|false (default true / reasoning on).
 # Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
 # EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
 base_model = "moonshotai/kimi-k2.6"
 name = "Kimi (latest)"
+status = "deprecated"
 reasoning_options = [{ type = "toggle" }]
 
 [modalities]

From 2db949bf037d2fe43b9245a90463d29ccc3d46e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=88=91=E4=BB=AC=E5=9C=A8?= <1244791533@qq.com>
Date: Thu, 17 Sep 2026 08:17:18 +0800
Subject: [PATCH 1470/2012] [missing-model] volcengine: add GLM-5.3-Flash
 (glm-5-3-flash-260828) (#7222)

* Add volcengine GLM-5.3-Flash (glm-5-3-flash-260828)

* Fix reasoning options and add cost per review
---
 .../models/glm-5-3-flash-260828.toml           | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/volcengine/models/glm-5-3-flash-260828.toml

diff --git a/providers/volcengine/models/glm-5-3-flash-260828.toml b/providers/volcengine/models/glm-5-3-flash-260828.toml
new file mode 100644
index 00000000000..98cb87ee9ba
--- /dev/null
+++ b/providers/volcengine/models/glm-5-3-flash-260828.toml
@@ -0,0 +1,18 @@
+# Model ID verified against POST /api/v3/chat/completions (2026-09-16): `glm-5-3-flash-260828`
+# CNY→USD rate: 6.737012, 2026-08-26, source: https://open.er-api.com/v6/latest/USD
+# CNY list price source: https://docs.volcengine.com/docs/82379/1449737 (深度思考) + 智谱官方价目表
+# GLM-5.3-Flash always reasons (thinking cannot be disabled on this model ID)
+# Effort: reasoning_effort = low|high|max
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.11875
+output = 0.41563
+cache_read = 0.03414

From 5d582f6d4a5d79ac484efa112df030cf23410901 Mon Sep 17 00:00:00 2001
From: 4ebuRushka <metelyov.vadim@gmail.com>
Date: Thu, 17 Sep 2026 07:17:30 +0700
Subject: [PATCH 1471/2012] feat(nvidia): add z-ai/glm-5.3 (hosted NIM) (#7211)

---
 providers/nvidia/models/z-ai/glm-5.3.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/nvidia/models/z-ai/glm-5.3.toml

diff --git a/providers/nvidia/models/z-ai/glm-5.3.toml b/providers/nvidia/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..7ac15f9946c
--- /dev/null
+++ b/providers/nvidia/models/z-ai/glm-5.3.toml
@@ -0,0 +1,14 @@
+# Hosted NIM: model listed at GET https://integrate.api.nvidia.com/v1/models
+# (verified 2026-09-16). Thinking is always enabled; reasoning budget controlled
+# with top-level `reasoning_effort`: low|high|max, default max.
+# https://docs.nvidia.com/nim/large-language-models/2.0.11/get-started/advanced/get-started-glm-5.3.html
+base_model = "zhipuai/glm-5.3"
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0
\ No newline at end of file

From 34c64881d99424a378e0e6bf7970ed861965f3d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 16 Sep 2026 19:40:06 -0500
Subject: [PATCH 1472/2012] feat: retry stalled issue fixes (#7213)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .github/workflows/retry-issue-fixer.yml | 102 ++++++++++++++++++++++++
 sync.md                                 |   2 +
 2 files changed, 104 insertions(+)
 create mode 100644 .github/workflows/retry-issue-fixer.yml

diff --git a/.github/workflows/retry-issue-fixer.yml b/.github/workflows/retry-issue-fixer.yml
new file mode 100644
index 00000000000..d10a3206c98
--- /dev/null
+++ b/.github/workflows/retry-issue-fixer.yml
@@ -0,0 +1,102 @@
+name: Retry stalled issue fixes
+
+on:
+  schedule:
+    - cron: "41 */6 * * *"
+  workflow_dispatch:
+
+permissions:
+  contents: write
+  issues: read
+  pull-requests: read
+
+concurrency:
+  group: retry-stalled-issue-fixes
+  cancel-in-progress: false
+
+jobs:
+  retry:
+    if: github.repository == 'anomalyco/models.dev'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Redispatch oldest issues without pull requests
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          set -euo pipefail
+
+          cutoff="$(date -u -d '1 hour ago' '+%Y-%m-%dT%H:%M:%SZ')"
+          owner="${GITHUB_REPOSITORY%/*}"
+          repo="${GITHUB_REPOSITORY#*/}"
+
+          issues="$({
+            gh issue list \
+              --repo "$GITHUB_REPOSITORY" \
+              --state open \
+              --label automation \
+              --label model-sync \
+              --label missing-model \
+              --limit 1000 \
+              --json number,createdAt,labels
+          } | jq --arg cutoff "$cutoff" '
+            map(select(.createdAt <= $cutoff)) | sort_by(.createdAt)
+          ')"
+
+          dispatched=0
+          dispatched_issues=()
+          while IFS= read -r issue; do
+            number="$(jq -r '.number' <<< "$issue")"
+            provider="$(jq -r '
+              [.labels[].name | select(startswith("provider:")) | ltrimstr("provider:")]
+              | first // empty
+            ' <<< "$issue")"
+
+            if [ -z "$provider" ] || [ "$provider" = "openai" ]; then
+              continue
+            fi
+
+            closing_prs="$(gh api graphql \
+              -F owner="$owner" \
+              -F repo="$repo" \
+              -F number="$number" \
+              -f query='query($owner: String!, $repo: String!, $number: Int!) {
+                repository(owner: $owner, name: $repo) {
+                  issue(number: $number) {
+                    closedByPullRequestsReferences(first: 1, includeClosedPrs: true) {
+                      totalCount
+                    }
+                  }
+                }
+              }' \
+              --jq '.data.repository.issue.closedByPullRequestsReferences.totalCount')"
+
+            if [ "$closing_prs" -gt 0 ]; then
+              continue
+            fi
+
+            jq -n \
+              --arg provider "$provider" \
+              --argjson issue_number "$number" \
+              '{
+                event_type: "missing-model",
+                client_payload: {
+                  provider: $provider,
+                  issue_number: $issue_number
+                }
+              }' \
+              | gh api "repos/$GITHUB_REPOSITORY/dispatches" --method POST --input -
+
+            dispatched=$((dispatched + 1))
+            dispatched_issues+=("#$number")
+            if [ "$dispatched" -eq 3 ]; then
+              break
+            fi
+          done < <(jq -c '.[]' <<< "$issues")
+
+          if [ "$dispatched" -eq 0 ]; then
+            echo "No stalled automated issues were eligible for retry." | tee -a "$GITHUB_STEP_SUMMARY"
+          else
+            printf 'Redispatched %d issue(s): %s\n' \
+              "$dispatched" "${dispatched_issues[*]}" \
+              | tee -a "$GITHUB_STEP_SUMMARY"
+          fi
diff --git a/sync.md b/sync.md
index 6bb7e47bb22..abfb3fac737 100644
--- a/sync.md
+++ b/sync.md
@@ -60,6 +60,8 @@ Providers that cannot safely auto-create TOMLs set `skipCreates: true`. In GitHu
 4. Dispatches the Issue Fixer explicitly so issues created with `GITHUB_TOKEN` can still produce PRs
 5. If listing fails, creates nothing (fail closed)
 
+Every six hours, a recovery workflow checks the oldest open issues with all three automation labels. It redispatches up to three issues that are at least one hour old and have no closing pull request, so a missed or failed Issue Fixer run does not leave them stranded. The one-hour delay keeps the recovery run from racing the initial dispatch.
+
 Providers that can auto-create most models may instead return an ID from `missingModelID` only for `translateModel` skips that need manual metadata. The runner preserves an existing local entry for that ID while the issue is handled. Intentional skips return `undefined` and do not open issues.
 
 Requires `GH_TOKEN` on the sync workflow step. Local runs are notice-only unless `--open-issues`. Use `--no-issues` / `--dry-run` to skip creates. Each newly opened issue explicitly dispatches the issue-fixer workflow so an agent can research the missing metadata and open a model PR.

From 62a8a3ea3976ee0ba0c5c4a316a74df6afc1884b Mon Sep 17 00:00:00 2001
From: Mykola Biron <mykolabiron@gmail.com>
Date: Thu, 17 Sep 2026 02:42:36 +0200
Subject: [PATCH 1473/2012] add AI21 Labs Provider (#7017)

* add ai21 model provider

* fix: use AI21 text-mark logo with currentColor and relative sizing

* fix: replace wordmark with square AI21 icon mark
---
 models/ai21/jamba-large.toml           | 24 ++++++++++++++++++++++++
 models/ai21/jamba-mini.toml            | 24 ++++++++++++++++++++++++
 providers/ai21/logo.svg                |  3 +++
 providers/ai21/models/jamba-large.toml |  5 +++++
 providers/ai21/models/jamba-mini.toml  |  5 +++++
 providers/ai21/provider.toml           |  5 +++++
 6 files changed, 66 insertions(+)
 create mode 100644 models/ai21/jamba-large.toml
 create mode 100644 models/ai21/jamba-mini.toml
 create mode 100644 providers/ai21/logo.svg
 create mode 100644 providers/ai21/models/jamba-large.toml
 create mode 100644 providers/ai21/models/jamba-mini.toml
 create mode 100644 providers/ai21/provider.toml

diff --git a/models/ai21/jamba-large.toml b/models/ai21/jamba-large.toml
new file mode 100644
index 00000000000..68bb57edeb9
--- /dev/null
+++ b/models/ai21/jamba-large.toml
@@ -0,0 +1,24 @@
+name = "Jamba Large"
+description = "AI21's hybrid SSM-Transformer long-context model for enterprise agents and grounded generation"
+family = "jamba"
+release_date = "2025-07-01"
+last_updated = "2025-07-01"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2024-08-22"
+open_weights = true
+
+[limit]
+context = 256_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/ai21labs/AI21-Jamba-Large-1.6"
diff --git a/models/ai21/jamba-mini.toml b/models/ai21/jamba-mini.toml
new file mode 100644
index 00000000000..c9c5c7d8fc6
--- /dev/null
+++ b/models/ai21/jamba-mini.toml
@@ -0,0 +1,24 @@
+name = "Jamba Mini"
+description = "AI21's efficient, lightweight hybrid SSM-Transformer model for a wide range of tasks"
+family = "jamba"
+release_date = "2026-01-01"
+last_updated = "2026-01-01"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2024-08-22"
+open_weights = true
+
+[limit]
+context = 256_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/ai21labs/AI21-Jamba-Mini-1.6"
diff --git a/providers/ai21/logo.svg b/providers/ai21/logo.svg
new file mode 100644
index 00000000000..b929375b0d2
--- /dev/null
+++ b/providers/ai21/logo.svg
@@ -0,0 +1,3 @@
+<svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg">
+<path d="M3 14.5A6.5 6.5 0 1 1 16 14.5A6.5 6.5 0 1 1 3 14.5Z M5.7 14.5A3.8 3.8 0 1 0 13.3 14.5A3.8 3.8 0 1 0 5.7 14.5Z M13 5H16.3V21H13Z M10.6 5H13V7.4H10.6Z"/>
+</svg>
diff --git a/providers/ai21/models/jamba-large.toml b/providers/ai21/models/jamba-large.toml
new file mode 100644
index 00000000000..dfa0edbeab9
--- /dev/null
+++ b/providers/ai21/models/jamba-large.toml
@@ -0,0 +1,5 @@
+base_model = "ai21/jamba-large"
+
+[cost]
+input = 2.00
+output = 8.00
diff --git a/providers/ai21/models/jamba-mini.toml b/providers/ai21/models/jamba-mini.toml
new file mode 100644
index 00000000000..164715bac23
--- /dev/null
+++ b/providers/ai21/models/jamba-mini.toml
@@ -0,0 +1,5 @@
+base_model = "ai21/jamba-mini"
+
+[cost]
+input = 0.20
+output = 0.40
diff --git a/providers/ai21/provider.toml b/providers/ai21/provider.toml
new file mode 100644
index 00000000000..3dd6911976e
--- /dev/null
+++ b/providers/ai21/provider.toml
@@ -0,0 +1,5 @@
+name = "AI21 Labs"
+env = ["AI21_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.ai21.com/studio/v1"
+doc = "https://docs.ai21.com/docs/jamba-foundation-models"

From 06c0b1c04dc31fd6d0d3e5114d80e69b1944a3b5 Mon Sep 17 00:00:00 2001
From: Yishuai Li <yishuai.li@pingcap.com>
Date: Thu, 17 Sep 2026 08:43:06 +0800
Subject: [PATCH 1474/2012] Add kimi-k3 to alibaba and alibaba-cn providers
 (#7149)

* Add kimi-k3 to alibaba and alibaba-cn providers

* Fix kimi-k3 reasoning options: effort low|high|max per zh API reference

* Use block-style reasoning_options, matching moonshotai/kimi-k3 style
---
 providers/alibaba-cn/models/kimi-k3.toml | 27 ++++++++++++++++++++++++
 providers/alibaba/models/kimi-k3.toml    | 27 ++++++++++++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 providers/alibaba-cn/models/kimi-k3.toml
 create mode 100644 providers/alibaba/models/kimi-k3.toml

diff --git a/providers/alibaba-cn/models/kimi-k3.toml b/providers/alibaba-cn/models/kimi-k3.toml
new file mode 100644
index 00000000000..0df890f8e11
--- /dev/null
+++ b/providers/alibaba-cn/models/kimi-k3.toml
@@ -0,0 +1,27 @@
+# Sources (accessed 2026-09-15):
+# https://www.alibabacloud.com/help/en/model-studio/kimi-k3 (China/Beijing USD
+# pricing, limits, capability table)
+# https://www.alibabacloud.com/help/zh/model-studio/qwen-api-via-openai-chat-completions
+# (reasoning_effort) and /help/en/model-studio/kimi-api (text+image input only,
+# no video; reasoning_content streamed in deltas)
+# Effort: reasoning_effort = low|high|max (default max). Thinking is always on:
+# enable_thinking only accepts true (passing false errors), no thinking_budget.
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.827
+output = 14.133
+cache_read = 0.283
+
+[modalities]
+input = ["text", "image"]
+
+[limit]
+output = 1_048_576
diff --git a/providers/alibaba/models/kimi-k3.toml b/providers/alibaba/models/kimi-k3.toml
new file mode 100644
index 00000000000..28505229989
--- /dev/null
+++ b/providers/alibaba/models/kimi-k3.toml
@@ -0,0 +1,27 @@
+# Sources (accessed 2026-09-15):
+# https://www.qwencloud.com/models/kimi-k3 (Singapore/international pricing, limits)
+# https://www.alibabacloud.com/help/en/model-studio/kimi-k3 (limits, capability table)
+# https://www.alibabacloud.com/help/zh/model-studio/qwen-api-via-openai-chat-completions
+# (reasoning_effort) and /help/en/model-studio/kimi-api (text+image input only,
+# no video; reasoning_content streamed in deltas)
+# Effort: reasoning_effort = low|high|max (default max). Thinking is always on:
+# enable_thinking only accepts true (passing false errors), no thinking_budget.
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 3.0
+output = 15.0
+cache_read = 0.3
+
+[modalities]
+input = ["text", "image"]
+
+[limit]
+output = 1_048_576

From 62ef4fab7fc06fc819cccf863386be02db5f7420 Mon Sep 17 00:00:00 2001
From: Tao <76082878+K-3-LT@users.noreply.github.com>
Date: Thu, 17 Sep 2026 08:43:46 +0800
Subject: [PATCH 1475/2012] feat: update IteraCompute model catalog (#7142)

* feat: update IteraCompute model catalog

* fix: sync IteraCompute prices with production catalog
---
 .../deepseek/deepseek-v4-flash-0731.toml      | 16 ++++++++++
 .../models/deepseek/deepseek-v4-pro-0813.toml | 16 ++++++++++
 .../iteracompute/ornith-1.5-35b-a3b.toml      | 31 -------------------
 .../models/iteracompute/qwen3.8-27b.toml      | 24 --------------
 .../models/minimax/minimax-m3.toml            | 19 ++++++++++++
 .../models/moonshotai/kimi-k3.toml            | 20 ++++++++++++
 .../models/ornith-ai/ornith-1.5-35b-a3b.toml  | 22 +++++++++++++
 .../models/qwen/qwen3.8-2.4t-a95b.toml        | 19 ++++++++++++
 .../iteracompute/models/qwen/qwen3.8-27b.toml | 19 ++++++++++++
 .../models/z-ai/glm-5.3-flash.toml            | 20 ++++++++++++
 .../iteracompute/models/z-ai/glm-5.3.toml     | 17 ++++++++++
 11 files changed, 168 insertions(+), 55 deletions(-)
 create mode 100644 providers/iteracompute/models/deepseek/deepseek-v4-flash-0731.toml
 create mode 100644 providers/iteracompute/models/deepseek/deepseek-v4-pro-0813.toml
 delete mode 100644 providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml
 delete mode 100644 providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
 create mode 100644 providers/iteracompute/models/minimax/minimax-m3.toml
 create mode 100644 providers/iteracompute/models/moonshotai/kimi-k3.toml
 create mode 100644 providers/iteracompute/models/ornith-ai/ornith-1.5-35b-a3b.toml
 create mode 100644 providers/iteracompute/models/qwen/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/iteracompute/models/qwen/qwen3.8-27b.toml
 create mode 100644 providers/iteracompute/models/z-ai/glm-5.3-flash.toml
 create mode 100644 providers/iteracompute/models/z-ai/glm-5.3.toml

diff --git a/providers/iteracompute/models/deepseek/deepseek-v4-flash-0731.toml b/providers/iteracompute/models/deepseek/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..de987dfd5fe
--- /dev/null
+++ b/providers/iteracompute/models/deepseek/deepseek-v4-flash-0731.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Reasoning is mandatory and the public catalog exposes no caller-selectable reasoning control.
+# Prices are in USD per million tokens.
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = []
+
+[cost]
+input = 0.34
+output = 1.05
+cache_read = 0.035
+
+[limit]
+context = 970_000
+output = 393_216
diff --git a/providers/iteracompute/models/deepseek/deepseek-v4-pro-0813.toml b/providers/iteracompute/models/deepseek/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..0199acb7930
--- /dev/null
+++ b/providers/iteracompute/models/deepseek/deepseek-v4-pro-0813.toml
@@ -0,0 +1,16 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Reasoning is mandatory and the public catalog exposes no caller-selectable reasoning control.
+# Prices are in USD per million tokens.
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = []
+
+[cost]
+input = 1.10
+output = 3.30
+cache_read = 0.11
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml b/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml
deleted file mode 100644
index 59d4eb1b5ea..00000000000
--- a/providers/iteracompute/models/iteracompute/ornith-1.5-35b-a3b.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-# Sources (accessed 2026-08-28):
-# https://api.iteracompute.com/v1/models (model ID, availability, limits, and pricing)
-# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
-# https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B (upstream capabilities)
-#
-# IteraCompute serves the language-only NVFP4 build with a 327,680-token
-# combined window, up to 262,144 input tokens, and up to 65,536 output tokens.
-# Reasoning toggle wire path: reasoning_effort = "none" disables thinking;
-# any accepted non-none effort enables it. Responses stream reasoning through
-# the reasoning_content delta field.
-# Prices are in USD per million tokens.
-base_model = "deepreinforce/ornith-1.5-35b-a3b"
-attachment = false
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0.30
-output = 3.00
-
-[limit]
-context = 327_680
-input = 262_144
-output = 65_536
-
-[modalities]
-input = ["text"]
diff --git a/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml b/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
deleted file mode 100644
index 8eb4170a5f0..00000000000
--- a/providers/iteracompute/models/iteracompute/qwen3.8-27b.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-# Sources (accessed 2026-09-01):
-# https://api.iteracompute.com/v1/models (model ID, availability, and input/output pricing)
-# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
-# https://huggingface.co/gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 (served checkpoint and image input)
-#
-# IteraCompute serves Qwen3.8 27B through its OpenAI-compatible API.
-# The NVFP4 deployment accepts text and image input and extends the combined
-# serving window to 327,680 tokens with up to 262,144 input and 65,536 output tokens.
-# Prices are in USD per million tokens.
-# Effort: reasoning_effort = none|low|medium|xhigh (xhigh is the default).
-base_model = "alibaba/qwen3.8-27b"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
-
-[cost]
-input = 0.30
-output = 2.50
-
-[limit]
-context = 327_680
-input = 262_144
-output = 65_536
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/iteracompute/models/minimax/minimax-m3.toml b/providers/iteracompute/models/minimax/minimax-m3.toml
new file mode 100644
index 00000000000..f9d9908e3f6
--- /dev/null
+++ b/providers/iteracompute/models/minimax/minimax-m3.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Reasoning is mandatory and the public catalog exposes no caller-selectable reasoning control.
+# Prices are in USD per million tokens.
+base_model = "minimax/MiniMax-M3"
+reasoning_options = []
+
+[cost]
+input = 0.29
+output = 1.20
+cache_read = 0.08
+
+[limit]
+context = 1_048_576
+output = 524_288
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/iteracompute/models/moonshotai/kimi-k3.toml b/providers/iteracompute/models/moonshotai/kimi-k3.toml
new file mode 100644
index 00000000000..4e4af1d1240
--- /dev/null
+++ b/providers/iteracompute/models/moonshotai/kimi-k3.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Reasoning is mandatory and the public catalog exposes no caller-selectable reasoning control.
+# Prices are in USD per million tokens.
+base_model = "moonshotai/kimi-k3"
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 3.00
+output = 14.90
+cache_read = 0.29
+
+[limit]
+context = 1_048_576
+output = 999_999
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/iteracompute/models/ornith-ai/ornith-1.5-35b-a3b.toml b/providers/iteracompute/models/ornith-ai/ornith-1.5-35b-a3b.toml
new file mode 100644
index 00000000000..e0ee4a079e5
--- /dev/null
+++ b/providers/iteracompute/models/ornith-ai/ornith-1.5-35b-a3b.toml
@@ -0,0 +1,22 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (canonical model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Toggle wire path: reasoning_effort = "none" disables thinking; responses expose reasoning_content.
+# Prices are in USD per million tokens.
+base_model = "deepreinforce/ornith-1.5-35b-a3b"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.30
+output = 3.00
+cache_read = 0.03
+
+[limit]
+context = 327_680
+input = 262_144
+output = 65_536
diff --git a/providers/iteracompute/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/iteracompute/models/qwen/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..1499c45a580
--- /dev/null
+++ b/providers/iteracompute/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Prices are in USD per million tokens.
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+attachment = true
+reasoning_options = [{ type = "toggle" }]
+
+[cost]
+input = 1.95
+output = 5.95
+cache_read = 0.20
+
+[limit]
+context = 970_000
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/iteracompute/models/qwen/qwen3.8-27b.toml b/providers/iteracompute/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..bb7bc16c39d
--- /dev/null
+++ b/providers/iteracompute/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (canonical model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Prices are in USD per million tokens.
+base_model = "alibaba/qwen3.8-27b"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
+
+[cost]
+input = 0.30
+output = 2.50
+cache_read = 0.03
+
+[limit]
+context = 327_680
+input = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/iteracompute/models/z-ai/glm-5.3-flash.toml b/providers/iteracompute/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..19503c99421
--- /dev/null
+++ b/providers/iteracompute/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, pricing, and capacity)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Effort wire path: reasoning.effort = low|high|max. Reasoning is mandatory.
+# Prices are in USD per million tokens.
+base_model = "zhipuai/glm-5.3-flash"
+structured_output = false
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 0.14
+output = 0.49
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/iteracompute/models/z-ai/glm-5.3.toml b/providers/iteracompute/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..2ba1f3cca79
--- /dev/null
+++ b/providers/iteracompute/models/z-ai/glm-5.3.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-15):
+# https://api.iteracompute.com/v1/models (model ID, availability, capabilities, limits, and pricing)
+# https://iteracompute.com/docs.html (OpenAI-compatible API documentation)
+# Reasoning is mandatory and the public catalog exposes no caller-selectable reasoning control.
+# Prices are in USD per million tokens.
+base_model = "zhipuai/glm-5.3"
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 1.20
+output = 3.50
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 131_072

From f76861302c380895fd864160cfec57850dcd6245 Mon Sep 17 00:00:00 2001
From: Fede Kamelhar <209537060+fede-kamel@users.noreply.github.com>
Date: Wed, 16 Sep 2026 20:44:11 -0400
Subject: [PATCH 1476/2012] Add OCI Generative AI provider (#7087)

* Add OCI Generative AI provider (OpenAI-compatible endpoint, API-key auth)

Signed-off-by: Federico Kamelhar <federico.kamelhar@oracle.com>

* oci: do not advertise tool calling for the Meta Llama models

OCI's OpenAI-compatible endpoint rejects tool-calling requests whose first
message is a system message for every Meta Llama model (400 "Tool calling for
this Llama model requires the first message to be a user message"). opencode
always sends its system prompt first together with its built-in tools, so with
tool_call = true these models fail on the first turn. Chat without tools works.
Re-enable once the endpoint accepts a leading system message.

Signed-off-by: Federico Kamelhar <federico.kamelhar@oracle.com>

---------

Signed-off-by: Federico Kamelhar <federico.kamelhar@oracle.com>
---
 providers/oci/logo.svg                        |  3 +++
 .../models/meta.llama-3.3-70b-instruct.toml   | 15 ++++++++++++
 ...lama-4-maverick-17b-128e-instruct-fp8.toml | 18 +++++++++++++++
 .../meta.llama-4-scout-17b-16e-instruct.toml  | 18 +++++++++++++++
 providers/oci/models/openai.gpt-oss-120b.toml | 20 ++++++++++++++++
 providers/oci/models/openai.gpt-oss-20b.toml  | 20 ++++++++++++++++
 .../models/xai.grok-4.20-non-reasoning.toml   | 21 +++++++++++++++++
 .../oci/models/xai.grok-4.20-reasoning.toml   | 23 +++++++++++++++++++
 providers/oci/models/xai.grok-4.3.toml        | 23 +++++++++++++++++++
 providers/oci/models/xai.grok-4.6.toml        | 19 +++++++++++++++
 providers/oci/provider.toml                   | 22 ++++++++++++++++++
 11 files changed, 202 insertions(+)
 create mode 100644 providers/oci/logo.svg
 create mode 100644 providers/oci/models/meta.llama-3.3-70b-instruct.toml
 create mode 100644 providers/oci/models/meta.llama-4-maverick-17b-128e-instruct-fp8.toml
 create mode 100644 providers/oci/models/meta.llama-4-scout-17b-16e-instruct.toml
 create mode 100644 providers/oci/models/openai.gpt-oss-120b.toml
 create mode 100644 providers/oci/models/openai.gpt-oss-20b.toml
 create mode 100644 providers/oci/models/xai.grok-4.20-non-reasoning.toml
 create mode 100644 providers/oci/models/xai.grok-4.20-reasoning.toml
 create mode 100644 providers/oci/models/xai.grok-4.3.toml
 create mode 100644 providers/oci/models/xai.grok-4.6.toml
 create mode 100644 providers/oci/provider.toml

diff --git a/providers/oci/logo.svg b/providers/oci/logo.svg
new file mode 100644
index 00000000000..a46ebf3f9ec
--- /dev/null
+++ b/providers/oci/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <path d="M8 5.5h8a6.5 6.5 0 0 1 0 13H8a6.5 6.5 0 0 1 0-13zm0 3a3.5 3.5 0 0 0 0 7h8a3.5 3.5 0 0 0 0-7H8z"/>
+</svg>
diff --git a/providers/oci/models/meta.llama-3.3-70b-instruct.toml b/providers/oci/models/meta.llama-3.3-70b-instruct.toml
new file mode 100644
index 00000000000..fabd7912f72
--- /dev/null
+++ b/providers/oci/models/meta.llama-3.3-70b-instruct.toml
@@ -0,0 +1,15 @@
+# OCI list price, pay as you go, USD (public price list, part B108080, accessed 2026-09-14):
+# $0.0018 per 10,000 transactions, where one transaction is one character of input or
+# output. OCI's model pages estimate four characters per token, so the per-token price used
+# here is $0.0018 / 10,000 * 4 * 1,000,000 = $0.72 per 1M input and output tokens.
+# Limits from https://docs.oracle.com/en-us/iaas/Content/generative-ai/meta-llama-3-3-70b.htm
+# The endpoint returns tool-call arguments as JSON strings and requires the first message
+# of a tool-calling request to be a user message.
+base_model = "meta/llama-3.3-70b-instruct"
+# opencode sends its system prompt first with tools; OCI rejects that for Llama models
+# (400 "Tool calling for this Llama model requires the first message to be a user message").
+tool_call = false
+
+[cost]
+input = 0.72
+output = 0.72
diff --git a/providers/oci/models/meta.llama-4-maverick-17b-128e-instruct-fp8.toml b/providers/oci/models/meta.llama-4-maverick-17b-128e-instruct-fp8.toml
new file mode 100644
index 00000000000..5b2d087b79c
--- /dev/null
+++ b/providers/oci/models/meta.llama-4-maverick-17b-128e-instruct-fp8.toml
@@ -0,0 +1,18 @@
+# OCI list price, pay as you go, USD (public price list, part B111036, accessed 2026-09-14):
+# $0.0018 per 10,000 transactions, where one transaction is one character of input or
+# output. OCI's model pages estimate four characters per token, so the per-token price used
+# here is $0.0018 / 10,000 * 4 * 1,000,000 = $0.72 per 1M input and output tokens.
+# Limits from https://docs.oracle.com/en-us/iaas/Content/generative-ai/meta-llama-4-maverick.htm
+# The endpoint returns tool-call arguments as JSON strings and requires the first message
+# of a tool-calling request to be a user message.
+base_model = "meta/llama-4-maverick-17b-instruct"
+# opencode sends its system prompt first with tools; OCI rejects that for Llama models
+# (400 "Tool calling for this Llama model requires the first message to be a user message").
+tool_call = false
+
+[cost]
+input = 0.72
+output = 0.72
+
+[limit]
+context = 512_000
diff --git a/providers/oci/models/meta.llama-4-scout-17b-16e-instruct.toml b/providers/oci/models/meta.llama-4-scout-17b-16e-instruct.toml
new file mode 100644
index 00000000000..a6e246a918f
--- /dev/null
+++ b/providers/oci/models/meta.llama-4-scout-17b-16e-instruct.toml
@@ -0,0 +1,18 @@
+# OCI list price, pay as you go, USD (public price list, part B111035, accessed 2026-09-14):
+# $0.0018 per 10,000 transactions, where one transaction is one character of input or
+# output. OCI's model pages estimate four characters per token, so the per-token price used
+# here is $0.0018 / 10,000 * 4 * 1,000,000 = $0.72 per 1M input and output tokens.
+# Limits from https://docs.oracle.com/en-us/iaas/Content/generative-ai/meta-llama-4-scout.htm
+# The endpoint returns tool-call arguments as JSON strings and requires the first message
+# of a tool-calling request to be a user message.
+base_model = "meta/llama-4-scout-17b-instruct"
+# opencode sends its system prompt first with tools; OCI rejects that for Llama models
+# (400 "Tool calling for this Llama model requires the first message to be a user message").
+tool_call = false
+
+[cost]
+input = 0.72
+output = 0.72
+
+[limit]
+context = 192_000
diff --git a/providers/oci/models/openai.gpt-oss-120b.toml b/providers/oci/models/openai.gpt-oss-120b.toml
new file mode 100644
index 00000000000..9822596fbb4
--- /dev/null
+++ b/providers/oci/models/openai.gpt-oss-120b.toml
@@ -0,0 +1,20 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list, parts B112004/B112005,
+# accessed 2026-09-14). Context and output limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/openai-gpt-oss-120b.htm: 128,000 tokens
+# prompt + response; Oracle documents a 16,000-token response cap in its playground and no
+# larger API cap, so 16,384 is used here.
+# Effort: reasoning_effort = low|medium|high (accepted on this endpoint, 2026-09-14).
+# Reasoning is streamed in `reasoning_content` deltas.
+base_model = "openai/gpt-oss-120b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.60
+
+[limit]
+context = 128_000
+output = 16_384
diff --git a/providers/oci/models/openai.gpt-oss-20b.toml b/providers/oci/models/openai.gpt-oss-20b.toml
new file mode 100644
index 00000000000..8d742f9081d
--- /dev/null
+++ b/providers/oci/models/openai.gpt-oss-20b.toml
@@ -0,0 +1,20 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list, parts B112006/B112007,
+# accessed 2026-09-14). Context and output limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/openai-gpt-oss-20b.htm: 128,000 tokens
+# prompt + response; Oracle documents a 16,000-token response cap in its playground and no
+# larger API cap, so 16,384 is used here.
+# Effort: reasoning_effort = low|medium|high (accepted on this endpoint, 2026-09-14).
+# Reasoning is streamed in `reasoning_content` deltas.
+base_model = "openai/gpt-oss-20b"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.07
+output = 0.30
+
+[limit]
+context = 128_000
+output = 16_384
diff --git a/providers/oci/models/xai.grok-4.20-non-reasoning.toml b/providers/oci/models/xai.grok-4.20-non-reasoning.toml
new file mode 100644
index 00000000000..8baf2ac03bc
--- /dev/null
+++ b/providers/oci/models/xai.grok-4.20-non-reasoning.toml
@@ -0,0 +1,21 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list "Grok 4.2" parts
+# B111910/B111911/B112076-B112079, accessed 2026-09-14); the higher band applies above 200K
+# prompt tokens. Model is hosted by xAI and called through OCI. Limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/xai-grok-4-20.htm (1M context).
+# Only text and base64 images were verified through this endpoint; PDF input is not declared.
+base_model = "xai/grok-4.20-0309-non-reasoning"
+
+[cost]
+input = 1.25
+output = 2.50
+cache_read = 0.20
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.50
+output = 5.00
+cache_read = 0.40
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/oci/models/xai.grok-4.20-reasoning.toml b/providers/oci/models/xai.grok-4.20-reasoning.toml
new file mode 100644
index 00000000000..72c4992fb8f
--- /dev/null
+++ b/providers/oci/models/xai.grok-4.20-reasoning.toml
@@ -0,0 +1,23 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list "Grok 4.2" parts
+# B111910/B111911/B112076-B112079, accessed 2026-09-14); the higher band applies above 200K
+# prompt tokens. Model is hosted by xAI and called through OCI. Limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/xai-grok-4-20.htm (1M context).
+# reasoning_effort is rejected with 400 by this endpoint for the reasoning variant: no caller control.
+# Only text and base64 images were verified through this endpoint; PDF input is not declared.
+base_model = "xai/grok-4.20-0309-reasoning"
+reasoning_options = []
+
+[cost]
+input = 1.25
+output = 2.50
+cache_read = 0.20
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.50
+output = 5.00
+cache_read = 0.40
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/oci/models/xai.grok-4.3.toml b/providers/oci/models/xai.grok-4.3.toml
new file mode 100644
index 00000000000..9bb0c653565
--- /dev/null
+++ b/providers/oci/models/xai.grok-4.3.toml
@@ -0,0 +1,23 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list, parts B112080-B112085,
+# accessed 2026-09-14); the higher band applies above 200K prompt tokens. Model is hosted by
+# xAI and called through OCI (see the OCI model page). Limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/xai-grok-4-3.htm (1M context).
+# Effort: reasoning_effort mirrors xAI's controls for this model.
+# Only text and base64 images were verified through this endpoint; PDF input is not declared.
+base_model = "xai/grok-4.3"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[cost]
+input = 1.25
+output = 2.50
+cache_read = 0.20
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.50
+output = 5.00
+cache_read = 0.40
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/oci/models/xai.grok-4.6.toml b/providers/oci/models/xai.grok-4.6.toml
new file mode 100644
index 00000000000..0821f873c24
--- /dev/null
+++ b/providers/oci/models/xai.grok-4.6.toml
@@ -0,0 +1,19 @@
+# OCI list price, pay as you go, USD per 1M tokens (public price list parts B112086/B112087/
+# B112589-B112592, accessed 2026-09-14); the higher band applies above 200K prompt tokens.
+# OCI also sells a priority tier at twice these prices (service_tier), not modelled here.
+# Model is hosted by xAI and called through OCI. Limits from
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/xai-grok-4-6.htm (500K context).
+# Effort: reasoning_effort mirrors xAI's controls for this model.
+base_model = "xai/grok-4.6"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2.00
+output = 6.00
+cache_read = 0.50
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4.00
+output = 12.00
+cache_read = 1.00
diff --git a/providers/oci/provider.toml b/providers/oci/provider.toml
new file mode 100644
index 00000000000..e623d53d68d
--- /dev/null
+++ b/providers/oci/provider.toml
@@ -0,0 +1,22 @@
+# OCI Generative AI, OpenAI-compatible endpoint (accessed 2026-09-14):
+# POST https://inference.generativeai.<region>.oci.oraclecloud.com/openai/v1/chat/completions
+# Auth: an OCI Generative AI API key sent as `Authorization: Bearer <key>` (create one with
+# `oci generative-ai api-key create`; keys are regional and compartment-scoped, no other
+# header is needed). The default URL below targets us-chicago-1; for the other supported
+# regions (us-ashburn-1, us-phoenix-1, eu-frankfurt-1, ap-hyderabad-1, ap-osaka-1) set the
+# provider baseURL option in the consuming tool.
+# Reasoning: reasoning_effort = low|medium|high on gpt-oss and Grok 4.3/4.6; the Grok 4.20
+# reasoning variants reject reasoning_effort (400). gpt-oss streams reasoning as
+# `reasoning_content` deltas. Streaming clients must accept gzip (the endpoint negotiates
+# zstd and compresses each streamed chunk as a separate frame).
+# Models below are the ones Oracle lists for API-key access to this endpoint (Meta Llama,
+# xAI Grok, OpenAI gpt-oss); Cohere models are served by the native OCI API only.
+# Sources:
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/openai-compatible-api.htm
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/api-keys.htm
+# https://docs.oracle.com/en-us/iaas/Content/generative-ai/pretrained-models.htm
+name = "OCI Generative AI"
+env = ["OCI_GENAI_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1"
+doc = "https://docs.oracle.com/en-us/iaas/Content/generative-ai/pretrained-models.htm"

From c7f79c4700dfdfe77531b0b9ad0e3eead58a3987 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=AD=A4=E6=96=B9=E4=B8=80=E6=B3=89?= <lg_329@163.com>
Date: Thu, 17 Sep 2026 08:50:33 +0800
Subject: [PATCH 1477/2012] fix(alibaba-cn): refresh Bailuan qwen3 prices and
 split CN coder-plus from intl symlink (#6723)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Verified against help.aliyun.com/zh/model-studio/model-pricing on 2026-09-11:

- qwen3-max: ¥9/54 base band (was launch ¥6/24), ¥15/90 over 128K as a tier
- qwen3.5-plus: ¥2/12 base band (was ¥4/24), ¥8/48 over 256K as a tier
- qwen3.5-flash: ¥1.2/7.2 base band (was ¥1.2/12), ¥4.8/28.8 over 256K
- qwen-vl-ocr: current alias price ¥0.3/0.5 (file held deprecated ¥5/5 snapshot)
- qwen3-coder-plus: replace the symlink to the international entry ($1/$5)
  with a CN-specific entry — Bailuan now charges ¥4/16 (0-32K), ¥6/24,
  ¥10/40, ¥20/200 across context bands

CNY→USD at the 0.1435 convention already used by the other alibaba-cn files.

Co-authored-by: guanchen.lin <guanchen.lin@kunlun-inc.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
---
 providers/alibaba-cn/models/qwen-vl-ocr.toml  |  6 +--
 .../alibaba-cn/models/qwen3-coder-plus.toml   | 42 ++++++++++++++++++-
 providers/alibaba-cn/models/qwen3-max.toml    | 11 +++--
 .../alibaba-cn/models/qwen3.5-flash.toml      | 12 ++++--
 providers/alibaba-cn/models/qwen3.5-plus.toml | 14 +++++--
 5 files changed, 71 insertions(+), 14 deletions(-)
 mode change 120000 => 100644 providers/alibaba-cn/models/qwen3-coder-plus.toml

diff --git a/providers/alibaba-cn/models/qwen-vl-ocr.toml b/providers/alibaba-cn/models/qwen-vl-ocr.toml
index fbef55eae88..2f10e823a50 100644
--- a/providers/alibaba-cn/models/qwen-vl-ocr.toml
+++ b/providers/alibaba-cn/models/qwen-vl-ocr.toml
@@ -2,7 +2,7 @@ name = "Qwen-VL OCR"
 description = "OCR model for extracting structured text from documents and screenshots"
 family = "qwen"
 release_date = "2024-10-28"
-last_updated = "2025-04-13"
+last_updated = "2026-09-11"
 attachment = false
 reasoning = false
 temperature = true
@@ -11,8 +11,8 @@ tool_call = false
 open_weights = false
 
 [cost]
-input = 0.717
-output = 0.717
+input = 0.043
+output = 0.072
 
 [limit]
 context = 34_096
diff --git a/providers/alibaba-cn/models/qwen3-coder-plus.toml b/providers/alibaba-cn/models/qwen3-coder-plus.toml
deleted file mode 120000
index f80009ca205..00000000000
--- a/providers/alibaba-cn/models/qwen3-coder-plus.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../alibaba/models/qwen3-coder-plus.toml
\ No newline at end of file
diff --git a/providers/alibaba-cn/models/qwen3-coder-plus.toml b/providers/alibaba-cn/models/qwen3-coder-plus.toml
new file mode 100644
index 00000000000..0491bf65275
--- /dev/null
+++ b/providers/alibaba-cn/models/qwen3-coder-plus.toml
@@ -0,0 +1,41 @@
+# China-site (Bailuan) pricing differs from the international console,
+# so this provider entry can no longer share the alibaba symlink ($1/$5).
+# https://help.aliyun.com/zh/model-studio/model-pricing (accessed 2026-09-11)
+name = "Qwen3 Coder Plus"
+description = "Hosted Qwen coder for software agents, repo edits, and long-context code"
+family = "qwen"
+release_date = "2025-07-23"
+last_updated = "2026-09-11"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.574
+output = 2.296
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.861
+output = 3.444
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 1.435
+output = 5.74
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 2.87
+output = 28.7
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/alibaba-cn/models/qwen3-max.toml b/providers/alibaba-cn/models/qwen3-max.toml
index 78f1355eab1..ffb4eb4582b 100644
--- a/providers/alibaba-cn/models/qwen3-max.toml
+++ b/providers/alibaba-cn/models/qwen3-max.toml
@@ -2,7 +2,7 @@ name = "Qwen3 Max"
 description = "Flagship Qwen model for complex reasoning, coding, and agentic workflows"
 family = "qwen"
 release_date = "2025-09-23"
-last_updated = "2025-09-23"
+last_updated = "2026-09-11"
 attachment = false
 reasoning = false
 temperature = true
@@ -11,8 +11,13 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0.861
-output = 3.441
+input = 1.291
+output = 7.749
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 2.153
+output = 12.915
 
 [limit]
 context = 262_144
diff --git a/providers/alibaba-cn/models/qwen3.5-flash.toml b/providers/alibaba-cn/models/qwen3.5-flash.toml
index 6cbf6860ad0..51651ed2e75 100644
--- a/providers/alibaba-cn/models/qwen3.5-flash.toml
+++ b/providers/alibaba-cn/models/qwen3.5-flash.toml
@@ -2,7 +2,7 @@ name = "Qwen3.5 Flash"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
 release_date = "2026-02-23"
-last_updated = "2026-02-23"
+last_updated = "2026-09-11"
 attachment = true
 reasoning = true
 structured_output = true
@@ -20,8 +20,14 @@ max = 81_920
 
 [cost]
 input = 0.172
-output = 1.72
-reasoning = 1.72
+output = 1.033
+reasoning = 1.033
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.689
+output = 4.133
+reasoning = 4.133
 
 [limit]
 context = 1_000_000
diff --git a/providers/alibaba-cn/models/qwen3.5-plus.toml b/providers/alibaba-cn/models/qwen3.5-plus.toml
index 732e524c4a4..56105ed505b 100644
--- a/providers/alibaba-cn/models/qwen3.5-plus.toml
+++ b/providers/alibaba-cn/models/qwen3.5-plus.toml
@@ -2,7 +2,7 @@ name = "Qwen3.5 Plus"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
 release_date = "2026-02-16"
-last_updated = "2026-02-16"
+last_updated = "2026-09-11"
 attachment = false
 reasoning = true
 temperature = true
@@ -18,9 +18,15 @@ type = "budget_tokens"
 max = 81_920
 
 [cost]
-input = 0.573
-output = 3.44
-reasoning = 3.44
+input = 0.287
+output = 1.722
+reasoning = 1.722
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 1.148
+output = 6.888
+reasoning = 6.888
 
 [limit]
 context = 1_000_000

From fc05b533def7b6bf8e84312434e4ef3be58dd048 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 01:00:05 +0000
Subject: [PATCH 1478/2012] chore(sync): update Kilo model catalog (#7282)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 1 -
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml    | 4 ++--
 providers/kilo/models/tencent/hy3.toml                    | 6 +++---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml      | 4 ++--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/kilo/models/~z-ai/glm-flash-latest.toml         | 6 +++---
 6 files changed, 14 insertions(+), 15 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 25c72a0b0fb..212e280b8a9 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 202_800
-output = 182_520
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index f491c2853c8..673b730eb39 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_000_000
-output = 393_216
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 5d5fb4fce66..e4ce4e5dfe6 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 1e20864604ceeae6974a65a6b34255b1f6c34db4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 01:00:10 +0000
Subject: [PATCH 1479/2012] chore(sync): update OpenRouter model catalog
 (#7281)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 .../openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml | 5 ++---
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml       | 6 +++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-flash-latest.toml   | 6 +++---
 9 files changed, 26 insertions(+), 28 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f7dc249d032..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0763
-output = 0.1526
-cache_read = 0.01526
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
index bf732029012..297dfb56987 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -7,9 +7,8 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.05
-output = 0.2
-cache_read = 0.03
+input = 0.06
+output = 0.24
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 5dcd117448e..f471e71439f 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
+input = 0.625
+output = 3.125
+cache_read = 0.1875
 
 [limit]
 context = 262_144
-output = 182_520
+output = 32_768
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index 4dac1f2fc80..f82829f96ea 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.09
+output = 0.3
+cache_read = 0.018
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 98fd23a037b..dc34ba0ccf7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 644122b25af..c32fca67d4b 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.07
-output = 0.2333
-cache_read = 0.014
+input = 0.075
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 1_310_720

From 71c38581c7121b865f7ade922348aca686045a4a Mon Sep 17 00:00:00 2001
From: islomjon-aiand <islomjon@aiand.com>
Date: Thu, 17 Sep 2026 10:19:19 +0900
Subject: [PATCH 1480/2012] feat(sync): add ai& sync module with auto-merge
 allowlist and motif family (#7040)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(sync): add ai& sync module with auto-merge allowlist and motif family

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* chore(sync): regenerate aiand catalog on rebased dev

Rerun of the ai& sync after cherry-picking onto current dev.
Interleaved reasoning_content blocks from #6839 are preserved by the
adapter (existing?.interleaved); motif-3 factors onto the new lab entry.

* docs(aiand): refresh kimi-k3 and gpt-oss-120b probe headers

Live reasoning_effort probes on 2026-09-14 (negative controls included)
confirm the authored effort sets are unchanged on the gateway.

* fix(aiand): factor motif-3 override-only onto its lab entry

The rebased regen kept dev's stale inline lab fields (open_weights = false,
release_date 2026-08-12) as overrides against the new lab file. Lab facts now
inherit from models/motif-technologies/motif-3.toml; the [interleaved] block
from #6839 is preserved.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): clear curated deprecated on omission; silence empty skip notice

The feed owns deprecation and absence means active, so a curated deprecated
no longer survives a feed that omits status (alpha/beta still do). A clean
sync no longer emits a zero-count skip notice into the report.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): publish interleaved for new reasoners, route lab-metadata skips to missing-model issues

Parse the feed's interleaved (authored fallback, gateway default of
message.reasoning_content for reasoners) so a new reasoner is never created
without its side channel; return skipped ids from missingModelID so the runner
preserves local entries and opens deduped issues. qwen3.8 header records the
2026-09-14 probe including the announced high→xhigh substitution.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): read overrides from the authored TOML, refresh all headers to the 2026-09-14 probes

Lab-owned fields on factored files come only from deltas the authored file
already carried on its base_model (context.authored, never the base-resolved
merge), so a full-inline file factored for the first time inherits the lab
entry instead of re-emitting stale values as overrides — the Motif regression
in general form. Host fields stay feed-authoritative. All 11 headers now cite
the 2026-09-14 probe matrix (11 models x 7 effort values).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): stop the runner re-injecting pre-factor descriptions; seed created files with a reasoning-wire header

preserveDescriptions is off so a full-inline → factored transition cannot
recreate a lab-identical description override (runner-level test covers the
Motif shape end to end). translateModel now returns a leading header for
creates naming the reasoning wire path, so a toggle control is never written
without it.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): fail a control-less reasoner for review instead of letting [] be written

A reasoner whose feed yields no schema-valid reasoning_options and has no
authored set to keep now throws MissingReasoningOptionsError: the runner
preserves the local file, reports the reason, and routes the id to the
missing-model issue flow. An explicit [] from the feed remains the feed's own
assertion. Runner-level test covers the untouched-file path.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(aiand): never write toggle or budget controls — ai& has a single reasoning wire path

reasoning_effort ("none" = off) is the only control this host exposes, so
toggle and budget_tokens are parsed for tolerance but dropped from feed and
authored options alike; an emptied set falls through to the existing
fail-for-review path. The create header states the single wire path instead
of describing controls the host does not have.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: fenil modi <fenilmodi4820@gmail.com>
---
 models/motif-technologies/motif-3.toml        |  33 ++
 package.json                                  |   1 +
 packages/core/src/family.ts                   |   3 +
 packages/core/src/sync/auto-merge.ts          |   1 +
 packages/core/src/sync/index.ts               |   5 +-
 packages/core/src/sync/providers/aiand.ts     | 401 ++++++++++++++++
 packages/core/test/aiand.test.ts              | 431 ++++++++++++++++++
 packages/core/test/auto-merge.test.ts         |   2 +-
 .../models/deepseek-ai/deepseek-v4-flash.toml |   5 +-
 .../models/deepseek-ai/deepseek-v4-pro.toml   |   5 +-
 .../aiand/models/google/gemma-4-31b-it.toml   |   4 +-
 .../models/moonshotai/kimi-k2.7-code.toml     |  14 +-
 .../aiand/models/moonshotai/kimi-k3.toml      |  12 +-
 .../models/motif-technologies/motif-3.toml    |  33 +-
 .../aiand/models/openai/gpt-oss-120b.toml     |   4 +-
 providers/aiand/models/qwen/qwen3.6-27b.toml  |   5 +-
 providers/aiand/models/qwen/qwen3.8-27b.toml  |   7 +-
 providers/aiand/models/zai-org/glm-5.2.toml   |   5 +-
 providers/aiand/models/zai-org/glm-5.3.toml   |   5 +-
 sync.md                                       |  13 +
 20 files changed, 936 insertions(+), 53 deletions(-)
 create mode 100644 models/motif-technologies/motif-3.toml
 create mode 100644 packages/core/src/sync/providers/aiand.ts
 create mode 100644 packages/core/test/aiand.test.ts

diff --git a/models/motif-technologies/motif-3.toml b/models/motif-technologies/motif-3.toml
new file mode 100644
index 00000000000..c42517fba3d
--- /dev/null
+++ b/models/motif-technologies/motif-3.toml
@@ -0,0 +1,33 @@
+# Source: https://huggingface.co/Motif-Technologies/Motif-3 (accessed
+# 2026-09-14): repo created 2026-08-07, 155 safetensors shards, ungated,
+# MIT license.
+# tool_call: the model card documents agentic tool use and ships a vLLM
+# --tool-call-parser motif; live probe on 2026-09-14 returned a real
+# tool_calls response (finish_reason = tool_calls).
+# temperature: the card evaluates at sampling temperature 1.0; live probe on
+# 2026-09-14 accepted 0.2 and 1.7, with 1.7 visibly changing the output.
+name = "Motif 3"
+family = "motif"
+description = "Motif 3 is a large-scale, decoder-only Mixture-of-Experts (MoE) language model with 314 billion total parameters and 13.2 billion parameters activated per token."
+release_date = "2026-08-07"
+last_updated = "2026-08-07"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+license = "MIT"
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Model weights"
+url = "https://huggingface.co/Motif-Technologies/Motif-3"
+format = "safetensors"
diff --git a/package.json b/package.json
index a1de7f5bb8e..099cbc05e8b 100644
--- a/package.json
+++ b/package.json
@@ -38,6 +38,7 @@
     "vercel:generate": "bun ./packages/core/script/sync-models.ts vercel",
     "wandb:generate": "bun ./packages/core/script/sync-models.ts wandb",
     "digitalocean:sync": "bun ./packages/core/script/sync-models.ts digitalocean",
+    "aiand:sync": "bun ./packages/core/script/sync-models.ts aiand",
     "fireworks:sync": "bun ./packages/core/script/sync-models.ts fireworks-ai",
     "ambient:sync": "bun ./packages/core/script/sync-models.ts ambient",
     "models:sync": "bun ./packages/core/script/sync-models.ts",
diff --git a/packages/core/src/family.ts b/packages/core/src/family.ts
index 9f719d80376..c7f641fe8ca 100644
--- a/packages/core/src/family.ts
+++ b/packages/core/src/family.ts
@@ -84,6 +84,9 @@ export const ModelFamilyValues = [
   "kimi-free",
   "kimi-thinking",
 
+  // Motif Technologies
+  "motif",
+
   // Poolside Laguna
   "laguna",
   "laguna-s",
diff --git a/packages/core/src/sync/auto-merge.ts b/packages/core/src/sync/auto-merge.ts
index 244d2e277d3..12629b95c9e 100644
--- a/packages/core/src/sync/auto-merge.ts
+++ b/packages/core/src/sync/auto-merge.ts
@@ -5,6 +5,7 @@ export const MAX_CREATED_MODELS = 10;
 export const MAX_DELETED_MODELS = 10;
 export const MAX_MODEL_CHURN = 15;
 const REVIEWED_REASONING_PROVIDERS = new Set([
+  "aiand",
   "crossmodel",
   "edenai",
   "empiriolabs",
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 9156827c689..5095f4124ef 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -6,6 +6,7 @@ import { z } from "zod";
 import { AuthoredModel, AuthoredModelShape, ModelMetadata } from "../schema.js";
 import { openMissingModelIssues } from "./missing-issues.js";
 import { MissingReasoningOptionsError } from "./missing-reasoning-options.js";
+import { aiand } from "./providers/aiand.js";
 import { ambient } from "./providers/ambient.js";
 import { anthropic } from "./providers/anthropic.js";
 import { baseten } from "./providers/baseten.js";
@@ -139,6 +140,7 @@ export interface SyncResult {
 }
 
 export const providers: {
+  aiand: SyncProvider<any>;
   ambient: SyncProvider<any>;
   anthropic: SyncProvider<any>;
   baseten: SyncProvider<any>;
@@ -177,6 +179,7 @@ export const providers: {
   wandb: SyncProvider<any>;
   xai: SyncProvider<any>;
 } = {
+  aiand,
   ambient,
   anthropic,
   baseten,
@@ -234,7 +237,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-ai-gateway", "cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "fireworks-ai", "friendli", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["aiand", "ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "fireworks-ai", "friendli", "github-copilot", "google", "hyper", "meta", "ollama-cloud", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;
diff --git a/packages/core/src/sync/providers/aiand.ts b/packages/core/src/sync/providers/aiand.ts
new file mode 100644
index 00000000000..9545bd91e0f
--- /dev/null
+++ b/packages/core/src/sync/providers/aiand.ts
@@ -0,0 +1,401 @@
+import { readdirSync } from "node:fs";
+import path from "node:path";
+import { z } from "zod";
+
+import { ModelFamily } from "../../family.js";
+import { ReasoningOption as CatalogReasoningOption } from "../../schema.js";
+import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
+import { MissingReasoningOptionsError } from "../missing-reasoning-options.js";
+import { factorBaseModel } from "./openrouter.js";
+
+const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+
+// ai&'s /v1/api.json publishes this repo's shape directly; the `aiand` key is
+// the attributable provider entry and lists only schema-complete models, so
+// translation is near-identity. Reference:
+// https://api.aiand.com/v1/api.json
+const API_ENDPOINT = process.env.AIAND_API_URL ?? "https://api.aiand.com/v1/api.json";
+
+const MODALITIES = ["text", "audio", "image", "video", "pdf"] as const;
+
+const GATEWAY_INTERLEAVED = { field: "reasoning_content" } as const;
+const INTERLEAVED_FIELDS = ["reasoning_content", "reasoning_details"] as const;
+type Modality = (typeof MODALITIES)[number];
+
+// The feed publishes the catalog's reasoning_options shape. Effort values are
+// parsed leniently (any string) so a vocabulary the schema doesn't know yet
+// is filtered per value instead of aborting the whole feed parse; toggle and
+// budget_tokens controls take the shared catalog schema as-is.
+const FeedReasoningOption = z.union([
+  z.object({ type: z.literal("effort"), values: z.array(z.string()) }).passthrough(),
+  CatalogReasoningOption,
+]);
+
+export const AiandModel = z
+  .object({
+    id: z.string().min(1),
+    name: z.string().min(1),
+    description: z.string().min(1).optional(),
+    family: z.string().optional(),
+    release_date: z.string(),
+    last_updated: z.string().optional(),
+    attachment: z.boolean(),
+    reasoning: z.boolean(),
+    reasoning_options: z.array(FeedReasoningOption).optional(),
+    temperature: z.boolean(),
+    tool_call: z.boolean(),
+    structured_output: z.boolean().optional(),
+    cost: z
+      .object({
+        input: z.number().nonnegative(),
+        output: z.number().nonnegative(),
+        cache_read: z.number().nonnegative().optional(),
+      })
+      .passthrough(),
+    limit: z
+      .object({
+        context: z.number().int().positive(),
+        output: z.number().int().positive(),
+      })
+      .passthrough(),
+    modalities: z
+      .object({
+        input: z.array(z.string()),
+        output: z.array(z.string()),
+      })
+      .passthrough(),
+    open_weights: z.boolean().optional(),
+    status: z.enum(["alpha", "beta", "deprecated"]).optional(),
+    interleaved: z.union([z.boolean(), z.object({ field: z.string() }).passthrough()]).optional(),
+  })
+  .passthrough();
+
+export const AiandResponse = z
+  .object({
+    aiand: z.object({ models: z.record(AiandModel) }).passthrough(),
+  })
+  .passthrough();
+
+export type AiandModel = z.infer<typeof AiandModel>;
+
+export const aiand = {
+  id: "aiand",
+  name: "ai&",
+  modelsDir: "providers/aiand/models",
+  // Factoring is explicit (factorBaseModel) so files stay override-only; the
+  // runner's default preservation keeps the base_model reference but does not
+  // drop fields identical to the base.
+  preserveBaseModels: false,
+  // The runner would otherwise re-inject the pre-factor authored description
+  // whenever the translator leaves it unset — recreating a lab-identical
+  // override on every full-inline → factored transition.
+  preserveDescriptions: false,
+  async fetchModels() {
+    const response = await fetch(API_ENDPOINT);
+    if (!response.ok) {
+      throw new Error(`ai& catalog request failed: ${response.status} ${response.statusText}`);
+    }
+    return response.json();
+  },
+  parseModels(raw) {
+    const models = Object.values(AiandResponse.parse(raw).aiand.models);
+    // deleteMissing runs at the runner default: an empty or truncated feed
+    // must fail loudly here rather than read as "delete the local catalog".
+    if (models.length === 0) {
+      throw new Error("ai& catalog returned no models; refusing an empty feed as authoritative");
+    }
+    return models;
+  },
+  translateModel(model, context) {
+    // `authored` is the raw TOML — the only thing that can carry a genuine
+    // override. `existing` is the base-resolved merge and would make inherited
+    // lab values look hand-written.
+    const authored = context.authored(model.id);
+    const baseModel = authored?.base_model ?? resolveAiandBaseModel(model.id, model.name);
+    // A new feed model with no resolvable lab entry must never be written as
+    // an unfactored full definition; it goes to the missing-model issue flow
+    // instead. An existing local file is always translated — skipping one
+    // would delete it.
+    if (authored === undefined && baseModel === undefined) return undefined;
+    const built = buildAiandModel(model, authored, baseModel);
+    // Existing headers win; this only seeds a create, and a toggle control
+    // must never be written without its wire path.
+    return { id: model.id, model: built, header: reasoningHeader(built) };
+  },
+  sourceID(model) {
+    return model.id;
+  },
+  // The only skip is "no lab base yet", never an intentional removal, so every
+  // skip enters the missing-model issue flow: the runner keeps any existing
+  // local entry and opens one deduped issue for the lab metadata.
+  missingModelID(model) {
+    return model.id;
+  },
+  skippedNotice(ids) {
+    if (ids.length === 0) return [];
+    return [
+      `Skipped ${ids.length} feed model(s) with no resolvable base model (missing-model issue flow): ${ids.join(", ")}`,
+    ];
+  },
+} satisfies SyncProvider<AiandModel>;
+
+type HostFields = Omit<
+  SyncedFullModel,
+  "name" | "description" | "family" | "release_date" | "last_updated" | "knowledge" | "open_weights"
+>;
+
+/**
+ * `authored` is the provider TOML as written (not base-resolved). Host facts —
+ * pricing, limits, controls, capability flags, modalities, status, the side
+ * channel — are feed-authoritative. Lab-owned facts are never asserted from
+ * the gateway on a factored file: they appear only as deltas the authored
+ * file already carried on top of its base_model, so a full-inline file being
+ * factored for the first time (or a brand-new file) inherits the lab entry
+ * outright instead of re-emitting the lab's values as overrides.
+ */
+export function buildAiandModel(
+  model: AiandModel,
+  authored: ExistingModel | undefined,
+  baseModel: string | null | undefined = authored?.base_model ?? resolveAiandBaseModel(model.id, model.name),
+  today = new Date().toISOString().slice(0, 10),
+): SyncedModel {
+  const limit = {
+    context: model.limit.context,
+    input: authored?.limit?.input,
+    output: model.limit.output,
+  };
+  const host: HostFields = {
+    attachment: model.attachment,
+    reasoning: model.reasoning,
+    // The schema refuses reasoning_options on a non-reasoner; a non-reasoning
+    // feed model must not stall the sync on that refine.
+    reasoning_options: model.reasoning ? requireReasoningOptions(model, authored) : undefined,
+    tool_call: model.tool_call,
+    structured_output: model.structured_output,
+    temperature: model.temperature,
+    // Canonical order so a feed-side reordering never churns a TOML.
+    modalities: {
+      input: sortModalities(model.modalities.input.filter(isModality)),
+      output: sortModalities(model.modalities.output.filter(isModality)),
+    },
+    limit,
+    cost: {
+      input: model.cost.input,
+      output: model.cost.output,
+      reasoning: authored?.cost?.reasoning,
+      cache_read: model.cost.cache_read,
+      cache_write: authored?.cost?.cache_write,
+      input_audio: authored?.cost?.input_audio,
+      output_audio: authored?.cost?.output_audio,
+      tiers: authored?.cost?.tiers,
+    },
+    // Absence means active on the feed, and the feed owns deprecation: a
+    // curated alpha/beta survives omission, a curated deprecated does not,
+    // or a route the gateway reactivated would stay marked retired forever.
+    status: model.status ?? (authored?.status === "deprecated" ? undefined : authored?.status),
+    // Every ai& reasoner streams its thinking in message.reasoning_content —
+    // a gateway-wide side channel — so a brand-new reasoner gets it even
+    // before the feed publishes `interleaved` itself.
+    interleaved: model.reasoning
+      ? (normalizeInterleaved(model.interleaved) ?? authored?.interleaved ?? GATEWAY_INTERLEAVED)
+      : undefined,
+  };
+
+  if (baseModel != null) {
+    // Only a file that already sat on this base can carry genuine lab-field
+    // deltas; anything else inherits the lab entry.
+    const deltas = authored?.base_model === baseModel ? authored : undefined;
+    return factorBaseModel(
+      baseModel,
+      {
+        ...host,
+        name: deltas?.name,
+        description: deltas?.description,
+        release_date: deltas?.release_date,
+        last_updated: deltas?.last_updated,
+        knowledge: deltas?.knowledge,
+        open_weights: deltas?.open_weights,
+      },
+      limit,
+      deltas?.base_model_omit,
+    );
+  }
+
+  // Standalone (no lab entry anywhere): curated values win for lab-owned
+  // fields, the feed fills the rest. Unknown families degrade to the curated
+  // value rather than fail validation.
+  const family = ModelFamily.safeParse(model.family);
+  return {
+    ...host,
+    name: authored?.name ?? model.name,
+    description: authored?.description ?? model.description ?? model.name,
+    family: family.success ? family.data : authored?.family,
+    // Release dates are lab metadata the gateway is not authoritative for.
+    release_date: authored?.release_date ?? model.release_date,
+    // The feed's last_updated tracks catalog-row edits, not model revisions.
+    last_updated: authored?.last_updated ?? model.last_updated ?? today,
+    knowledge: authored?.knowledge,
+    open_weights: model.open_weights ?? authored?.open_weights ?? false,
+  };
+}
+
+const DOCS_URL = "https://docs.aiand.com/models/catalog/";
+
+/**
+ * Leading comment block for a created file. ai& exposes one reasoning control
+ * — `reasoning_effort` on /v1/chat/completions and `reasoning.effort` on
+ * /v1/responses, enforced per model, with "none" as off — so the header names
+ * that path and states that no separate toggle or token budget exists.
+ */
+function reasoningHeader(model: SyncedModel): string | undefined {
+  const options = model.reasoning_options;
+  if (options === undefined || options.length === 0) return undefined;
+  const lines = [
+    `# Pricing: GET https://api.aiand.com/v1/api.json (synced hourly by the aiand module)`,
+    `# Docs: ${DOCS_URL}`,
+  ];
+  for (const option of options) {
+    if (option.type === "effort" && option.values.length > 0) {
+      lines.push(`# Effort: reasoning_effort = ${option.values.map((value) => `"${value}"`).join(" | ")}`);
+    }
+  }
+  lines.push(
+    "# Controls: reasoning_effort only (`reasoning_effort` on /v1/chat/completions, `reasoning.effort` on /v1/responses; \"none\" = off) — no separate toggle or token-budget field on this host.",
+  );
+  lines.push("# Reasoning side channel: message.reasoning_content");
+  return `${lines.join("\n")}\n`;
+}
+
+interface MetadataEntry {
+  id: string;
+  normalizedFull: string;
+  normalizedFilename: string;
+}
+
+let metadataEntries: MetadataEntry[] | undefined;
+
+/**
+ * ai& publishes lab-prefixed ids ("deepseek-ai/deepseek-v4-flash") whose lab
+ * segment can differ from the models/ directory ("deepseek/…"), so new ids
+ * resolve like Venice's: a unique normalized match on the full id first, then
+ * on the filename alone.
+ */
+export function resolveAiandBaseModel(id: string, name: string): string | undefined {
+  const entries = getMetadataEntries();
+  for (const candidate of [...new Set([id, name])]) {
+    const normalizedFull = normalize(candidate);
+    const normalizedFilename = normalize(candidate.split("/").pop() ?? candidate);
+    const ranked = [
+      entries.filter((entry) => entry.normalizedFull === normalizedFull),
+      entries.filter((entry) => entry.normalizedFilename === normalizedFilename),
+    ];
+    const match = ranked.find((matches) => matches.length === 1)?.[0]?.id;
+    if (match !== undefined) return match;
+  }
+  return undefined;
+}
+
+function getMetadataEntries() {
+  if (metadataEntries !== undefined) return metadataEntries;
+  metadataEntries = [];
+  for (const provider of readdirSync(MODELS_DIR, { withFileTypes: true })) {
+    if (!provider.isDirectory()) continue;
+    for (const file of readdirSync(path.join(MODELS_DIR, provider.name), { withFileTypes: true })) {
+      if (!file.isFile() || !file.name.endsWith(".toml")) continue;
+      const filename = file.name.slice(0, -5);
+      metadataEntries.push({
+        id: `${provider.name}/${filename}`,
+        normalizedFull: normalize(`${provider.name}/${filename}`),
+        normalizedFilename: normalize(filename),
+      });
+    }
+  }
+  return metadataEntries;
+}
+
+function normalize(value: string) {
+  return value.toLowerCase().replaceAll(/[^a-z0-9]/g, "");
+}
+
+type ReasoningEffortValue = Extract<
+  Extract<z.infer<typeof CatalogReasoningOption>, { type: "effort" }>["values"][number],
+  string
+>;
+
+function isReasoningEffort(value: string): value is ReasoningEffortValue {
+  return CatalogReasoningOption.safeParse({ type: "effort", values: [value] }).success;
+}
+
+/**
+ * A reasoner must never be written with an invented empty control set: `[]`
+ * means "no caller control", not uncertainty (AGENTS.md → Reasoning options).
+ * When the feed yields no schema-valid controls and nothing authored can be
+ * kept, the model fails sync for manual review — the runner preserves the
+ * local file and routes the id to the missing-model issue flow.
+ */
+function requireReasoningOptions(
+  model: AiandModel,
+  authored: ExistingModel | undefined,
+): SyncedFullModel["reasoning_options"] {
+  const options = resolveReasoningOptions(model, authored);
+  if (options !== undefined) return options;
+  throw new MissingReasoningOptionsError(
+    model.id,
+    "feed publishes reasoning = true without a schema-valid reasoning_options set and no authored controls exist to keep; research the ai& effort set before listing",
+  );
+}
+
+/**
+ * Omit, empty, and vocabulary-miss are three different assertions: an omitted
+ * feed list asserts nothing (authored options stay), an explicit [] asserts
+ * "no caller controls", and a non-empty list whose values the schema doesn't
+ * know yet keeps the authored options rather than inventing "no control" and
+ * auto-merging it.
+ */
+function resolveReasoningOptions(
+  model: AiandModel,
+  authored: ExistingModel | undefined,
+): SyncedFullModel["reasoning_options"] {
+  if (model.reasoning_options === undefined) return authoredReasoningOptions(authored);
+  if (model.reasoning_options.length === 0) return [];
+  const feed = model.reasoning_options.flatMap((option) => {
+    // ai& has exactly one reasoning control — `reasoning_effort`, enforced per
+    // model, where "off" is the "none" level. A toggle or token budget cannot
+    // be true of this host, so neither is ever written; parsing them keeps a
+    // feed that publishes one from aborting the run.
+    if (option.type !== "effort") return [];
+    const values = option.values.filter(
+      (value): value is ReasoningEffortValue => typeof value === "string" && isReasoningEffort(value),
+    );
+    return values.length > 0 ? [{ type: "effort" as const, values }] : [];
+  });
+  return feed.length > 0 ? feed : authoredReasoningOptions(authored);
+}
+
+function authoredReasoningOptions(
+  authored: ExistingModel | undefined,
+): SyncedFullModel["reasoning_options"] {
+  const options = (authored?.reasoning_options ?? [])
+    .map((option) => CatalogReasoningOption.safeParse(option))
+    .flatMap((result) => (result.success && result.data.type === "effort" ? [result.data] : []));
+  return options.length > 0 ? options : undefined;
+}
+
+function isModality(value: string): value is Modality {
+  return (MODALITIES as readonly string[]).includes(value);
+}
+
+function normalizeInterleaved(
+  value: AiandModel["interleaved"],
+): SyncedFullModel["interleaved"] | undefined {
+  if (value === true) return true;
+  if (value === undefined || value === false) return undefined;
+  const field = value.field;
+  return (INTERLEAVED_FIELDS as readonly string[]).includes(field)
+    ? { field: field as (typeof INTERLEAVED_FIELDS)[number] }
+    : undefined;
+}
+
+function sortModalities(values: Modality[]): Modality[] {
+  return [...new Set(values)].sort((a, b) => MODALITIES.indexOf(a) - MODALITIES.indexOf(b));
+}
diff --git a/packages/core/test/aiand.test.ts b/packages/core/test/aiand.test.ts
new file mode 100644
index 00000000000..d4e0c74c1fb
--- /dev/null
+++ b/packages/core/test/aiand.test.ts
@@ -0,0 +1,431 @@
+import { expect, test } from "bun:test";
+import { copyFile, mkdir, mkdtemp, readFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import path from "node:path";
+
+import { syncProvider } from "../src/sync/index.js";
+import { MissingReasoningOptionsError } from "../src/sync/missing-reasoning-options.js";
+
+import {
+  aiand,
+  AiandModel,
+  AiandResponse,
+  buildAiandModel,
+  resolveAiandBaseModel,
+} from "../src/sync/providers/aiand.js";
+
+function aiandModel(overrides: Partial<AiandModel> = {}): AiandModel {
+  return {
+    id: "deepseek-ai/deepseek-v4-flash",
+    name: "deepseek-ai/DeepSeek-V4-Flash",
+    description: "Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work",
+    family: "deepseek",
+    release_date: "2026-04-24",
+    last_updated: "2026-09-14",
+    attachment: false,
+    reasoning: true,
+    reasoning_options: [{ type: "effort", values: ["none", "high", "max"] }],
+    temperature: true,
+    tool_call: true,
+    structured_output: true,
+    cost: { input: 0.15, output: 0.25, cache_read: 0.08 },
+    limit: { context: 1_048_576, output: 384_000 },
+    modalities: { input: ["text"], output: ["text"] },
+    open_weights: true,
+    ...overrides,
+  };
+}
+
+test("translates a feed model near-identity when nothing is authored", () => {
+  const built = buildAiandModel(aiandModel(), undefined, null);
+  expect(built).toMatchObject({
+    name: "deepseek-ai/DeepSeek-V4-Flash",
+    family: "deepseek",
+    release_date: "2026-04-24",
+    reasoning: true,
+    reasoning_options: [{ type: "effort", values: ["none", "high", "max"] }],
+    structured_output: true,
+    cost: { input: 0.15, output: 0.25, cache_read: 0.08 },
+    limit: { context: 1_048_576, output: 384_000 },
+    open_weights: true,
+  });
+});
+
+test("curated name, knowledge, and last_updated win over the feed", () => {
+  const built = buildAiandModel(
+    aiandModel(),
+    { name: "DeepSeek V4 Flash", knowledge: "2025-05", last_updated: "2026-07-31" },
+    null,
+  );
+  expect(built.name).toBe("DeepSeek V4 Flash");
+  expect(built.knowledge).toBe("2025-05");
+  expect(built.last_updated).toBe("2026-07-31");
+});
+
+test("the feed's last_updated is row-edit noise: today only fills a blank", () => {
+  const built = buildAiandModel(aiandModel({ last_updated: undefined }), undefined, null, "2026-09-14");
+  expect(built.last_updated).toBe("2026-09-14");
+});
+
+test("a curated release_date wins over the feed's", () => {
+  const built = buildAiandModel(
+    aiandModel({ release_date: "2026-05-01" }),
+    { release_date: "2026-04-24" },
+    null,
+  );
+  expect(built.release_date).toBe("2026-04-24");
+});
+
+test("a vocabulary miss preserves authored reasoning options instead of publishing none", () => {
+  const built = buildAiandModel(
+    aiandModel({ reasoning_options: [{ type: "effort", values: ["turbo"] }] }),
+    { reasoning_options: [{ type: "effort", values: ["high"] }] },
+    null,
+  );
+  expect(built.reasoning_options).toEqual([{ type: "effort", values: ["high"] }]);
+});
+
+test("a family the enum does not know falls back to the curated value", () => {
+  const unknown = buildAiandModel(aiandModel({ family: "not-a-family" }), undefined, null);
+  expect(unknown.family).toBeUndefined();
+  const curated = buildAiandModel(aiandModel({ family: "not-a-family" }), { family: "deepseek" }, null);
+  expect(curated.family).toBe("deepseek");
+});
+
+test("unknown modalities are dropped", () => {
+  const built = buildAiandModel(
+    aiandModel({ modalities: { input: ["text", "smell"], output: ["text"] } }),
+    undefined,
+    null,
+  );
+  expect(built.modalities?.input).toEqual(["text"]);
+});
+
+test("a reasoner with no schema-valid controls and nothing authored fails instead of writing []", () => {
+  const vocabularyMiss = aiandModel({ reasoning_options: [{ type: "effort", values: ["turbo"] }] });
+  expect(() => buildAiandModel(vocabularyMiss, undefined, null)).toThrow(MissingReasoningOptionsError);
+  const omitted = aiandModel({ reasoning_options: undefined });
+  expect(() => buildAiandModel(omitted, undefined, null)).toThrow(MissingReasoningOptionsError);
+  // Authored controls are the escape hatch, and an explicit [] is the feed's own assertion.
+  expect(buildAiandModel(omitted, { reasoning_options: [{ type: "effort", values: ["high"] }] }, null).reasoning_options)
+    .toEqual([{ type: "effort", values: ["high"] }]);
+  expect(buildAiandModel(aiandModel({ reasoning_options: [] }), undefined, null).reasoning_options).toEqual([]);
+});
+
+test("omitted reasoning_options assert nothing: authored options stay", () => {
+  const built = buildAiandModel(aiandModel({ reasoning_options: undefined }), {
+    reasoning_options: [{ type: "effort", values: ["high"] }],
+  });
+  expect(built.reasoning_options).toEqual([{ type: "effort", values: ["high"] }]);
+});
+
+test("an explicit empty list asserts no caller controls and is written as-is", () => {
+  const built = buildAiandModel(
+    aiandModel({ reasoning_options: [] }),
+    { reasoning_options: [{ type: "effort", values: ["high"] }] },
+    null,
+  );
+  expect(built.reasoning_options).toEqual([]);
+});
+
+test("parseModels refuses an empty feed instead of authorizing catalog deletion", () => {
+  expect(() => aiand.parseModels({ aiand: { models: {} } })).toThrow(/refusing an empty feed/);
+});
+
+test("modalities keep canonical order regardless of feed order", () => {
+  const built = buildAiandModel(
+    aiandModel({ modalities: { input: ["video", "text", "pdf", "image"], output: ["text"] } }),
+    undefined,
+    null,
+  );
+  expect(built.modalities?.input).toEqual(["text", "image", "video", "pdf"]);
+});
+
+test("factors against an authored base_model and never overrides family", () => {
+  const built = buildAiandModel(aiandModel(), {
+    base_model: "deepseek/deepseek-v4-flash",
+    base_model_omit: ["limit.input"],
+  });
+  expect(built).toMatchObject({
+    base_model: "deepseek/deepseek-v4-flash",
+    base_model_omit: ["limit.input"],
+  });
+  expect("family" in built ? built.family : undefined).toBeUndefined();
+});
+
+test("a curated alpha/beta survives a feed that omits status; a feed status wins", () => {
+  const kept = buildAiandModel(aiandModel(), { status: "beta" }, null);
+  expect(kept.status).toBe("beta");
+  const overridden = buildAiandModel(aiandModel({ status: "deprecated" }), { status: "beta" }, null);
+  expect(overridden.status).toBe("deprecated");
+});
+
+test("the feed owns deprecation: an omitted status clears a curated deprecated", () => {
+  const reactivated = buildAiandModel(aiandModel(), { status: "deprecated" }, null);
+  expect(reactivated.status).toBeUndefined();
+});
+
+test("skippedNotice stays silent on a clean sync", () => {
+  expect(aiand.skippedNotice([])).toEqual([]);
+  expect(aiand.skippedNotice(["unknown-lab/mystery-9"])).toHaveLength(1);
+});
+
+test("authored-only cost fields ride along; feed prices are authoritative", () => {
+  const built = buildAiandModel(
+    aiandModel(),
+    { cost: { input: 9, output: 9, cache_write: 0.5 }, limit: { context: 1, input: 128_000, output: 1 } },
+    null,
+  );
+  expect(built.cost).toMatchObject({ input: 0.15, output: 0.25, cache_write: 0.5 });
+  expect(built.limit).toEqual({ context: 1_048_576, input: 128_000, output: 384_000 });
+});
+
+test("a non-reasoning feed model omits reasoning_options entirely", () => {
+  const built = buildAiandModel(
+    aiandModel({ reasoning: false, reasoning_options: undefined }),
+    undefined,
+    null,
+  );
+  expect(built.reasoning).toBe(false);
+  expect(built.reasoning_options).toBeUndefined();
+});
+
+test("a new feed id resolves its lab base model despite a different lab prefix", () => {
+  expect(resolveAiandBaseModel("deepseek-ai/deepseek-v4-flash", "DeepSeek V4 Flash")).toBe(
+    "deepseek/deepseek-v4-flash",
+  );
+  expect(resolveAiandBaseModel("openai/gpt-oss-120b", "GPT OSS 120B")).toBe("openai/gpt-oss-120b");
+  expect(resolveAiandBaseModel("unknown-lab/mystery-9", "Mystery 9")).toBeUndefined();
+});
+
+test("translateModel skips a new id with no resolvable base instead of writing a full definition", () => {
+  const context = { existing: () => undefined, authored: () => undefined };
+  const skipped = aiand.translateModel(
+    aiandModel({ id: "unknown-lab/mystery-9", name: "Mystery 9" }),
+    context,
+  );
+  expect(skipped).toBeUndefined();
+
+  const resolved = aiand.translateModel(aiandModel(), context);
+  expect(resolved?.model).toMatchObject({ base_model: "deepseek/deepseek-v4-flash" });
+});
+
+test("a new factored file inherits every lab-owned field instead of asserting the feed's", () => {
+  const context = { existing: () => undefined, authored: () => undefined };
+  const created = aiand.translateModel(
+    aiandModel({ description: "gateway blurb", release_date: "2099-01-01", open_weights: false }),
+    context,
+  );
+  expect(created?.model).toMatchObject({ base_model: "deepseek/deepseek-v4-flash" });
+  for (const field of ["name", "description", "family", "release_date", "last_updated", "open_weights"]) {
+    expect(created?.model).not.toHaveProperty(field);
+  }
+});
+
+test("a curated description wins over the feed's on a standalone file", () => {
+  const built = buildAiandModel(aiandModel(), { description: "curated" }, null);
+  expect(built.description).toBe("curated");
+});
+
+test("non-effort controls parse but are never written: ai& has no toggle or budget wire path", () => {
+  const model = AiandModel.parse({
+    ...aiandModel(),
+    reasoning_options: [{ type: "toggle" }, { type: "budget_tokens", min: 1024, max: 32_768 }, { type: "effort", values: ["high"] }],
+  });
+  expect(buildAiandModel(model, undefined, null).reasoning_options).toEqual([{ type: "effort", values: ["high"] }]);
+
+  // Dropping them can't leave an invented [] behind: with nothing authored the model fails for review.
+  const toggleOnly = AiandModel.parse({ ...aiandModel(), reasoning_options: [{ type: "toggle" }] });
+  expect(() => buildAiandModel(toggleOnly, undefined, null)).toThrow(MissingReasoningOptionsError);
+
+  // A stale authored toggle is not carried forward on update either.
+  const authoredStale = { reasoning_options: [{ type: "toggle" as const }, { type: "effort" as const, values: ["high" as const] }] };
+  expect(buildAiandModel(aiandModel({ reasoning_options: undefined }), authoredStale, null).reasoning_options)
+    .toEqual([{ type: "effort", values: ["high"] }]);
+});
+
+test("a new reasoner gets the gateway's reasoning_content side channel; feed and authored values win over it", () => {
+  const created = buildAiandModel(aiandModel(), undefined, null);
+  expect(created.interleaved).toEqual({ field: "reasoning_content" });
+  const authored = buildAiandModel(aiandModel(), { interleaved: true }, null);
+  expect(authored.interleaved).toBe(true);
+  const fromFeed = buildAiandModel(
+    AiandModel.parse({ ...aiandModel(), interleaved: { field: "reasoning_details" } }),
+    { interleaved: true },
+    null,
+  );
+  expect(fromFeed.interleaved).toEqual({ field: "reasoning_details" });
+  const nonReasoner = buildAiandModel(
+    aiandModel({ reasoning: false, reasoning_options: undefined }),
+    { interleaved: true },
+    null,
+  );
+  expect(nonReasoner.interleaved).toBeUndefined();
+});
+
+test("an unresolvable new id enters the missing-model issue flow", () => {
+  expect(aiand.missingModelID(aiandModel({ id: "unknown-lab/mystery-9" }))).toBe("unknown-lab/mystery-9");
+});
+
+test("first factor of a full-inline file inherits the lab entry instead of re-emitting its stale lab fields", () => {
+  const fullInline = {
+    name: "Motif 3",
+    description: "old inline description",
+    release_date: "2026-08-12",
+    last_updated: "2026-09-11",
+    attachment: false,
+    reasoning: true,
+    temperature: false,
+    tool_call: false,
+    structured_output: false,
+    open_weights: false,
+    reasoning_options: [],
+    interleaved: { field: "reasoning_content" as const },
+    cost: { input: 0.5, output: 2 },
+    limit: { context: 262_144, output: 262_144 },
+    modalities: { input: ["text" as const], output: ["text" as const] },
+  };
+  const built = buildAiandModel(
+    aiandModel({
+      id: "motif-technologies/motif-3",
+      name: "Motif-Technologies/Motif-3",
+      reasoning_options: [{ type: "effort", values: ["none", "high"] }],
+      structured_output: false,
+      cost: { input: 0.5, output: 2, cache_read: 0.2 },
+      limit: { context: 262_144, output: 262_144 },
+    }),
+    fullInline,
+    "motif-technologies/motif-3",
+  );
+  expect(built).toMatchObject({ base_model: "motif-technologies/motif-3" });
+  for (const field of ["name", "description", "release_date", "last_updated", "open_weights", "family", "temperature", "tool_call"]) {
+    expect(built).not.toHaveProperty(field);
+  }
+  expect(built).toMatchObject({ cost: { cache_read: 0.2 }, interleaved: { field: "reasoning_content" } });
+});
+
+test("an already-factored file keeps its authored lab-field deltas and omit list", () => {
+  const built = buildAiandModel(aiandModel(), {
+    base_model: "deepseek/deepseek-v4-flash",
+    base_model_omit: ["limit.input"],
+    name: "DeepSeek V4 Flash (ai& lane)",
+    knowledge: "2025-06",
+  });
+  // Deltas that differ from the lab entry survive; identical values would be
+  // dropped by factoring as redundant, which is the point.
+  expect(built).toMatchObject({
+    base_model: "deepseek/deepseek-v4-flash",
+    base_model_omit: ["limit.input"],
+    name: "DeepSeek V4 Flash (ai& lane)",
+    knowledge: "2025-06",
+  });
+});
+
+test("a created reasoner gets a leading header naming the single effort wire path and no other control", () => {
+  const context = { existing: () => undefined, authored: () => undefined };
+  const created = aiand.translateModel(
+    AiandModel.parse({ ...aiandModel(), reasoning_options: [{ type: "toggle" }, { type: "effort", values: ["high", "max"] }] }),
+    context,
+  );
+  expect(created?.header).toContain('# Effort: reasoning_effort = "high" | "max"');
+  expect(created?.header).toContain("no separate toggle or token-budget field");
+  expect(created?.header).not.toContain("# Toggle:");
+  expect(created?.header).not.toContain("# Budget:");
+  expect(created?.header?.endsWith("\n")).toBe(true);
+});
+
+test("runner-level: a full-inline file factored for the first time is written without a lab-identical description", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "models-dev-aiand-"));
+  const modelsDir = path.join(root, "providers", "aiand", "models");
+  const labSource = path.join(import.meta.dirname, "..", "..", "..", "models", "motif-technologies", "motif-3.toml");
+  const labDest = path.join(root, "models", "motif-technologies", "motif-3.toml");
+  await mkdir(path.dirname(labDest), { recursive: true });
+  await copyFile(labSource, labDest);
+  const providerFile = path.join(modelsDir, "motif-technologies", "motif-3.toml");
+  await mkdir(path.dirname(providerFile), { recursive: true });
+  await Bun.write(providerFile, [
+    "# First-party Motif host entry (no shared lab base_model in catalog yet).",
+    'name = "Motif 3"',
+    'description = "Motif 3 is a large-scale, decoder-only Mixture-of-Experts (MoE) language model with 314 billion total parameters and 13.2 billion parameters activated per token."',
+    'release_date = "2026-08-12"',
+    'last_updated = "2026-09-11"',
+    "attachment = false",
+    "reasoning = true",
+    "temperature = false",
+    "tool_call = false",
+    "structured_output = false",
+    "open_weights = false",
+    "",
+    "[[reasoning_options]]",
+    'type = "effort"',
+    'values = ["none", "high"]',
+    "",
+    "[interleaved]",
+    'field = "reasoning_content"',
+    "",
+    "[cost]",
+    "input = 0.5",
+    "output = 2",
+    "",
+    "[limit]",
+    "context = 262_144",
+    "output = 262_144",
+    "",
+    "[modalities]",
+    'input = ["text"]',
+    'output = ["text"]',
+    "",
+  ].join("\n"));
+
+  const feed = aiandModel({
+    id: "motif-technologies/motif-3",
+    name: "Motif-Technologies/Motif-3",
+    reasoning_options: [{ type: "effort", values: ["none", "high"] }],
+    structured_output: false,
+    cost: { input: 0.5, output: 2, cache_read: 0.2 },
+    limit: { context: 262_144, output: 262_144 },
+  });
+  await syncProvider(
+    { ...aiand, modelsDir, fetchModels: async () => ({ aiand: { models: { [feed.id]: feed } } }) },
+    { openIssues: false },
+  );
+
+  const written = await readFile(providerFile, "utf8");
+  expect(written).toContain('base_model = "motif-technologies/motif-3"');
+  expect(written).not.toMatch(/^description = /m);
+  expect(written).not.toMatch(/^open_weights = /m);
+  expect(written).not.toMatch(/^release_date = /m);
+  expect(written).toContain("cache_read = 0.2");
+  expect(written).toContain('field = "reasoning_content"');
+  expect(written.startsWith("# First-party Motif host entry")).toBe(true);
+});
+
+test("runner-level: a reasoner the feed leaves without controls is reported and its local file is left untouched", async () => {
+  const root = await mkdtemp(path.join(tmpdir(), "models-dev-aiand-missing-"));
+  const modelsDir = path.join(root, "providers", "aiand", "models");
+  const labDest = path.join(root, "models", "motif-technologies", "motif-3.toml");
+  await mkdir(path.dirname(labDest), { recursive: true });
+  await copyFile(path.join(import.meta.dirname, "..", "..", "..", "models", "motif-technologies", "motif-3.toml"), labDest);
+  const providerFile = path.join(modelsDir, "motif-technologies", "motif-3.toml");
+  await mkdir(path.dirname(providerFile), { recursive: true });
+  const original = ['base_model = "motif-technologies/motif-3"', "", "[cost]", "input = 0.5", "output = 2", ""].join("\n");
+  await Bun.write(providerFile, original);
+
+  const feed = aiandModel({ id: "motif-technologies/motif-3", name: "Motif-Technologies/Motif-3", reasoning_options: undefined });
+  const result = await syncProvider(
+    { ...aiand, modelsDir, fetchModels: async () => ({ aiand: { models: { [feed.id]: feed } } }) },
+    { openIssues: false },
+  );
+
+  expect(await readFile(providerFile, "utf8")).toBe(original);
+  expect(result.notices.join(" ")).toContain("motif-technologies/motif-3");
+  expect(result.notices.join(" ")).toContain("no authored controls");
+});
+
+test("parses the provider entry from the full api.json document", () => {
+  const parsed = AiandResponse.parse({
+    opencode: { models: {} },
+    aiand: { models: { "deepseek-ai/deepseek-v4-flash": aiandModel() } },
+  });
+  expect(Object.keys(parsed.aiand.models)).toEqual(["deepseek-ai/deepseek-v4-flash"]);
+});
diff --git a/packages/core/test/auto-merge.test.ts b/packages/core/test/auto-merge.test.ts
index 1b4a44e3fd5..51eb633de48 100644
--- a/packages/core/test/auto-merge.test.ts
+++ b/packages/core/test/auto-merge.test.ts
@@ -94,7 +94,7 @@ test("does not inspect deleted models", async () => {
 });
 
 test("allows reviewed providers with explicit reasoning options", async () => {
-  for (const provider of ["crossmodel", "edenai", "empiriolabs", "hyper", "kilo", "llmgateway", "llmgateway-providers", "merge-gateway", "nano-gpt", "openrouter", "venice"]) {
+  for (const provider of ["aiand", "crossmodel", "edenai", "empiriolabs", "hyper", "kilo", "llmgateway", "llmgateway-providers", "merge-gateway", "nano-gpt", "openrouter", "venice"]) {
     const decision = await classifyAutoMerge(
       [{ status: "updated", path: `providers/${provider}/models/reasoner.toml` }],
       async () => fullModel(true, 'reasoning_options = [{ type = "toggle" }]'),
diff --git a/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml b/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
index 9b8bea586bd..50018b3d802 100644
--- a/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
+++ b/providers/aiand/models/deepseek-ai/deepseek-v4-flash.toml
@@ -1,7 +1,8 @@
 # Effort: reasoning_effort = none|high|max
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high|max → 200; minimal|low|medium|xhigh → 400.
+# Live probe 2026-09-14: none|high|max → 200; minimal|low|medium|xhigh → 400.
+# context_window = 1048576 on this host overrides the lab's 1_000_000.
 # Reasoning side channel: message.reasoning_content
 base_model = "deepseek/deepseek-v4-flash"
 
diff --git a/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml b/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
index 88498c5c29a..4beec143d13 100644
--- a/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
+++ b/providers/aiand/models/deepseek-ai/deepseek-v4-pro.toml
@@ -1,7 +1,8 @@
 # Effort: reasoning_effort = none|high|max
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high|max → 200; low → 400.
+# Live probe 2026-09-14: none|high|max → 200; minimal|low|medium|xhigh → 400.
+# context_window = 1048576 on this host overrides the lab's 1_000_000.
 # Reasoning side channel: message.reasoning_content
 base_model = "deepseek/deepseek-v4-pro"
 
diff --git a/providers/aiand/models/google/gemma-4-31b-it.toml b/providers/aiand/models/google/gemma-4-31b-it.toml
index 312e02bf19b..834d710d9db 100644
--- a/providers/aiand/models/google/gemma-4-31b-it.toml
+++ b/providers/aiand/models/google/gemma-4-31b-it.toml
@@ -1,7 +1,7 @@
 # Effort: reasoning_effort = none|high
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high → 200; minimal|low|xhigh → 400.
+# Live probe 2026-09-14: none|high → 200; minimal|low|medium|xhigh|max → 400.
 # Modalities: image, video, and PDF accepted on this host (2026-07-24 probes).
 # Reasoning side channel: message.reasoning_content
 base_model = "google/gemma-4-31b-it"
diff --git a/providers/aiand/models/moonshotai/kimi-k2.7-code.toml b/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
index a384aa028b6..de7ff0f4876 100644
--- a/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/aiand/models/moonshotai/kimi-k2.7-code.toml
@@ -1,18 +1,20 @@
-# Effort: reasoning_effort = high (always-on; none/low rejected)
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Effort: reasoning_effort = high (always-on)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: high → 200; none|low → 400.
+# Live probe 2026-09-14: high → 200; none|minimal|low|medium|xhigh|max → 400.
 # Modalities: image + PDF accepted; video rejected (2026-07-24 probes).
+# temperature = true: the host accepts the parameter (live probe 2026-09-14: 200 at 0.0 and 1.8); the lab file records Moonshot's own API as false.
 # Reasoning side channel: message.reasoning_content
 base_model = "moonshotai/kimi-k2.7-code"
+temperature = true
+
+[interleaved]
+field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
 values = ["high"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.75
 output = 3.5
diff --git a/providers/aiand/models/moonshotai/kimi-k3.toml b/providers/aiand/models/moonshotai/kimi-k3.toml
index b28cbd68903..0a249611cd8 100644
--- a/providers/aiand/models/moonshotai/kimi-k3.toml
+++ b/providers/aiand/models/moonshotai/kimi-k3.toml
@@ -1,18 +1,20 @@
 # Effort: reasoning_effort = low|high|max (always-on; no none)
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# K3 accepts only low|high|max (lab docs + host 400 on none/minimal/medium/xhigh).
+# Live probe 2026-09-14: low|high|max → 200; none|minimal|medium|xhigh → 400.
 # Modalities: text + image + PDF; video rejected on this host.
+# temperature = true: the host accepts the parameter (live probe 2026-09-14: 200 at 0.0 and 1.8); the lab file records Moonshot's own API as false.
 # Reasoning side channel: message.reasoning_content
 base_model = "moonshotai/kimi-k3"
+temperature = true
+
+[interleaved]
+field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
 values = ["low", "high", "max"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 3
 output = 12.5
diff --git a/providers/aiand/models/motif-technologies/motif-3.toml b/providers/aiand/models/motif-technologies/motif-3.toml
index d0dcf873cc7..e8b372d3399 100644
--- a/providers/aiand/models/motif-technologies/motif-3.toml
+++ b/providers/aiand/models/motif-technologies/motif-3.toml
@@ -1,36 +1,21 @@
 # Effort: reasoning_effort = none|high
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high → 200; low → 400.
-# First-party Motif host entry (no shared lab base_model in catalog yet).
+# Live probe 2026-09-14: none|high → 200; minimal|low|medium|xhigh|max → 400.
+# Lab facts (release date, open weights, tool calling, temperature) live in
+# models/motif-technologies/motif-3.toml. structured_output stays false: live
+# probe 2026-09-14 — response_format json_schema is ignored (prose reply).
 # Reasoning side channel: message.reasoning_content
-name = "Motif 3"
-description = "Motif 3 is a large-scale, decoder-only Mixture-of-Experts (MoE) language model with 314 billion total parameters and 13.2 billion parameters activated per token."
-release_date = "2026-08-12"
-last_updated = "2026-09-11"
-attachment = false
-reasoning = true
-temperature = false
-tool_call = false
-structured_output = false
-open_weights = false
+base_model = "motif-technologies/motif-3"
+
+[interleaved]
+field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
 values = ["none", "high"]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
 input = 0.5
 output = 2
 cache_read = 0.2
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/aiand/models/openai/gpt-oss-120b.toml b/providers/aiand/models/openai/gpt-oss-120b.toml
index 7083c30e20d..c36d59ddb11 100644
--- a/providers/aiand/models/openai/gpt-oss-120b.toml
+++ b/providers/aiand/models/openai/gpt-oss-120b.toml
@@ -1,7 +1,7 @@
 # Effort: reasoning_effort = low|medium|high
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-07-24: low|medium|high → 200; none|minimal|xhigh → 400.
+# Live probe 2026-09-14: low|medium|high → 200; none|minimal|xhigh|max → 400.
 # Reasoning side channel: message.reasoning_content
 base_model = "openai/gpt-oss-120b"
 
diff --git a/providers/aiand/models/qwen/qwen3.6-27b.toml b/providers/aiand/models/qwen/qwen3.6-27b.toml
index 1f874fc1bc4..114703224f2 100644
--- a/providers/aiand/models/qwen/qwen3.6-27b.toml
+++ b/providers/aiand/models/qwen/qwen3.6-27b.toml
@@ -1,7 +1,8 @@
 # Effort: reasoning_effort = none|high
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high → 200; minimal|low|xhigh → 400.
+# Live probe 2026-09-14: none|high → 200; minimal|low|medium|xhigh|max → 400.
+# Modalities: image, video, and PDF accepted on this host (catalog vision|video|document).
 # Reasoning side channel: message.reasoning_content
 base_model = "alibaba/qwen3.6-27b"
 
diff --git a/providers/aiand/models/qwen/qwen3.8-27b.toml b/providers/aiand/models/qwen/qwen3.8-27b.toml
index 487eb3e6f98..cdb5d7d084c 100644
--- a/providers/aiand/models/qwen/qwen3.8-27b.toml
+++ b/providers/aiand/models/qwen/qwen3.8-27b.toml
@@ -1,7 +1,10 @@
 # Effort: reasoning_effort = none|low|medium|xhigh
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-08-29: none|low|medium|xhigh → 200; minimal|high|max → 400.
+# Live probe 2026-09-14: none|low|medium|xhigh → 200; minimal|max → 400. high also
+# returns 200 via ai&'s announced substitution (X-Reasoning-Effort: xhigh) and is
+# deliberately not listed.
+# Modalities: image, video, and PDF accepted on this host (catalog vision|video|document).
 # Reasoning side channel: message.reasoning_content
 base_model = "alibaba/qwen3.8-27b"
 
diff --git a/providers/aiand/models/zai-org/glm-5.2.toml b/providers/aiand/models/zai-org/glm-5.2.toml
index ed3b498d41c..3c8e485f19d 100644
--- a/providers/aiand/models/zai-org/glm-5.2.toml
+++ b/providers/aiand/models/zai-org/glm-5.2.toml
@@ -1,7 +1,8 @@
 # Effort: reasoning_effort = none|high|max
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: none|high|max → 200; low|xhigh → 400.
+# Live probe 2026-09-14: none|high|max → 200; minimal|low|medium|xhigh → 400.
+# context_window = 1048576 on this host overrides the lab's 1_000_000.
 # Reasoning side channel: message.reasoning_content
 base_model = "zhipuai/glm-5.2"
 
diff --git a/providers/aiand/models/zai-org/glm-5.3.toml b/providers/aiand/models/zai-org/glm-5.3.toml
index d7d6ad36c6d..29850bab7b8 100644
--- a/providers/aiand/models/zai-org/glm-5.3.toml
+++ b/providers/aiand/models/zai-org/glm-5.3.toml
@@ -1,7 +1,8 @@
 # Effort: reasoning_effort = low|high|max (always-on; no none)
-# Pricing: GET https://api.aiand.com/v1/models (accessed 2026-09-11)
+# Pricing: GET https://api.aiand.com/v1/api.json (accessed 2026-09-14); synced hourly by the aiand module.
 # Docs: https://docs.aiand.com/models/catalog/
-# Live probe 2026-09-11: low|high|max → 200; none|xhigh → 400.
+# Live probe 2026-09-14: low|high|max → 200; none|minimal|medium|xhigh → 400.
+# context_window = 1048576 on this host overrides the lab's 1_000_000.
 # Reasoning side channel: message.reasoning_content
 base_model = "zhipuai/glm-5.3"
 
diff --git a/sync.md b/sync.md
index abfb3fac737..f2bb9a5130e 100644
--- a/sync.md
+++ b/sync.md
@@ -9,6 +9,7 @@ The grouped sync targets are available for local convenience, but CI syncs each
 ## Commands
 
 - `bun models:sync aggregators` syncs every provider in the `aggregators` group.
+- `bun models:sync aiand` syncs only ai&.
 - `bun models:sync openrouter` syncs only OpenRouter.
 - `bun models:sync cloudflare-workers-ai` syncs only Cloudflare Workers AI.
 - `bun models:sync cloudflare-ai-gateway` syncs only Cloudflare AI Gateway's proxied catalog.
@@ -361,3 +362,15 @@ Venice is implemented in `packages/core/src/sync/providers/venice.ts`.
 ## Standalone Generators
 
 Some provider scripts in `packages/core/script/generate-*.ts` are not wired into `bun models:sync`. When updating those scripts, preserve existing `base_model` and `base_model_omit` fields for generated TOMLs that already use model metadata inheritance. New inheritance-aware output should use `base_model`; do not reintroduce legacy `[extends]` syntax.
+
+## ai& Notes
+
+- Endpoint: `GET https://api.aiand.com/v1/api.json` (public, no auth). The module reads the `aiand` provider entry; `AIAND_API_URL` overrides the endpoint for staging dry runs.
+- The feed publishes this repo's `api.json` shape, so translation is near-identity. The feed is authoritative for prices (including `cache_read`), limits, capability flags, modalities, gateway-enforced `reasoning_options`, and `deprecated` status; the `aiand` entry lists only models whose catalog metadata is complete.
+- Curated values win for `name`, `description`, `knowledge`, `release_date`, and `last_updated` — the feed's `last_updated` tracks catalog-row edits, not model revisions, and release dates are lab metadata the gateway is not authoritative for. A curated alpha/beta `status` survives a feed that omits one; a curated `deprecated` does not, since the feed owns deprecation and absence means active.
+- On base-factored files, lab-owned fields (`name`, `description`, `family`, `release_date`, `last_updated`, `knowledge`, `open_weights`) are never asserted from the feed: they appear only as deltas the authored TOML already carried on top of its `base_model` (read via `context.authored()`, never the base-resolved merge), so a full-inline file being factored for the first time — or a new file — inherits the lab entry outright. A new feed id resolves its `base_model` by normalized match against `models/` and is skipped with a report notice when nothing resolves — an unfactored full definition is never created. An empty feed fails the run rather than deleting the local catalog.
+- `family` passes through `ModelFamily.safeParse` and is omitted when unknown.
+- A skipped new id is returned from `missingModelID` (the only skip is "no lab base yet"), so the runner preserves any existing local entry and opens a deduped missing-model issue for the lab metadata.
+- `preserveDescriptions` is off: the runner must not re-inject a pre-factor authored description when the translator leaves it unset, or a full-inline → factored transition would recreate a lab-identical override. Created files get a leading header from `translateModel` naming the single reasoning wire path (`reasoning_effort`, "none" = off). `toggle` and `budget_tokens` controls are parsed (so a feed that publishes one never aborts the run) but never written, from feed or authored file: ai& has no separate on/off or token-budget field, so they cannot be true of this host.
+- `interleaved` comes from the feed when published, else the authored value, else `{ field = "reasoning_content" }` for reasoners — every ai& reasoner streams thinking in `message.reasoning_content`, so a new reasoner is never created without its side channel.
+- A reasoner never gets an invented `[]`: an omitted feed list keeps the authored controls, and a non-empty list whose effort values the schema doesn't know yet does too; when neither yields a schema-valid set the model fails with `MissingReasoningOptionsError` — the runner preserves the local file and routes the id to the missing-model issue flow. Only an explicit `[]` published by the feed is written as "no caller control".

From 040dd0f3872bad713889069320a324a62b629d52 Mon Sep 17 00:00:00 2001
From: Nikolas Sturm <code@nsturm.me>
Date: Thu, 17 Sep 2026 03:19:42 +0200
Subject: [PATCH 1481/2012] feat(providers): add Inco (#6978)

* feat(providers): add Inco

Inco (inco.ai) is a prepaid OpenAI- and Anthropic-compatible inference
provider. Adds 7 model routes, all inheriting lab metadata via
base_model with provider-specific cost and reasoning controls:

- deepseek-v4.1-flash:fast ($0.60/$2.40)
- glm-5.3 ($1.40/$4.40), glm-5.3:fast ($2.80/$8.80)
- glm-5.3-flash:fast ($0.15/$0.50)
- kimi-k3:fast ($6.00/$30.00)
- minimax-m3 ($0.30/$1.20), minimax-m3:fast ($0.60/$2.40)

Reasoning controls verified against the live API:
- GLM models toggle via chat_template_kwargs.enable_thinking
- DeepSeek and Kimi toggle via chat_template_kwargs.thinking
- reasoning_effort and thinking.type are accepted but ignored
- MiniMax M3 always reasons; no caller control (reasoning_options = [])

All models return interleaved reasoning in reasoning_content. Logo is
the "inco" Montserrat Alternates SemiBold wordmark knocked out of a
square, in currentColor.

* fix(providers): document Inco toggle wire path on glm-5.3:fast
---
 providers/inco/logo.svg                           |  3 +++
 .../inco/models/deepseek-v4.1-flash:fast.toml     | 15 +++++++++++++++
 providers/inco/models/glm-5.3-flash:fast.toml     | 15 +++++++++++++++
 providers/inco/models/glm-5.3.toml                | 15 +++++++++++++++
 providers/inco/models/glm-5.3:fast.toml           | 15 +++++++++++++++
 providers/inco/models/kimi-k3:fast.toml           | 14 ++++++++++++++
 providers/inco/models/minimax-m3.toml             | 13 +++++++++++++
 providers/inco/models/minimax-m3:fast.toml        | 12 ++++++++++++
 providers/inco/provider.toml                      |  5 +++++
 9 files changed, 107 insertions(+)
 create mode 100644 providers/inco/logo.svg
 create mode 100644 providers/inco/models/deepseek-v4.1-flash:fast.toml
 create mode 100644 providers/inco/models/glm-5.3-flash:fast.toml
 create mode 100644 providers/inco/models/glm-5.3.toml
 create mode 100644 providers/inco/models/glm-5.3:fast.toml
 create mode 100644 providers/inco/models/kimi-k3:fast.toml
 create mode 100644 providers/inco/models/minimax-m3.toml
 create mode 100644 providers/inco/models/minimax-m3:fast.toml
 create mode 100644 providers/inco/provider.toml

diff --git a/providers/inco/logo.svg b/providers/inco/logo.svg
new file mode 100644
index 00000000000..e60b4418f08
--- /dev/null
+++ b/providers/inco/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 48 48" fill="currentColor">
+  <path fill-rule="evenodd" d="M0 0 H48 V48 H0 Z M10.66 22.04H12.28V28.95H10.66ZM10.44 19.93Q10.44 19.52 10.74 19.24Q11.03 18.96 11.47 18.96Q11.91 18.96 12.21 19.23Q12.51 19.49 12.51 19.89Q12.51 20.32 12.22 20.61Q11.93 20.90 11.47 20.90Q11.03 20.90 10.74 20.62Q10.44 20.35 10.44 19.93Z M21.20 24.99V28.95H19.58V25.20Q19.58 24.29 19.16 23.83Q18.73 23.37 17.94 23.37Q17.05 23.37 16.53 23.91Q16.01 24.45 16.01 25.45V28.95H14.40V22.04H15.94V22.93Q16.34 22.45 16.95 22.21Q17.55 21.96 18.32 21.96Q19.62 21.96 20.41 22.73Q21.20 23.49 21.20 24.99Z M22.69 25.49Q22.69 24.47 23.17 23.67Q23.64 22.87 24.49 22.42Q25.34 21.96 26.42 21.96Q27.45 21.96 28.22 22.38Q28.99 22.79 29.37 23.57L28.13 24.29Q27.83 23.81 27.39 23.57Q26.94 23.33 26.41 23.33Q25.51 23.33 24.91 23.92Q24.32 24.51 24.32 25.49Q24.32 26.48 24.91 27.07Q25.49 27.65 26.41 27.65Q26.94 27.65 27.39 27.41Q27.83 27.17 28.13 26.70L29.37 27.42Q28.97 28.20 28.20 28.62Q27.43 29.04 26.42 29.04Q25.35 29.04 24.50 28.58Q23.64 28.13 23.17 27.32Q22.69 26.52 22.69 25.49Z M30.20 25.49Q30.20 24.47 30.67 23.67Q31.15 22.87 31.99 22.42Q32.83 21.96 33.87 21.96Q34.93 21.96 35.78 22.42Q36.62 22.87 37.09 23.67Q37.56 24.47 37.56 25.49Q37.56 26.52 37.09 27.32Q36.62 28.13 35.78 28.58Q34.93 29.04 33.87 29.04Q32.83 29.04 31.99 28.58Q31.15 28.13 30.67 27.32Q30.20 26.52 30.20 25.49ZM35.93 25.49Q35.93 24.52 35.35 23.93Q34.77 23.33 33.87 23.33Q32.98 23.33 32.41 23.93Q31.83 24.52 31.83 25.49Q31.83 26.46 32.41 27.06Q32.98 27.65 33.87 27.65Q34.77 27.65 35.35 27.06Q35.93 26.46 35.93 25.49Z"/>
+</svg>
diff --git a/providers/inco/models/deepseek-v4.1-flash:fast.toml b/providers/inco/models/deepseek-v4.1-flash:fast.toml
new file mode 100644
index 00000000000..2b762eedb7e
--- /dev/null
+++ b/providers/inco/models/deepseek-v4.1-flash:fast.toml
@@ -0,0 +1,15 @@
+# Toggle: chat_template_kwargs.thinking = true|false (verified against Inco;
+# reasoning_effort and thinking.type are accepted but ignored). Reasoning is
+# off by default on this route; send thinking=true to enable it.
+# reasoning_content interleaved field verified against Inco responses.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.6
+output = 2.4
diff --git a/providers/inco/models/glm-5.3-flash:fast.toml b/providers/inco/models/glm-5.3-flash:fast.toml
new file mode 100644
index 00000000000..e4b52c4169f
--- /dev/null
+++ b/providers/inco/models/glm-5.3-flash:fast.toml
@@ -0,0 +1,15 @@
+# Toggle: chat_template_kwargs.enable_thinking = true|false (verified against
+# Inco; reasoning_effort and thinking.type are accepted but ignored). Reasons
+# by default; reasoning_content interleaved field verified against Inco
+# responses.
+base_model = "zhipuai/glm-5.3-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
diff --git a/providers/inco/models/glm-5.3.toml b/providers/inco/models/glm-5.3.toml
new file mode 100644
index 00000000000..bfc8998ba01
--- /dev/null
+++ b/providers/inco/models/glm-5.3.toml
@@ -0,0 +1,15 @@
+# Toggle: chat_template_kwargs.enable_thinking = true|false (verified against
+# Inco; reasoning_effort and thinking.type are accepted but ignored). Reasons
+# by default; reasoning_content interleaved field verified against Inco
+# responses.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
diff --git a/providers/inco/models/glm-5.3:fast.toml b/providers/inco/models/glm-5.3:fast.toml
new file mode 100644
index 00000000000..63d16c7dfe8
--- /dev/null
+++ b/providers/inco/models/glm-5.3:fast.toml
@@ -0,0 +1,15 @@
+# Standard-speed endpoint at 2x credit pricing.
+# Toggle: chat_template_kwargs.enable_thinking = true|false
+# Verified against Inco; reasons by default.
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 Fast"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.8
+output = 8.8
diff --git a/providers/inco/models/kimi-k3:fast.toml b/providers/inco/models/kimi-k3:fast.toml
new file mode 100644
index 00000000000..ebe03653543
--- /dev/null
+++ b/providers/inco/models/kimi-k3:fast.toml
@@ -0,0 +1,14 @@
+# Toggle: chat_template_kwargs.thinking = true|false (verified against Inco;
+# reasoning_effort and thinking.type are accepted but ignored). Reasons by
+# default; reasoning_content interleaved field verified against Inco responses.
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 6
+output = 30
diff --git a/providers/inco/models/minimax-m3.toml b/providers/inco/models/minimax-m3.toml
new file mode 100644
index 00000000000..dee42c58f8c
--- /dev/null
+++ b/providers/inco/models/minimax-m3.toml
@@ -0,0 +1,13 @@
+# Always reasons on Inco; no caller control — toggle fields (thinking.type,
+# chat_template_kwargs.thinking/enable_thinking, reasoning) were tested and are
+# ignored. reasoning_content interleaved field verified against Inco responses.
+base_model = "minimax/MiniMax-M3"
+
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
diff --git a/providers/inco/models/minimax-m3:fast.toml b/providers/inco/models/minimax-m3:fast.toml
new file mode 100644
index 00000000000..02e06cf191d
--- /dev/null
+++ b/providers/inco/models/minimax-m3:fast.toml
@@ -0,0 +1,12 @@
+# Standard-speed endpoint at 2x credit pricing, see minimax-m3.toml for controls.
+base_model = "minimax/MiniMax-M3"
+name = "MiniMax M3 Fast"
+
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.6
+output = 2.4
diff --git a/providers/inco/provider.toml b/providers/inco/provider.toml
new file mode 100644
index 00000000000..1cbf0c9ba78
--- /dev/null
+++ b/providers/inco/provider.toml
@@ -0,0 +1,5 @@
+name = "Inco"
+npm = "@ai-sdk/openai-compatible"
+env = ["INCO_API_KEY"]
+api = "https://api.inco.ai/v1"
+doc = "https://platform.inco.ai/docs"

From 54d2d5f3460fdd946e9b43b0eee2d36aba211586 Mon Sep 17 00:00:00 2001
From: Yanis Miraoui <yanis.miraoui@gmail.com>
Date: Wed, 16 Sep 2026 18:20:05 -0700
Subject: [PATCH 1482/2012] fix(inception): correct provider logo and Mercury
 Edit 2 metadata (#6749)

The provider logo was not the Inception mark. Replace it with the official
mark (the same artwork inceptionlabs.ai serves as its favicon), converted to
currentColor with no fixed width/height.

Mercury Edit 2 is served only on /v1/edit/completions and /v1/fim/completions,
where context is capped at 32K rather than the 128K chat limit. It exposes no
reasoning control on either endpoint, and the docs model table lists no
tool-calling or structured-output support. Its description had also been copied
from Mercury 2 and advertised tool use that the model does not have.

Fix the provider `doc` link, which returned 403.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/inception/logo.svg                  | 16 +++++++++---
 .../inception/models/mercury-edit-2.toml      | 26 ++++++++++++-------
 providers/inception/provider.toml             |  9 ++++---
 3 files changed, 35 insertions(+), 16 deletions(-)

diff --git a/providers/inception/logo.svg b/providers/inception/logo.svg
index f70ffbc785b..c8f9d452ccf 100644
--- a/providers/inception/logo.svg
+++ b/providers/inception/logo.svg
@@ -1,4 +1,14 @@
-<svg width="24" height="24" viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">
-<path d="M14.3794 2.99999H8.71219L3.04492 8.66725V14.3345H8.71219V8.66725H14.3794V2.99999Z" fill="currentColor"/>
-<path d="M15.2877 9.66548V15.3327H9.62048V21H15.2877L20.955 15.3327V9.66548H15.2877Z" fill="currentColor"/>
+<svg viewBox="0 0 322 329" fill="currentColor" xmlns="http://www.w3.org/2000/svg">
+<path d="M66.6056 12.2522C78.2802 10.9725 90.1467 9.81141 102.205 8.76904C101.283 18.7387 100.344 29.0241 99.3908 39.6253C86.7998 40.7234 74.4657 41.933 62.3887 43.254C63.8026 32.6301 65.2082 22.2961 66.6056 12.2522Z"/>
+<path d="M190.537 0.314209C209.075 1.76321 227.197 3.48983 244.902 5.49407C247.173 20.8365 249.457 36.8426 251.754 53.5123C233.188 51.4234 213.965 49.5885 194.085 48.0074C192.87 31.3521 191.687 15.4544 190.537 0.314209Z"/>
+<path d="M3.32567 83.1635C10.678 82.1789 18.1872 81.1902 25.8533 80.1974C24.9079 88.9946 24.0245 97.9239 23.203 106.985C15.5534 107.735 8.07104 108.476 0.755859 109.208C1.55054 100.419 2.40714 91.7378 3.32567 83.1635Z"/>
+<path d="M125.097 61.4632C142.838 61.1247 160.579 61.1247 178.32 61.4632C178.987 77.2949 179.652 93.8005 180.314 110.98C161.244 110.635 142.174 110.635 123.104 110.98C123.766 93.8005 124.431 77.2949 125.097 61.4632Z"/>
+<path d="M257.047 58.7478C276.794 61.4951 295.649 64.229 313.611 66.9495C315.968 87.5947 317.949 108.787 319.555 130.526C301.897 129.345 283.187 128.121 263.425 126.854C261.637 103.218 259.511 80.5158 257.047 58.7478Z"/>
+<path d="M53.7907 142.982C66.3384 142.561 79.3041 142.139 92.6877 141.716C92.4689 156.637 92.4689 171.559 92.6877 186.48C79.3041 186.057 66.3384 185.635 53.7907 185.214C53.5802 171.137 53.5802 157.059 53.7907 142.982Z"/>
+<path d="M198.812 129.277C219.842 130.206 239.886 131.151 258.946 132.113C259.435 153.437 259.435 174.762 258.946 196.086C239.886 197.048 219.842 197.993 198.812 198.922C199.298 175.707 199.298 152.492 198.812 129.277Z"/>
+<path d="M0.755859 218.987C8.07104 219.72 15.5534 220.461 23.203 221.21C24.0245 230.271 24.9079 239.201 25.8533 247.998C18.1872 247.005 10.678 246.016 3.32567 245.032C2.40714 236.457 1.55054 227.776 0.755859 218.987Z"/>
+<path d="M123.104 217.221C142.174 217.566 161.244 217.566 180.314 217.221C179.652 234.4 178.987 250.906 178.32 266.738C160.579 267.076 142.838 267.076 125.097 266.738C124.431 250.906 123.766 234.4 123.104 217.221Z"/>
+<path d="M263.425 201.342C283.187 200.075 301.897 198.851 319.555 197.67C317.949 219.409 315.968 240.602 313.611 261.247C295.649 263.967 276.794 266.701 257.047 269.448C259.511 247.68 261.637 224.978 263.425 201.342Z"/>
+<path d="M62.3887 284.943C74.4657 286.264 86.7998 287.473 99.3908 288.571C100.344 299.173 101.283 309.458 102.205 319.428C90.1467 318.385 78.2802 317.224 66.6056 315.944C65.2082 305.9 63.8026 295.567 62.3887 284.943Z"/>
+<path d="M194.085 280.187C213.965 278.606 233.188 276.771 251.754 274.682C249.457 291.351 247.173 307.358 244.902 322.7C227.197 324.704 209.075 326.431 190.537 327.88C191.687 312.74 192.87 296.842 194.085 280.187Z"/>
 </svg>
diff --git a/providers/inception/models/mercury-edit-2.toml b/providers/inception/models/mercury-edit-2.toml
index 52bcc2bdf83..83ac2191068 100644
--- a/providers/inception/models/mercury-edit-2.toml
+++ b/providers/inception/models/mercury-edit-2.toml
@@ -1,15 +1,23 @@
-name = "Mercury Edit 2"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-# Model exception (accessed 2026-06-25): the /v1/edit/completions and
-# /v1/fim/completions request schemas expose no reasoning-control field.
+# Sources (accessed 2026-09-10): Mercury Edit 2 is served only on
+# /v1/edit/completions and /v1/fim/completions, which cap context at 32K and
+# output at 8,192 tokens; it is not offered on chat completions. Neither the
+# edit nor the FIM request/response schema exposes a reasoning field, and
+# `reasoning_effort` is documented as applying to mercury-2.5 and mercury-2 on
+# chat completions only. The docs model table lists no tool-calling or
+# structured-output support for this model.
+# https://docs.inceptionlabs.ai/get-started/models
+# https://docs.inceptionlabs.ai/capabilities/reasoning-efforts
 # https://api.inceptionlabs.ai/openapi.json
+name = "Mercury Edit 2"
+description = "Code editing dLLM for autocomplete (FIM) and next-edit suggestions"
+family = "mercury"
 release_date = "2026-03-30"
 last_updated = "2026-03-30"
 attachment = false
-reasoning = true
-reasoning_options = []
+reasoning = false
 temperature = true
 tool_call = false
+structured_output = false
 open_weights = false
 
 [cost]
@@ -18,9 +26,9 @@ output = 0.75
 cache_read = 0.025
 
 [limit]
-context = 128000
-output = 8192
+context = 32_000
+output = 8_192
 
 [modalities]
 input = ["text"]
-output = ["text"]
\ No newline at end of file
+output = ["text"]
diff --git a/providers/inception/provider.toml b/providers/inception/provider.toml
index 362d02d8ffa..90591ade98f 100644
--- a/providers/inception/provider.toml
+++ b/providers/inception/provider.toml
@@ -1,11 +1,12 @@
 name = "Inception"
 env = ["INCEPTION_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
-# Reasoning HTTP format (accessed 2026-06-25): POST /v1/chat/completions accepts
+# Reasoning HTTP format (accessed 2026-09-10): POST /v1/chat/completions accepts
 # top-level `reasoning_effort` = instant, low, medium, or high; default medium.
 # It documents no off value or numeric reasoning-token budget. `instant` is the
-# near-realtime Mercury 2 mode for low-latency turns that do not require tools.
+# near-realtime mode for low-latency turns. Effort applies to mercury-2.5 and
+# mercury-2 on chat completions only.
 # https://docs.inceptionlabs.ai/api-reference/chat/create-a-chat-completion
-# https://docs.inceptionlabs.ai/capabilities/instant
+# https://docs.inceptionlabs.ai/capabilities/reasoning-efforts
 api = "https://api.inceptionlabs.ai/v1/"
-doc = "https://platform.inceptionlabs.ai/docs"
+doc = "https://docs.inceptionlabs.ai/get-started/models"

From 45486bdd643d1abc206082cf1607af1c03df2b78 Mon Sep 17 00:00:00 2001
From: Wassel Alazhar <wassel.alazhar@gmail.com>
Date: Thu, 17 Sep 2026 03:20:36 +0200
Subject: [PATCH 1483/2012] umans-ai + coding-plan: add GLM 5.3 Flash, drop
 K2.7 + GLM 5.2 (sunset 2026-09-10) (#6581)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* umans-ai + coding-plan: add GLM 5.3 Flash, drop K2.7 + GLM 5.2 (sunset 2026-09-10)

Add umans-glm-5.3-flash (Z.ai GLM-5.3-Flash, 320B-A18B MoE, native
multimodal, 1M context) to both Umans providers: pay-per-token at the
publisher's list rates ($0.15/$0.50/$0.03 per 1M) on umans-ai, cost 0 on
the coding plan. Reasoning is effort-only low|high|max (thinking is always
on — no toggle), matching the live can_disable=false card.

Drop umans-kimi-k2.7 and umans-glm-5.2 from both providers: both are
deprecated and removed from the Umans catalog on 2026-09-10 (sunset
announced on https://api.code.umans.ai/v1/models/info —
deprecation.sunset_date), succeeded by umans-kimi-k3 and umans-glm-5.3
respectively.

* umans-ai + coding-plan: repoint umans-coder to GLM 5.3 Flash

The coding alias rides the GLM-5.3-Flash release since 2026-09-09 (it routed
to Kimi K2.7 Code before; K2.7 sunsets 2026-09-10): base_model moves to
zhipuai/glm-5.3-flash, reasoning becomes effort-only low|high|max (thinking
always on — can_disable=false), the pay-per-token entry carries the flash's
rates ($0.15/$0.50/$0.03) and its 1M context / 131071 output cap, and the
modalities gain the native vision (image + video).
---
 .../models/umans-coder.toml                   | 15 ++++++----
 .../models/umans-glm-5.2.toml                 | 20 -------------
 .../models/umans-glm-5.3-flash.toml           | 29 ++++++++++++++++++
 .../models/umans-kimi-k2.7.toml               | 14 ---------
 providers/umans-ai/models/umans-coder.toml    | 20 +++++++------
 providers/umans-ai/models/umans-glm-5.2.toml  | 19 ------------
 .../umans-ai/models/umans-glm-5.3-flash.toml  | 30 +++++++++++++++++++
 .../umans-ai/models/umans-kimi-k2.7.toml      | 16 ----------
 8 files changed, 80 insertions(+), 83 deletions(-)
 delete mode 100644 providers/umans-ai-coding-plan/models/umans-glm-5.2.toml
 create mode 100644 providers/umans-ai-coding-plan/models/umans-glm-5.3-flash.toml
 delete mode 100644 providers/umans-ai-coding-plan/models/umans-kimi-k2.7.toml
 delete mode 100644 providers/umans-ai/models/umans-glm-5.2.toml
 create mode 100644 providers/umans-ai/models/umans-glm-5.3-flash.toml
 delete mode 100644 providers/umans-ai/models/umans-kimi-k2.7.toml

diff --git a/providers/umans-ai-coding-plan/models/umans-coder.toml b/providers/umans-ai-coding-plan/models/umans-coder.toml
index 2d6c1c8b2c2..ca486077ae9 100644
--- a/providers/umans-ai-coding-plan/models/umans-coder.toml
+++ b/providers/umans-ai-coding-plan/models/umans-coder.toml
@@ -1,10 +1,11 @@
-base_model = "moonshotai/kimi-k2.7-code"
+base_model = "zhipuai/glm-5.3-flash"
 name = "Umans Coder"
 temperature = false
-# This alias currently routes to Kimi K2.7 Code: reasoning is always on
-# (`can_disable=false`) with no effort levels (accessed 2026-06-25).
+# This alias routes to GLM-5.3-Flash since 2026-09-09 (it rode Kimi K2.7 Code
+# before): reasoning is always on (`can_disable=false`) WITH effort levels
+# low|high|max, default max (accessed 2026-09-08):
 # https://api.code.umans.ai/v1/models/info
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -15,5 +16,9 @@ output = 0
 cache_read = 0
 cache_write = 0
 
+[limit]
+context = 1_048_576
+output = 131_071
+
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
diff --git a/providers/umans-ai-coding-plan/models/umans-glm-5.2.toml b/providers/umans-ai-coding-plan/models/umans-glm-5.2.toml
deleted file mode 100644
index ccc697fa08b..00000000000
--- a/providers/umans-ai-coding-plan/models/umans-glm-5.2.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2"
-# Live metadata: levels none/high/max, default high, and `can_disable=true`
-# (accessed 2026-06-25): https://api.code.umans.ai/v1/models/info
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
-
-[limit]
-context = 405_504
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/umans-ai-coding-plan/models/umans-glm-5.3-flash.toml b/providers/umans-ai-coding-plan/models/umans-glm-5.3-flash.toml
new file mode 100644
index 00000000000..96bbd094e14
--- /dev/null
+++ b/providers/umans-ai-coding-plan/models/umans-glm-5.3-flash.toml
@@ -0,0 +1,29 @@
+# Effort only: reasoning_effort = low|high|max — thinking is ALWAYS ON (the
+# checkpoint's chat template has no non-thinking mode; verified against the
+# released repo 2026-08-26 and the live lab card since then). The released
+# id's live metadata mirrors that shape: levels low/high/max, default max,
+# `can_disable=false`: https://api.code.umans.ai/v1/models/info
+# Coding-plan keys call the same pay-per-token catalog id (released
+# 2026-09-09, llm-gateway 00350); plan usage is flat-fee, so cost is 0 here.
+# limit.output is cap-1 on purpose: the gateway 400s max_tokens >=
+# max_completion_tokens (131072), so 131071 is the largest value that serves
+# (its advertised `recommended_max_tokens`).
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_048_576
+output = 131_071
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/umans-ai-coding-plan/models/umans-kimi-k2.7.toml b/providers/umans-ai-coding-plan/models/umans-kimi-k2.7.toml
deleted file mode 100644
index 43ef003cf3a..00000000000
--- a/providers/umans-ai-coding-plan/models/umans-kimi-k2.7.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "moonshotai/kimi-k2.7-code"
-name = "Kimi K2.7 Code"
-# Live metadata says reasoning is always on (`can_disable=false`) with no effort
-# levels (accessed 2026-06-25): https://api.code.umans.ai/v1/models/info
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
diff --git a/providers/umans-ai/models/umans-coder.toml b/providers/umans-ai/models/umans-coder.toml
index 7bec758f0a0..f4f91982bbb 100644
--- a/providers/umans-ai/models/umans-coder.toml
+++ b/providers/umans-ai/models/umans-coder.toml
@@ -1,21 +1,23 @@
-base_model = "moonshotai/kimi-k2.7-code"
+base_model = "zhipuai/glm-5.3-flash"
 name = "Umans Coder"
 temperature = false
-# This alias currently routes to Kimi K2.7 Code: reasoning is always on
-# (`can_disable=false`) with no effort levels (accessed 2026-06-25).
+# This alias routes to GLM-5.3-Flash since 2026-09-09 (it rode Kimi K2.7 Code
+# before): reasoning is always on (`can_disable=false`) WITH effort levels
+# low|high|max, default max (accessed 2026-09-08):
 # https://api.code.umans.ai/v1/models/info
-reasoning_options = []
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.95
-output = 4.00
-cache_read = 0.19
+input = 0.15
+output = 0.50
+cache_read = 0.03
 
 [limit]
-output = 32_768
+context = 1_048_576
+output = 131_071
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
diff --git a/providers/umans-ai/models/umans-glm-5.2.toml b/providers/umans-ai/models/umans-glm-5.2.toml
deleted file mode 100644
index 9e2184d3818..00000000000
--- a/providers/umans-ai/models/umans-glm-5.2.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2"
-# Live metadata: levels none/high/max, default high, and `can_disable=true`
-# (accessed 2026-06-25): https://api.code.umans.ai/v1/models/info
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.40
-output = 4.40
-cache_read = 0.26
-
-[limit]
-context = 405_504
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/umans-ai/models/umans-glm-5.3-flash.toml b/providers/umans-ai/models/umans-glm-5.3-flash.toml
new file mode 100644
index 00000000000..60262713ba9
--- /dev/null
+++ b/providers/umans-ai/models/umans-glm-5.3-flash.toml
@@ -0,0 +1,30 @@
+# Effort only: reasoning_effort = low|high|max — thinking is ALWAYS ON (the
+# checkpoint's chat template has no non-thinking mode; verified against the
+# released repo 2026-08-26 and the live lab card since then). The released
+# id's live metadata mirrors that shape: levels low/high/max, default max,
+# `can_disable=false`: https://api.code.umans.ai/v1/models/info
+# Pay-per-token (llm-gateway 00350, released 2026-09-09): Umans AI's public
+# rate, which is what GET /v1/models renders — the publisher's list rates
+# ($0.15 / $0.50 / $0.03 per 1M); the 50% launch promo (to 2026-09-09) is not
+# mirrored.
+# limit.output is cap-1 on purpose: the gateway 400s max_tokens >=
+# max_completion_tokens (131072), so 131071 is the largest value that serves
+# (its advertised `recommended_max_tokens`).
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.50
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 131_071
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/umans-ai/models/umans-kimi-k2.7.toml b/providers/umans-ai/models/umans-kimi-k2.7.toml
deleted file mode 100644
index 913774482d5..00000000000
--- a/providers/umans-ai/models/umans-kimi-k2.7.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "moonshotai/kimi-k2.7-code"
-name = "Kimi K2.7 Code"
-# Live metadata says reasoning is always on (`can_disable=false`) with no effort
-# levels (accessed 2026-06-25): https://api.code.umans.ai/v1/models/info
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.95
-output = 4.00
-cache_read = 0.19
-
-[limit]
-output = 32_768

From 86cfd2450e8f994bfccb29bc7324a2148cc12735 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 01:36:10 +0000
Subject: [PATCH 1484/2012] chore(sync): update Kilo model catalog (#7284)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..283ade8c995 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From 29e03d7f4b77c056c816f9efe2f343ea97f43570 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 01:36:11 +0000
Subject: [PATCH 1485/2012] chore(sync): update NanoGPT model catalog (#7283)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From ca68f80970b9d9be7b21cc4fbe9cf205c3f5238d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 01:36:15 +0000
Subject: [PATCH 1486/2012] chore(sync): update OpenRouter model catalog
 (#7285)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..75641321a8f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From 6092333750ba7d422ad45e52ed1ad0c3e8451fa8 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Wed, 16 Sep 2026 20:40:27 -0500
Subject: [PATCH 1487/2012] fix(fireworks): clean up serverless catalog (#7286)

---
 .../accounts/fireworks/models/deepseek-v4-flash-0731.toml   | 2 ++
 .../fireworks/models/deepseek-v4-flash-vision-exp.toml      | 2 ++
 .../accounts/fireworks/models/deepseek-v4-pro-0813.toml     | 2 ++
 .../models/accounts/fireworks/models/deepseek-v4-pro.toml   | 2 ++
 .../models/accounts/fireworks/models/glm-5p2.toml           | 2 ++
 .../models/accounts/fireworks/models/kimi-k2p6.toml         | 2 ++
 .../models/accounts/fireworks/models/kimi-k2p7-code.toml    | 2 ++
 .../models/accounts/fireworks/models/minimax-m2p7.toml      | 2 ++
 .../accounts/fireworks/models/mistral-large-3-fp8.toml      | 6 ------
 .../models/accounts/fireworks/models/muse-glimmer-30b.toml  | 2 ++
 10 files changed, 18 insertions(+), 6 deletions(-)
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
index 7bb811f6613..276c95f212a 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
@@ -1,4 +1,6 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 base_model = "deepseek/deepseek-v4-flash-0731"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml
index e3955a13717..0e06ace6ae7 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-vision-exp.toml
@@ -1,8 +1,10 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 # Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
 # Effort: reasoning_effort = low|high|max (same surface as deepseek-v4-flash-0731 on Fireworks)
 # Pricing/model: https://fireworks.ai/models/deepseek-ai/deepseek-v4-flash-vision-exp (accessed 2026-09-02)
 # Peer: providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash-0731.toml
 base_model = "deepseek/deepseek-v4-flash-vision-exp"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "toggle"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
index 031dd2a013b..38881be88e1 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
@@ -1,8 +1,10 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 # Toggle: thinking.type = enabled|disabled (reasoning_effort = none|false also disables)
 # Effort: reasoning_effort = high|max; low/medium promote to high, xhigh to max
 # Reasoning: https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-08-14)
 # Pricing: https://fireworks.ai/models/deepseek-ai/deepseek-v4-pro-0813 (accessed 2026-08-14)
 base_model = "deepseek/deepseek-v4-pro-0813"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
index 84e0b5d334e..6af4675f1f0 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro.toml
@@ -1,3 +1,4 @@
+# Deprecated effective 2026-08-27: https://docs.fireworks.ai/updates/changelog
 # GET /v1/serverless/models lists DeepSeek V4 Pro on Standard and Priority
 # at $1.20 / $0.60 / $1.20 per 1M (input / cached input / output) on
 # 2026-09-16. Both serving modes currently have the same price.
@@ -6,6 +7,7 @@
 # Reasoning API: https://docs.fireworks.ai/api-reference/post-chatcompletions (accessed 2026-09-15)
 # Peer surface: providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-pro-0813.toml
 base_model = "deepseek/deepseek-v4-pro"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
index bf70c694ba9..8cfa16eeca2 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml
@@ -1,3 +1,4 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 # Fireworks' published schema does not document this model's toggle or "max"
 # effort mapping. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
 name = "GLM 5.2"
@@ -10,6 +11,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = true
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
index cf312414120..f2e97c2cc21 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p6.toml
@@ -1,3 +1,4 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 # Fireworks documents thinking.type = "enabled", but no explicit disabled
 # value for this model. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
 name = "Kimi K2.6"
@@ -10,6 +11,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = true
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
index 4db378e3e5e..b69c7ada8f2 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml
@@ -1,3 +1,4 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 # Fireworks documents thinking.type = "enabled", but no explicit disabled
 # value for this model. https://docs.fireworks.ai/guides/reasoning (accessed 2026-06-25)
 name = "Kimi K2.7 Code"
@@ -10,6 +11,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = true
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
index b64465a927f..bbf3b9c635c 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/minimax-m2p7.toml
@@ -1,9 +1,11 @@
+# Deprecated effective 2026-08-27: https://docs.fireworks.ai/updates/changelog
 # GET /v1/serverless/models lists MiniMax M2.7 as Priority-only at
 # $1.20 / $0.60 / $1.20 per 1M (input / cached input / output) with a
 # 196,608-token context on 2026-09-16.
 # $.reasoning_effort = "low" | "medium" | "high" (same Fireworks surface as MiniMax M3).
 # https://docs.fireworks.ai/guides/reasoning
 base_model = "minimax/MiniMax-M2.7"
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml b/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml
deleted file mode 100644
index 4f5831e7930..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/models/mistral-large-3-fp8.toml
+++ /dev/null
@@ -1,6 +0,0 @@
-# Fireworks identifies this on-demand FP8 endpoint as Mistral Large 3 675B Instruct 2512,
-# with 256K context, function calling, and image input. It is not offered serverlessly,
-# so there is no per-token price to record:
-# https://fireworks.ai/models/fireworks/mistral-large-3-fp8 (accessed 2026-09-05)
-base_model = "mistral/mistral-large-2512"
-name = "Mistral Large 3 675B Instruct 2512"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/models/muse-glimmer-30b.toml b/providers/fireworks-ai/models/accounts/fireworks/models/muse-glimmer-30b.toml
index f86f459d872..4a73f9bf085 100644
--- a/providers/fireworks-ai/models/accounts/fireworks/models/muse-glimmer-30b.toml
+++ b/providers/fireworks-ai/models/accounts/fireworks/models/muse-glimmer-30b.toml
@@ -1,4 +1,6 @@
+# Serverless deprecation announced for 2026-09-25: https://docs.fireworks.ai/updates/changelog
 base_model = "meta/muse-glimmer-30b"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "effort"

From d4073ed8eb65ae20f0b0f3a0cb25e5387706fb4a Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Wed, 16 Sep 2026 21:06:59 -0500
Subject: [PATCH 1488/2012] fix(core): reconcile Fireworks inventory (#7287)

---
 .../core/src/sync/providers/fireworks-ai.ts   | 106 +++++++++++++++---
 packages/core/test/fireworks-ai-sync.test.ts  |  98 ++++++++++++++++
 2 files changed, 188 insertions(+), 16 deletions(-)

diff --git a/packages/core/src/sync/providers/fireworks-ai.ts b/packages/core/src/sync/providers/fireworks-ai.ts
index e6fc5cf6bc5..96e28194d97 100644
--- a/packages/core/src/sync/providers/fireworks-ai.ts
+++ b/packages/core/src/sync/providers/fireworks-ai.ts
@@ -4,6 +4,7 @@ import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "
 import { factorBaseModel } from "./openrouter.js";
 
 const API_ENDPOINT = "https://api.fireworks.ai/v1/serverless/models";
+const INVENTORY_ENDPOINT = "https://api.fireworks.ai/v1/accounts/fireworks/models";
 
 const FireworksPrice = z.object({
   sku: z.string().min(1),
@@ -33,55 +34,84 @@ export const FireworksResponse = z.object({
   data: z.array(FireworksModel),
 }).passthrough();
 
+export const FireworksInventoryModel = z.object({
+  name: z.string().min(1),
+  kind: z.string().min(1),
+  supportsServerless: z.boolean(),
+}).passthrough();
+
+export const FireworksInventoryResponse = z.object({
+  models: z.array(FireworksInventoryModel),
+  nextPageToken: z.string().optional(),
+}).passthrough();
+
+const FireworksSyncResponse = z.object({
+  serverless: FireworksResponse,
+  inventory: z.array(FireworksInventoryModel),
+});
+
 export type FireworksModel = z.infer<typeof FireworksModel>;
+export type FireworksInventoryModel = z.infer<typeof FireworksInventoryModel>;
 export type FireworksCatalogModel = FireworksModel & {
   catalogId: string;
   flagModes: FireworksModel[];
 };
+export type FireworksInventoryCatalogModel = {
+  catalogId: string;
+  inventoryOnly: true;
+};
+
+type FireworksSourceModel = FireworksCatalogModel | FireworksInventoryCatalogModel;
 
 export const fireworksAi = {
   id: "fireworks-ai",
   name: "Fireworks AI",
   modelsDir: "providers/fireworks-ai/models",
   skipCreates: true,
-  // The endpoint describes the public serverless catalog, but it still lacks
-  // enough intrinsic metadata and reasoning controls to create safe entries.
-  deleteMissing: false,
+  // The pricing feed supplies invocation IDs and serving modes, while List
+  // Models fills inventory gaps. A model absent from both is not serverless.
+  deleteMissing: true,
   sourceID(model) {
     return supportsCatalogModel(model) ? model.catalogId : undefined;
   },
   skippedNotice(ids) {
     if (ids.length === 0) return [];
     return [
-      `${ids.length} Fireworks serverless text/vision IDs were not created because the endpoint does not yet provide output limits, reasoning controls, tool support, or open-weight status. Existing models are still updated from API-authoritative fields.`,
+      `${ids.length} Fireworks serverless text/vision IDs were not created because the sources do not yet provide enough output limits, reasoning controls, tool support, or open-weight status. Existing models are still updated from API-authoritative fields.`,
       `Skipped remote IDs: ${ids.map((id) => `\`${id}\``).join(", ")}`,
     ];
   },
-  missingNotice(paths) {
-    if (paths.length === 0) return [];
-    return [
-      `${paths.length} local Fireworks models were absent from the serverless catalog and were retained for manual lifecycle review.`,
-      `Retained local paths: ${paths.map((item) => `\`${item}\``).join(", ")}`,
-    ];
-  },
   async fetchModels() {
     const key = process.env.FIREWORKS_API_KEY;
     if (key === undefined) throw new Error("Fireworks AI sync requires FIREWORKS_API_KEY");
-    return fetchFireworksModels(key);
+    const [serverless, inventory] = await Promise.all([
+      fetchFireworksModels(key),
+      fetchFireworksInventory(key),
+    ]);
+    return { serverless, inventory };
   },
   parseModels(raw) {
-    return expandFireworksModels(FireworksResponse.parse(raw).data);
+    const parsed = FireworksSyncResponse.parse(raw);
+    return mergeFireworksModels(parsed.serverless.data, parsed.inventory);
   },
   translateModel(model, context) {
     if (!supportsCatalogModel(model)) return undefined;
     const existing = context.existing(model.catalogId);
     if (existing === undefined) return undefined;
+    const authored = context.authored(model.catalogId);
+    if ("inventoryOnly" in model && authored === undefined) {
+      throw new Error(`Fireworks AI model ${model.catalogId} has no local TOML to preserve`);
+    }
+    if ("inventoryOnly" in model) {
+      // The shared runner validates translated models before writing them.
+      return { id: model.catalogId, model: authored as SyncedModel };
+    }
     return {
       id: model.catalogId,
       model: buildFireworksModel(model, existing),
     };
   },
-} satisfies SyncProvider<FireworksCatalogModel>;
+} satisfies SyncProvider<FireworksSourceModel>;
 
 export async function fetchFireworksModels(
   key: string,
@@ -96,6 +126,50 @@ export async function fetchFireworksModels(
   return FireworksResponse.parse(await response.json());
 }
 
+export async function fetchFireworksInventory(
+  key: string,
+  fetcher: typeof fetch = fetch,
+  pageToken?: string,
+): Promise<FireworksInventoryModel[]> {
+  const url = new URL(INVENTORY_ENDPOINT);
+  url.searchParams.set("filter", "supports_serverless=true");
+  url.searchParams.set("pageSize", "200");
+  if (pageToken !== undefined) url.searchParams.set("pageToken", pageToken);
+  const response = await fetcher(url, {
+    headers: { Authorization: `Bearer ${key}` },
+  });
+  if (!response.ok) {
+    throw new Error(`Fireworks AI inventory request failed: ${response.status} ${response.statusText}`);
+  }
+  const page = FireworksInventoryResponse.parse(await response.json());
+  if (page.nextPageToken === undefined || page.nextPageToken.length === 0) return page.models;
+  return [...page.models, ...await fetchFireworksInventory(key, fetcher, page.nextPageToken)];
+}
+
+export function mergeFireworksModels(
+  serverless: FireworksModel[],
+  inventory: FireworksInventoryModel[],
+): FireworksSourceModel[] {
+  if (serverless.length === 0 || inventory.length === 0) {
+    throw new Error("Fireworks AI returned an empty serverless source; refusing destructive sync");
+  }
+  const expanded = expandFireworksModels(serverless);
+  const ids = new Set(expanded.map((model) => model.catalogId));
+  return [
+    ...expanded,
+    ...inventory
+      .filter((model) =>
+        model.kind === "HF_BASE_MODEL"
+        && model.supportsServerless
+        && !ids.has(model.name)
+      )
+      .map((model): FireworksInventoryCatalogModel => ({
+        catalogId: model.name,
+        inventoryOnly: true,
+      })),
+  ];
+}
+
 export function expandFireworksModels(models: FireworksModel[]): FireworksCatalogModel[] {
   const expanded = new Map<string, FireworksCatalogModel>();
   const grouped = Map.groupBy(models, (model) => model.id);
@@ -125,8 +199,8 @@ export function expandFireworksModels(models: FireworksModel[]): FireworksCatalo
   }
 }
 
-function supportsCatalogModel(model: FireworksCatalogModel) {
-  return model.output_modalities.includes("text");
+function supportsCatalogModel(model: FireworksSourceModel) {
+  return "inventoryOnly" in model || model.output_modalities.includes("text");
 }
 
 type Modality = SyncedFullModel["modalities"]["input"][number];
diff --git a/packages/core/test/fireworks-ai-sync.test.ts b/packages/core/test/fireworks-ai-sync.test.ts
index 4f352794ab7..57b95a2c9f6 100644
--- a/packages/core/test/fireworks-ai-sync.test.ts
+++ b/packages/core/test/fireworks-ai-sync.test.ts
@@ -4,9 +4,12 @@ import type { ExistingModel } from "../src/sync/index.js";
 import {
   buildFireworksModel,
   expandFireworksModels,
+  fetchFireworksInventory,
   fetchFireworksModels,
+  type FireworksInventoryModel,
   FireworksResponse,
   fireworksAi,
+  mergeFireworksModels,
   type FireworksCatalogModel,
   type FireworksModel,
 } from "../src/sync/providers/fireworks-ai.js";
@@ -40,6 +43,92 @@ test("parses the Fireworks serverless model list", () => {
   });
 });
 
+test("fetches Fireworks serverless inventory with the documented filter", async () => {
+  let request: Request | undefined;
+  const fetcher = (async (input: string | URL | Request, init?: RequestInit) => {
+    request = input instanceof Request
+      ? new Request(input, init)
+      : new Request(input.toString(), init);
+    return Response.json({ models: [inventoryModel()] });
+  }) as unknown as typeof fetch;
+
+  await fetchFireworksInventory("test-key", fetcher);
+
+  expect(request?.url).toStartWith("https://api.fireworks.ai/v1/accounts/fireworks/models?");
+  expect(new URL(request!.url).searchParams.get("filter")).toBe("supports_serverless=true");
+  expect(new URL(request!.url).searchParams.get("pageSize")).toBe("200");
+  expect(request?.headers.get("authorization")).toBe("Bearer test-key");
+});
+
+test("fetches every Fireworks serverless inventory page", async () => {
+  const requests: Request[] = [];
+  const fetcher = (async (input: string | URL | Request, init?: RequestInit) => {
+    const request = input instanceof Request
+      ? new Request(input, init)
+      : new Request(input.toString(), init);
+    requests.push(request);
+    return Response.json(
+      request.url.includes("pageToken=next")
+        ? { models: [inventoryModel({ name: "accounts/fireworks/models/two" })] }
+        : {
+          models: [inventoryModel({ name: "accounts/fireworks/models/one" })],
+          nextPageToken: "next",
+        },
+    );
+  }) as unknown as typeof fetch;
+
+  const models = await fetchFireworksInventory("test-key", fetcher);
+
+  expect(models.map((model) => model.name)).toEqual([
+    "accounts/fireworks/models/one",
+    "accounts/fireworks/models/two",
+  ]);
+  expect(new URL(requests[1]!.url).searchParams.get("pageToken")).toBe("next");
+});
+
+test("unions pricing IDs with generation models from serverless inventory", () => {
+  const models = mergeFireworksModels(
+    [fireworksModel()],
+    [
+      inventoryModel({ name: "accounts/fireworks/models/example" }),
+      inventoryModel({ name: "accounts/fireworks/models/inventory-only" }),
+      inventoryModel({ name: "accounts/fireworks/models/embedding", kind: "EMBEDDING_MODEL" }),
+      inventoryModel({ name: "accounts/fireworks/models/on-demand", supportsServerless: false }),
+    ],
+  );
+
+  expect(models.map((model) => model.catalogId)).toEqual([
+    "accounts/fireworks/models/example",
+    "accounts/fireworks/models/inventory-only",
+  ]);
+  expect(models[1]).toEqual({
+    catalogId: "accounts/fireworks/models/inventory-only",
+    inventoryOnly: true,
+  });
+});
+
+test("refuses destructive sync when either Fireworks source is empty", () => {
+  expect(() => mergeFireworksModels([], [inventoryModel()])).toThrow("empty serverless source");
+  expect(() => mergeFireworksModels([fireworksModel()], [])).toThrow("empty serverless source");
+});
+
+test("preserves inventory-only models while enabling deletion for models absent from both sources", () => {
+  const authored = {
+    base_model: "example/example",
+    cost: { input: 1, output: 2 },
+  };
+  const translated = fireworksAi.translateModel({
+    catalogId: "accounts/fireworks/models/example",
+    inventoryOnly: true,
+  }, {
+    existing: () => existingModel(),
+    authored: () => authored,
+  });
+
+  expect(fireworksAi.deleteMissing).toBe(true);
+  expect(translated?.model).toEqual(authored);
+});
+
 test("expands usage identifiers and aliases and attaches flag-only modes", () => {
   const models = expandFireworksModels([
     fireworksModel({ aliases: ["accounts/fireworks/routers/example-latest"] }),
@@ -190,6 +279,15 @@ function fireworksModel(overrides: Partial<FireworksModel> = {}): FireworksModel
   };
 }
 
+function inventoryModel(overrides: Partial<FireworksInventoryModel> = {}): FireworksInventoryModel {
+  return {
+    name: "accounts/fireworks/models/example",
+    kind: "HF_BASE_MODEL",
+    supportsServerless: true,
+    ...overrides,
+  };
+}
+
 function catalogModel(overrides: Partial<FireworksCatalogModel> = {}): FireworksCatalogModel {
   const model = fireworksModel(overrides);
   return {

From 60bca35cfb0001ffe0507279ded570c301706779 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 03:30:58 +0000
Subject: [PATCH 1489/2012] chore(sync): update OpenRouter model catalog
 (#7289)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 4 ++--
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 12c99d28b38..f6dda11044a 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.1
+input = 0.03
+output = 0.13
 cache_read = 0.01
 
 [limit]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 9d68ba72961..0d21da4aeaf 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.833
-output = 2.618
-cache_read = 0.13685
+input = 0.8925
+output = 2.805
+cache_read = 0.146625
 
 [limit]
 context = 1_310_720

From 655eadbbc721e331e8f544efc7a5c09c281270c4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 03:31:04 +0000
Subject: [PATCH 1490/2012] chore(sync): update Kilo model catalog (#7288)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 4 ++--
 providers/kilo/models/~z-ai/glm-latest.toml                 | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index b586ef3f991..473252b6895 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.1
+input = 0.03
+output = 0.13
 cache_read = 0.01
 
 [limit]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7effc6dceef..da0527bbfae 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.833
-output = 2.618
-cache_read = 0.13685
+input = 0.8925
+output = 2.805
+cache_read = 0.146625
 
 [limit]
 context = 1_048_576

From ab28419166a8ed62579de5d73c735bbbd5bbddf4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 04:30:57 +0000
Subject: [PATCH 1491/2012] chore(sync): update Kilo model catalog (#7290)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 283ade8c995..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index da0527bbfae..38aa9fed430 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8925
-output = 2.805
-cache_read = 0.146625
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 09a20adc4551fc9f74e7333b32b214283eeff3b1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 04:30:59 +0000
Subject: [PATCH 1492/2012] chore(sync): update NanoGPT model catalog (#7292)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From 00aa2be0d5fb20993e4b2d13a0fcfd58de95c978 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 04:31:08 +0000
Subject: [PATCH 1493/2012] chore(sync): update OpenRouter model catalog
 (#7291)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 75641321a8f..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0d21da4aeaf..3e3ce108967 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8925
-output = 2.805
-cache_read = 0.146625
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From 1e171ec5732b75cdd13d0b7fb37e79a54944a6b3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 05:27:26 +0000
Subject: [PATCH 1494/2012] chore(sync): update Eden AI model catalog (#7295)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/amazon/amazon.nova-lite-v1:0.toml     |  1 +
 .../models/amazon/amazon.nova-lite-v1:0@us.toml  |  1 +
 .../models/amazon/amazon.nova-micro-v1:0.toml    |  1 +
 .../models/amazon/amazon.nova-micro-v1:0@us.toml |  1 +
 .../models/amazon/amazon.nova-pro-v1:0.toml      |  1 +
 .../models/amazon/amazon.nova-pro-v1:0@us.toml   |  1 +
 .../nebius/deepseek-ai/DeepSeek-V4.1-Flash.toml  | 16 ++++++++++++++++
 7 files changed, 22 insertions(+)
 create mode 100644 providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
index ea127cc7b46..f7657f661d2 100644
--- a/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
+++ b/providers/edenai/models/amazon/amazon.nova-lite-v1:0.toml
@@ -4,6 +4,7 @@ structured_output = true
 [cost]
 input = 0.06
 output = 0.24
+cache_read = 0.015
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
index db196fcd529..61fa496d779 100644
--- a/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
+++ b/providers/edenai/models/amazon/amazon.nova-lite-v1:0@us.toml
@@ -5,6 +5,7 @@ structured_output = true
 [cost]
 input = 0.06
 output = 0.24
+cache_read = 0.015
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
index f742ca81b6b..0b42211b2af 100644
--- a/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
+++ b/providers/edenai/models/amazon/amazon.nova-micro-v1:0.toml
@@ -4,3 +4,4 @@ structured_output = true
 [cost]
 input = 0.035
 output = 0.14
+cache_read = 0.00875
diff --git a/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
index c0229f11c40..307e8239aee 100644
--- a/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
+++ b/providers/edenai/models/amazon/amazon.nova-micro-v1:0@us.toml
@@ -5,3 +5,4 @@ structured_output = true
 [cost]
 input = 0.035
 output = 0.14
+cache_read = 0.00875
diff --git a/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml b/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
index d8ba4575921..5dda1e65f8b 100644
--- a/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
+++ b/providers/edenai/models/amazon/amazon.nova-pro-v1:0.toml
@@ -4,6 +4,7 @@ structured_output = true
 [cost]
 input = 0.8
 output = 3.2
+cache_read = 0.2
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml b/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
index 25bbe563d46..5185388f1ca 100644
--- a/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
+++ b/providers/edenai/models/amazon/amazon.nova-pro-v1:0@us.toml
@@ -5,6 +5,7 @@ structured_output = true
 [cost]
 input = 0.8
 output = 3.2
+cache_read = 0.2
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..48e159fa2bc
--- /dev/null
+++ b/providers/edenai/models/nebius/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,16 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Nebius)"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.3
+
+[limit]
+context = 1_048_000

From 7bd0ac37a1d839bac81df43feb0755e918790f1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 06:42:06 +0000
Subject: [PATCH 1495/2012] chore(sync): update Kilo model catalog (#7299)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..283ade8c995 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From 65b5344ac1b63f3c71dfcd1f2cfc1d60930638e3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 06:42:10 +0000
Subject: [PATCH 1496/2012] chore(sync): update OpenRouter model catalog
 (#7298)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..75641321a8f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From 621c2b24fd4c0642fcacca04433cffaed4238c01 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 06:42:14 +0000
Subject: [PATCH 1497/2012] chore(sync): update NanoGPT model catalog (#7300)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From 7d2481e74c496c9bb1b86eaee052c637ee85d0ae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 07:30:46 +0000
Subject: [PATCH 1498/2012] chore(sync): update OpenRouter model catalog
 (#7302)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-4.6.toml | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.5
+output = 2
+cache_read = 0.1

From 18ee3a4b6b838e56cf394b527ef76e7b67f633b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 07:30:53 +0000
Subject: [PATCH 1499/2012] chore(sync): update Kilo model catalog (#7301)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fd68fe0c68b 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
+input = 0.5
+output = 2
+cache_read = 0.1
 
 [limit]
-context = 198_000
-output = 16_384
+context = 202_752

From bdcd68d2156517b550d7b5090b6ab5628849cd04 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 08:35:03 +0000
Subject: [PATCH 1500/2012] chore(sync): update Kilo model catalog (#7304)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 2a55dc9d9d710e4f17352a1b72cbdbb55aa3faa7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 08:35:09 +0000
Subject: [PATCH 1501/2012] chore(sync): update OpenRouter model catalog
 (#7305)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From cc3dcc91d87a6f81800097cddc4f5afa56ee51f2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:29:38 +0000
Subject: [PATCH 1502/2012] chore(sync): update Kilo model catalog (#7308)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
 
 [limit]
-context = 202_752
+context = 198_000
+output = 16_384

From 8c9bdfa5cdc7033a17f6821a11d636592894d1eb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:29:48 +0000
Subject: [PATCH 1503/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#7310)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ovhcloud/models/qwen3.8-27b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/ovhcloud/models/qwen3.8-27b.toml b/providers/ovhcloud/models/qwen3.8-27b.toml
index 45ae7920a40..cc5da21fd15 100644
--- a/providers/ovhcloud/models/qwen3.8-27b.toml
+++ b/providers/ovhcloud/models/qwen3.8-27b.toml
@@ -22,8 +22,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0
-output = 0
+input = 0.47
+output = 3.19
 
 [limit]
 context = 262_144

From cbf13755fd223585bbbc84c56f21dc51963f8bed Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:29:52 +0000
Subject: [PATCH 1504/2012] chore(sync): update Cortecs model catalog (#7311)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/llama-3.1-405b-instruct.toml       | 25 -------------------
 1 file changed, 25 deletions(-)
 delete mode 100644 providers/cortecs/models/llama-3.1-405b-instruct.toml

diff --git a/providers/cortecs/models/llama-3.1-405b-instruct.toml b/providers/cortecs/models/llama-3.1-405b-instruct.toml
deleted file mode 100644
index cc85efa5deb..00000000000
--- a/providers/cortecs/models/llama-3.1-405b-instruct.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Llama 3.1 405B Instruct"
-description = "Open Llama instruction model for multilingual chat, reasoning, and coding"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2023-12"
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 1.95
-output = 1.95
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

From f70ae6b4fd14270feca45dc7333219f5d0121bd6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:30:02 +0000
Subject: [PATCH 1505/2012] chore(sync): update OpenRouter model catalog
 (#7309)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-4.6.toml | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384

From 84c3bf3eda0e2b084a94877302db753d99c0c9d0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 10:27:23 +0000
Subject: [PATCH 1506/2012] chore(sync): update OpenRouter model catalog
 (#7313)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml  | 6 +++---
 providers/openrouter/models/meta/muse-glimmer-30b.toml   | 4 ++--
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 +++------
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 5 files changed, 14 insertions(+), 17 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index 3b46ee2987a..12578d9072c 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -10,8 +10,8 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.35
-output = 1.5
+input = 0.3
+output = 1.1
 cache_read = 0.04
 
 [limit]
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.5
+output = 2
+cache_read = 0.1
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 75641321a8f..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From f56290e2730b6931ae7780dca7c34c537c90090f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 10:27:28 +0000
Subject: [PATCH 1507/2012] chore(sync): update Kilo model catalog (#7312)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml                  | 9 ++++-----
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fd68fe0c68b 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
+input = 0.5
+output = 2
+cache_read = 0.1
 
 [limit]
-context = 198_000
-output = 16_384
+context = 202_752
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 283ade8c995..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From 1f096331ea83a1ff88e7d58b8661c9ce02ad631e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 10:27:42 +0000
Subject: [PATCH 1508/2012] chore(sync): update NanoGPT model catalog (#7314)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From 98048c8e3648cd4299c6741184f62197eefc0abf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 12:37:18 +0000
Subject: [PATCH 1509/2012] chore(sync): update Kilo model catalog (#7317)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
 
 [limit]
-context = 202_752
+context = 198_000
+output = 16_384

From e300daf726676c68c166404af6ef907b904ba9d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 12:37:31 +0000
Subject: [PATCH 1510/2012] chore(sync): update OpenRouter model catalog
 (#7316)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-4.6.toml | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384

From 602f31df3b811cacfdd298ed9aa0549f1527408e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 13:26:41 +0000
Subject: [PATCH 1511/2012] chore(sync): update OpenRouter model catalog
 (#7318)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..3eee6643f71 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.07
+output = 0.14
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..50d7ce57993 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 2c80148927e930cd6769130b548eda5fecde92dc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 13:26:55 +0000
Subject: [PATCH 1512/2012] chore(sync): update Kilo model catalog (#7319)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 1 +
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..957bc0a82fa 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 2a6bf78db2f55d715e60f0f38086863b2838796a Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:22:12 -0500
Subject: [PATCH 1513/2012] fix: Add Fireworks AI US-only Serverless router
 models and pricing (#7294)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../routers/deepseek-v4-flash-0731-us.toml    | 30 ++++++++++++++
 .../fireworks/routers/glm-5p2-fast-us.toml    | 33 +++++++++++++++
 .../fireworks/routers/glm-5p2-us.toml         | 34 +++++++++++++++
 .../fireworks/routers/glm-5p3-flash-us.toml   | 32 +++++++++++++++
 .../fireworks/routers/glm-5p3-us.toml         | 30 ++++++++++++++
 .../fireworks/routers/kimi-k3-us.toml         | 41 +++++++++++++++++++
 6 files changed, 200 insertions(+)
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
 create mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
new file mode 100644
index 00000000000..846c51d942e
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
@@ -0,0 +1,30 @@
+# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
+# global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
+# Base DeepSeek V4 Flash (0731) Standard: $0.22 / $0.007 / $0.66 → US
+# $0.33 / $0.0105 / $0.99 (USD/MTok, input / cached input / output).
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash (0731) US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.33
+output = 0.99
+cache_read = 0.0105
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
new file mode 100644
index 00000000000..edbace0f24a
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
@@ -0,0 +1,33 @@
+# US-only Serverless Fast router. Must be called via us.api.fireworks.ai (not
+# the global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# GLM 5.2 Fast US is an explicit exception to the 1.5x US premium and matches
+# global GLM 5.2 Fast: $2.10 / $0.21 / $6.60 (USD/MTok, input / cached input /
+# output). https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = high|max
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2 Fast US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 2.1
+output = 6.6
+cache_read = 0.21
+
+[limit]
+context = 1_048_575
+output = 131_072
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
new file mode 100644
index 00000000000..2e6423c0de4
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
@@ -0,0 +1,34 @@
+# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
+# global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
+# Base GLM 5.2 Standard: $1.40 / $0.14 / $4.40 → US $2.10 / $0.21 / $6.60
+# (USD/MTok, input / cached input / output).
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = high|max
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2 US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 2.1
+output = 6.6
+cache_read = 0.21
+
+[limit]
+context = 1_048_575
+output = 131_072
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
new file mode 100644
index 00000000000..aa9054e1cf0
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
@@ -0,0 +1,32 @@
+# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
+# global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
+# Base GLM 5.3 Flash Standard: $0.15 / $0.03 / $0.50 → US $0.225 / $0.045 /
+# $0.75 (USD/MTok, input / cached input / output).
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Thinking-only on Fireworks (no toggle): reasoning_effort = low|high|max.
+# Limits/modalities aligned with global fireworks glm-5p3-flash.
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.225
+output = 0.75
+cache_read = 0.045
+
+[limit]
+context = 1_048_573
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
new file mode 100644
index 00000000000..3fd7ef4be2d
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
@@ -0,0 +1,30 @@
+# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
+# global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
+# Base GLM 5.3 Standard: $1.40 / $0.26 / $4.40 → US $2.10 / $0.39 / $6.60
+# (USD/MTok, input / cached input / output).
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Thinking-only on Fireworks (no toggle): reasoning_effort = low|high|max.
+# Limits aligned with global fireworks glm-5p3 serving probes.
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 2.1
+output = 6.6
+cache_read = 0.39
+
+[limit]
+context = 1_048_573
+output = 262_144
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml
new file mode 100644
index 00000000000..f1970262aad
--- /dev/null
+++ b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml
@@ -0,0 +1,41 @@
+# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
+# global api.fireworks.ai endpoint).
+# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
+# Standard + Priority pricing (USD/MTok, input / cached input / output):
+# $4.50 / $0.45 / $22.50 and $5.625 / $0.5625 / $28.125
+# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|medium|high|max
+# Budget: thinking.budget_tokens (N >= 1024; conflicts with reasoning_effort)
+# https://docs.fireworks.ai/guides/reasoning
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 US"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1_024
+
+[cost]
+input = 4.5
+output = 22.5
+cache_read = 0.45
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+api = "https://us.api.fireworks.ai/inference/v1/"
+
+[experimental.modes.priority]
+cost = { input = 5.625, output = 28.125, cache_read = 0.5625 }
+provider = { body = { service_tier = "priority" } }

From 94e6ff9a5ecbcdfb542776de3ae98674c02f3898 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E8=BD=BB=E5=B0=98?=
 <53001305+SlightDust@users.noreply.github.com>
Date: Thu, 17 Sep 2026 22:22:21 +0800
Subject: [PATCH 1514/2012] feat: add DeepSeek-V4.1-Flash to SCNet Token Plan
 (#7297)

---
 .../models/DeepSeek-V4.1-Flash.toml             | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/scnet-token-plan/models/DeepSeek-V4.1-Flash.toml

diff --git a/providers/scnet-token-plan/models/DeepSeek-V4.1-Flash.toml b/providers/scnet-token-plan/models/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..3b224be63f4
--- /dev/null
+++ b/providers/scnet-token-plan/models/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,17 @@
+# Toggle: enable_thinking = true|false
+# Effort: reasoning_effort = high|max
+# https://www.scnet.cn/ac/openapi/doc/2.0/moduleapi/api/chat.html (accessed 2026-09-15)
+base_model = "deepseek/deepseek-v4.1-flash"
+attachment = false
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[modalities]
+input = ["text"]

From 09fd4392b4b4d3c35da816585cc3ad7a86c607d1 Mon Sep 17 00:00:00 2001
From: Haseeb Ahmed <thehaseebahmed@outlook.com>
Date: Thu, 17 Sep 2026 16:23:10 +0200
Subject: [PATCH 1515/2012] feat(azure): add GPT Image 2.5 models (#7315)

* feat(azure): add GPT Image 2.5 models

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(models): restore GPT Image 2.5 capabilities

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---
 models/openai/gpt-image-2.5-flare.toml         | 18 ++++++++++++++++++
 models/openai/gpt-image-2.5-sunburst.toml      | 18 ++++++++++++++++++
 .../azure/models/gpt-image-2.5-flare.toml      |  6 ++++++
 .../azure/models/gpt-image-2.5-sunburst.toml   |  6 ++++++
 .../models/openai/gpt-image-2.5-flare.toml     | 14 +-------------
 .../models/openai/gpt-image-2.5-sunburst.toml  | 14 +-------------
 6 files changed, 50 insertions(+), 26 deletions(-)
 create mode 100644 models/openai/gpt-image-2.5-flare.toml
 create mode 100644 models/openai/gpt-image-2.5-sunburst.toml
 create mode 100644 providers/azure/models/gpt-image-2.5-flare.toml
 create mode 100644 providers/azure/models/gpt-image-2.5-sunburst.toml

diff --git a/models/openai/gpt-image-2.5-flare.toml b/models/openai/gpt-image-2.5-flare.toml
new file mode 100644
index 00000000000..e74eb10baa8
--- /dev/null
+++ b/models/openai/gpt-image-2.5-flare.toml
@@ -0,0 +1,18 @@
+name = "GPT Image 2.5 Flare"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "gpt-image"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/models/openai/gpt-image-2.5-sunburst.toml b/models/openai/gpt-image-2.5-sunburst.toml
new file mode 100644
index 00000000000..871d743973f
--- /dev/null
+++ b/models/openai/gpt-image-2.5-sunburst.toml
@@ -0,0 +1,18 @@
+name = "GPT Image 2.5 Sunburst"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "gpt-image"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/providers/azure/models/gpt-image-2.5-flare.toml b/providers/azure/models/gpt-image-2.5-flare.toml
new file mode 100644
index 00000000000..9b010ac6446
--- /dev/null
+++ b/providers/azure/models/gpt-image-2.5-flare.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-image-2.5-flare"
+
+[cost]
+input = 5.00
+output = 30.00
+cache_read = 1.25
diff --git a/providers/azure/models/gpt-image-2.5-sunburst.toml b/providers/azure/models/gpt-image-2.5-sunburst.toml
new file mode 100644
index 00000000000..ad9ee64e47d
--- /dev/null
+++ b/providers/azure/models/gpt-image-2.5-sunburst.toml
@@ -0,0 +1,6 @@
+base_model = "openai/gpt-image-2.5-sunburst"
+
+[cost]
+input = 5.00
+output = 30.00
+cache_read = 1.25
diff --git a/providers/vercel/models/openai/gpt-image-2.5-flare.toml b/providers/vercel/models/openai/gpt-image-2.5-flare.toml
index b64fffc4712..139e7e11fd5 100644
--- a/providers/vercel/models/openai/gpt-image-2.5-flare.toml
+++ b/providers/vercel/models/openai/gpt-image-2.5-flare.toml
@@ -1,22 +1,10 @@
-name = "GPT Image 2.5 Flare"
-description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "gpt-image"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
+base_model = "openai/gpt-image-2.5-flare"
 attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
 
 [cost]
 input = 5
 output = 30
 cache_read = 1.25
 
-[limit]
-context = 0
-output = 0
-
 [modalities]
 input = ["text"]
-output = ["image"]
diff --git a/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml b/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml
index 948747ae075..001417a7de5 100644
--- a/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml
+++ b/providers/vercel/models/openai/gpt-image-2.5-sunburst.toml
@@ -1,22 +1,10 @@
-name = "GPT Image 2.5 Sunburst"
-description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "gpt-image"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
+base_model = "openai/gpt-image-2.5-sunburst"
 attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
 
 [cost]
 input = 5
 output = 30
 cache_read = 1.25
 
-[limit]
-context = 0
-output = 0
-
 [modalities]
 input = ["text"]
-output = ["image"]

From b1a35cde4ff5ca327fc8d20d98609b7ca5fdffc3 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:23:27 -0500
Subject: [PATCH 1516/2012] fix: [Alibaba Token Plan | Alibaba Token Plan
 (China)] New model GLM-5.3 available (#7321)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../alibaba-token-plan-cn/models/glm-5.3.toml | 22 +++++++++++++++++++
 .../alibaba-token-plan/models/glm-5.3.toml    | 22 +++++++++++++++++++
 2 files changed, 44 insertions(+)
 create mode 100644 providers/alibaba-token-plan-cn/models/glm-5.3.toml
 create mode 100644 providers/alibaba-token-plan/models/glm-5.3.toml

diff --git a/providers/alibaba-token-plan-cn/models/glm-5.3.toml b/providers/alibaba-token-plan-cn/models/glm-5.3.toml
new file mode 100644
index 00000000000..3640af94643
--- /dev/null
+++ b/providers/alibaba-token-plan-cn/models/glm-5.3.toml
@@ -0,0 +1,22 @@
+# Alibaba Token Plan (China / Beijing) lists model ID glm-5.3 (Zhipu AI).
+# Subscription plan: no per-token USD price (Credits billing).
+# Always-on thinking; effort via reasoning_effort = low|high|max (default max).
+# Same control surface as Bailian glm-5.3 / lab peers; thinking cannot be
+# disabled (no toggle). Interleaved reasoning on reasoning_content.
+# https://docs.bailian.console.aliyun.com/zh/model-studio/token-plan-personal-overview
+# https://docs.bailian.console.aliyun.com/zh/model-studio/token-plan-team-overview
+# https://docs.z.ai/guides/llm/glm-5.3
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
diff --git a/providers/alibaba-token-plan/models/glm-5.3.toml b/providers/alibaba-token-plan/models/glm-5.3.toml
new file mode 100644
index 00000000000..bcde01d6519
--- /dev/null
+++ b/providers/alibaba-token-plan/models/glm-5.3.toml
@@ -0,0 +1,22 @@
+# Alibaba Token Plan (Singapore) lists model ID glm-5.3 (Zhipu AI).
+# Subscription plan: no per-token USD price (Credits billing).
+# Always-on thinking; effort via reasoning_effort = low|high|max (default max).
+# Same control surface as Model Studio glm-5.3 / lab peers; thinking cannot be
+# disabled (no toggle). Interleaved reasoning on reasoning_content.
+# https://docs.modelstudio.console.alibabacloud.com/en/model-studio/token-plan-personal-overview
+# https://docs.modelstudio.console.alibabacloud.com/en/model-studio/token-plan-team-overview
+# https://docs.z.ai/guides/llm/glm-5.3
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From da2dffdbe780259be3e915f95baaa25d261f1257 Mon Sep 17 00:00:00 2001
From: MarioF <79393843+mario03690@users.noreply.github.com>
Date: Thu, 17 Sep 2026 22:27:16 +0800
Subject: [PATCH 1517/2012] Add provider: ainetcafe (Kimi K3) (#7226)

* Add provider: ainetcafe (Kimi K3)

* ainetcafe: currentColor logo, reasoning options

* ainetcafe: effort-only reasoning_options with wire path; drop toggle

* ainetcafe: cite host source for 32K output cap

---------

Co-authored-by: Mario <mario@MariodeMac-mini.local>
---
 providers/ainetcafe/logo.svg            |  1 +
 providers/ainetcafe/models/Kimi-K3.toml | 24 ++++++++++++++++++++++++
 providers/ainetcafe/provider.toml       | 11 +++++++++++
 3 files changed, 36 insertions(+)
 create mode 100644 providers/ainetcafe/logo.svg
 create mode 100644 providers/ainetcafe/models/Kimi-K3.toml
 create mode 100644 providers/ainetcafe/provider.toml

diff --git a/providers/ainetcafe/logo.svg b/providers/ainetcafe/logo.svg
new file mode 100644
index 00000000000..06147492c10
--- /dev/null
+++ b/providers/ainetcafe/logo.svg
@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor"><path d="M14.25 1.3 4.1 14.05h6.95L9.75 22.7 19.9 9.95h-6.95Z"/></svg>
diff --git a/providers/ainetcafe/models/Kimi-K3.toml b/providers/ainetcafe/models/Kimi-K3.toml
new file mode 100644
index 00000000000..1e5d11c4350
--- /dev/null
+++ b/providers/ainetcafe/models/Kimi-K3.toml
@@ -0,0 +1,24 @@
+# Effort: reasoning_effort = low|high|max (top-level OpenAI Chat field, passed through as-is).
+# No toggle: thinking is always on at this gateway; thinking.type = "disabled" is accepted but ignored.
+# model = "Kimi-K3" (case-sensitive at this gateway).
+# Default context 256K; 1M is enabled per account on request, so the limit here is the default.
+# Max output 32K per the host guides (https://ainetcafe.com/k3/guides/kilo-code.html, cline.html, openhands.html).
+# Pricing: https://ainetcafe.com/k3/compare.html
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.10
+output = 10.50
+cache_read = 0.30
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/ainetcafe/provider.toml b/providers/ainetcafe/provider.toml
new file mode 100644
index 00000000000..dcce1fc1cae
--- /dev/null
+++ b/providers/ainetcafe/provider.toml
@@ -0,0 +1,11 @@
+# ainetcafe serves Kimi K3 from its own cluster (native MXFP4 weights).
+# OpenAI Chat: POST https://microquickjs.com/v1/chat/completions
+# Anthropic Messages: POST https://microquickjs.com/v1/messages
+# Reasoning is always on for Kimi-K3 (reasoning_content is returned; thinking.type = "disabled"
+# is accepted but ignored). Depth is set with top-level reasoning_effort = low|high|max.
+# Docs: https://ainetcafe.com/k3/guides/
+name = "ainetcafe"
+env = ["AINETCAFE_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://microquickjs.com/v1"
+doc = "https://ainetcafe.com/k3/guides/"

From ae6a14bf66c13bf001de2073fd3b5eff6cec5a16 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 14:29:06 +0000
Subject: [PATCH 1518/2012] chore(sync): update Kilo model catalog (#7323)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 1 -
 .../kilo/models/~deepseek/deepseek-flash-latest.toml      | 4 ++--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..cf89931c6cd 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
 output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 673b730eb39..f491c2853c8 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_000_000
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 957bc0a82fa..bd4537b7471 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
-context = 1_048_576
+context = 1_000_000
 output = 393_216
 
 [modalities]

From 90e4c4c1293b25da3202bc98c025c05056400c5e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 14:29:25 +0000
Subject: [PATCH 1519/2012] chore(sync): update Fireworks AI model catalog
 (#7324)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../routers/deepseek-v4-flash-0731-us.toml    | 30 --------------
 .../fireworks/routers/glm-5p2-fast-us.toml    | 33 ---------------
 .../fireworks/routers/glm-5p2-us.toml         | 34 ---------------
 .../fireworks/routers/glm-5p3-flash-us.toml   | 32 ---------------
 .../fireworks/routers/glm-5p3-us.toml         | 30 --------------
 .../fireworks/routers/kimi-k3-us.toml         | 41 -------------------
 6 files changed, 200 deletions(-)
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
 delete mode 100644 providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml

diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
deleted file mode 100644
index 846c51d942e..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/deepseek-v4-flash-0731-us.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
-# global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
-# Base DeepSeek V4 Flash (0731) Standard: $0.22 / $0.007 / $0.66 → US
-# $0.33 / $0.0105 / $0.99 (USD/MTok, input / cached input / output).
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = low|high|max
-# https://docs.fireworks.ai/guides/reasoning
-base_model = "deepseek/deepseek-v4-flash-0731"
-name = "DeepSeek V4 Flash (0731) US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 0.33
-output = 0.99
-cache_read = 0.0105
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
deleted file mode 100644
index edbace0f24a..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-fast-us.toml
+++ /dev/null
@@ -1,33 +0,0 @@
-# US-only Serverless Fast router. Must be called via us.api.fireworks.ai (not
-# the global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# GLM 5.2 Fast US is an explicit exception to the 1.5x US premium and matches
-# global GLM 5.2 Fast: $2.10 / $0.21 / $6.60 (USD/MTok, input / cached input /
-# output). https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = high|max
-# https://docs.fireworks.ai/guides/reasoning
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2 Fast US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
-
-[cost]
-input = 2.1
-output = 6.6
-cache_read = 0.21
-
-[limit]
-context = 1_048_575
-output = 131_072
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
deleted file mode 100644
index 2e6423c0de4..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p2-us.toml
+++ /dev/null
@@ -1,34 +0,0 @@
-# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
-# global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
-# Base GLM 5.2 Standard: $1.40 / $0.14 / $4.40 → US $2.10 / $0.21 / $6.60
-# (USD/MTok, input / cached input / output).
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = high|max
-# https://docs.fireworks.ai/guides/reasoning
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2 US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
-
-[cost]
-input = 2.1
-output = 6.6
-cache_read = 0.21
-
-[limit]
-context = 1_048_575
-output = 131_072
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
deleted file mode 100644
index aa9054e1cf0..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-flash-us.toml
+++ /dev/null
@@ -1,32 +0,0 @@
-# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
-# global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
-# Base GLM 5.3 Flash Standard: $0.15 / $0.03 / $0.50 → US $0.225 / $0.045 /
-# $0.75 (USD/MTok, input / cached input / output).
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Thinking-only on Fireworks (no toggle): reasoning_effort = low|high|max.
-# Limits/modalities aligned with global fireworks glm-5p3-flash.
-base_model = "zhipuai/glm-5.3-flash"
-name = "GLM 5.3 Flash US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 0.225
-output = 0.75
-cache_read = 0.045
-
-[limit]
-context = 1_048_573
-
-[modalities]
-input = ["text", "image"]
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
deleted file mode 100644
index 3fd7ef4be2d..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/glm-5p3-us.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
-# global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# Beginning 2026-09-01, US-only models are 1.5x base serverless prices.
-# Base GLM 5.3 Standard: $1.40 / $0.26 / $4.40 → US $2.10 / $0.39 / $6.60
-# (USD/MTok, input / cached input / output).
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Thinking-only on Fireworks (no toggle): reasoning_effort = low|high|max.
-# Limits aligned with global fireworks glm-5p3 serving probes.
-base_model = "zhipuai/glm-5.3"
-name = "GLM 5.3 US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 2.1
-output = 6.6
-cache_read = 0.39
-
-[limit]
-context = 1_048_573
-output = 262_144
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
diff --git a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml b/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml
deleted file mode 100644
index f1970262aad..00000000000
--- a/providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k3-us.toml
+++ /dev/null
@@ -1,41 +0,0 @@
-# US-only Serverless router. Must be called via us.api.fireworks.ai (not the
-# global api.fireworks.ai endpoint).
-# Model ID + endpoint: https://docs.fireworks.ai/serverless/us-only-serverless
-# Standard + Priority pricing (USD/MTok, input / cached input / output):
-# $4.50 / $0.45 / $22.50 and $5.625 / $0.5625 / $28.125
-# https://docs.fireworks.ai/serverless/pricing (accessed 2026-09-17)
-# Toggle: thinking.type = enabled|disabled
-# Effort: reasoning_effort = low|medium|high|max
-# Budget: thinking.budget_tokens (N >= 1024; conflicts with reasoning_effort)
-# https://docs.fireworks.ai/guides/reasoning
-base_model = "moonshotai/kimi-k3"
-name = "Kimi K3 US"
-
-[interleaved]
-field = "reasoning_content"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "max"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-
-[cost]
-input = 4.5
-output = 22.5
-cache_read = 0.45
-
-[modalities]
-input = ["text", "image"]
-
-[provider]
-api = "https://us.api.fireworks.ai/inference/v1/"
-
-[experimental.modes.priority]
-cost = { input = 5.625, output = 28.125, cache_read = 0.5625 }
-provider = { body = { service_tier = "priority" } }

From f2b3a6b9f550ec298a175d33c1a7514bd9b49c87 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 14:29:34 +0000
Subject: [PATCH 1520/2012] chore(sync): update OpenRouter model catalog
 (#7322)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..5b3d39c7244 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index dc34ba0ccf7..98fd23a037b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 50d7ce57993..e0b3ef90d2e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
 context = 1_048_576

From edf46481d65dfae6efdf5c7e802435aa8edde085 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:34:57 -0500
Subject: [PATCH 1521/2012] fix: [missing-model] ofox: bailian/qwen3.8-flash
 (#5622)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/bailian/qwen3.8-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/ofox/models/bailian/qwen3.8-flash.toml

diff --git a/providers/ofox/models/bailian/qwen3.8-flash.toml b/providers/ofox/models/bailian/qwen3.8-flash.toml
new file mode 100644
index 00000000000..f8768c1fb51
--- /dev/null
+++ b/providers/ofox/models/bailian/qwen3.8-flash.toml
@@ -0,0 +1,14 @@
+# Sources (accessed 2026-08-27):
+# https://ofox.ai/en/models/bailian/qwen3.8-flash (pricing, context 1M/131K, capabilities)
+# https://www.qwencloud.com/models/qwen3.8-flash (lab pricing/limits; enable_thinking example)
+# https://docs.qwencloud.com/developer-guides/text-generation/thinking
+# Toggle: enable_thinking true|false (Bailian/DashScope chat path via Ofox; same surface as ofox bailian/qwen3.5-flash, qwen3.6-flash)
+# Budget: thinking_budget
+base_model = "alibaba/qwen3.8-flash"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2

From 858998e89602121d0567dc18c697eae4d5820193 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 17 Sep 2026 09:52:44 -0500
Subject: [PATCH 1522/2012] fix(sync): tolerate new Vercel model types
 (evaluation) (#7325)

* fix(sync): tolerate new Vercel model types (evaluation)

Vercel added type=evaluation (typesafe-ai/jev) without notice, which broke
the sync with a ZodError on the strict ModelType enum.

- Add evaluation to the known model types
- Widen the schema to KnownModelType.or(z.string()) so future unknown
  types fall through to the default text/text handling instead of
  failing the whole sync
- Add regression test covering evaluation + unknown types

* fix(sync): use word-boundary family matching in Vercel sync

The naive substring + subsequence fallback stamped bogus families:
single-letter "o" matched anything containing that letter (26 files:
cohere rerank, fish-audio, alibaba wan, openai tts, ...), and the
subsequence fallback matched scattered letters ("typesafe-ai/jev"
-> "yi").

Align with the other gateway syncs (OpenRouter et al.): word-boundary
regex, with "o" requiring a following digit (o1/o3-style IDs).
Existing bogus "o" stamps self-heal on the next sync.

Also tracks the new typesafe-ai/jev evaluation model (tool_call=false)
with a curated description, since the generated chat-model fallback
does not fit an evaluation model.

---------

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 packages/core/src/sync/providers/vercel.ts    | 47 +++++++----
 packages/core/test/sync.test.ts               | 77 +++++++++++++++++++
 .../models/alibaba/wan-v2.5-t2v-preview.toml  |  1 -
 .../models/alibaba/wan-v2.6-i2v-flash.toml    |  1 -
 .../vercel/models/alibaba/wan-v2.6-i2v.toml   |  1 -
 .../models/alibaba/wan-v2.6-r2v-flash.toml    |  1 -
 .../vercel/models/alibaba/wan-v2.6-r2v.toml   |  1 -
 .../vercel/models/alibaba/wan-v2.6-t2v.toml   |  1 -
 .../vercel/models/alibaba/wan-v2.7-r2v.toml   |  1 -
 .../vercel/models/alibaba/wan-v2.7-t2v.toml   |  1 -
 .../models/alibaba/wan-v3.0-video-prime.toml  |  1 -
 .../vercel/models/alibaba/wan-v3.0-video.toml |  1 -
 .../vercel/models/cohere/rerank-v3.5.toml     |  1 -
 .../vercel/models/cohere/rerank-v4-fast.toml  |  1 -
 .../vercel/models/cohere/rerank-v4-pro.toml   |  1 -
 .../vercel/models/fish-audio/s1-free.toml     |  1 -
 providers/vercel/models/fish-audio/s1.toml    |  1 -
 .../vercel/models/fish-audio/s2-pro-free.toml |  1 -
 .../vercel/models/fish-audio/s2-pro.toml      |  1 -
 .../models/fish-audio/s2.1-pro-free.toml      |  1 -
 .../vercel/models/fish-audio/s2.1-pro.toml    |  1 -
 .../models/fish-audio/transcribe-1-free.toml  |  1 -
 .../models/fish-audio/transcribe-1.toml       |  1 -
 .../inference-net/schematron-v2-small.toml    |  1 -
 .../inference-net/schematron-v2-turbo.toml    |  1 -
 providers/vercel/models/openai/tts-1-hd.toml  |  1 -
 providers/vercel/models/openai/tts-1.toml     |  1 -
 .../vercel/models/quiverai/arrow-1.1.toml     |  1 -
 providers/vercel/models/typesafe-ai/jev.toml  | 20 +++++
 29 files changed, 128 insertions(+), 42 deletions(-)
 create mode 100644 providers/vercel/models/typesafe-ai/jev.toml

diff --git a/packages/core/src/sync/providers/vercel.ts b/packages/core/src/sync/providers/vercel.ts
index 714f0e9bf4a..fdf8be631dd 100644
--- a/packages/core/src/sync/providers/vercel.ts
+++ b/packages/core/src/sync/providers/vercel.ts
@@ -7,7 +7,7 @@ import { factorBaseModel, resolveCanonicalBaseModel } from "./openrouter.js";
 
 const API_ENDPOINT = "https://ai-gateway.vercel.sh/v1/models";
 
-const ModelType = z.enum([
+const KnownModelType = z.enum([
   "language",
   "embedding",
   "image",
@@ -16,6 +16,7 @@ const ModelType = z.enum([
   "transcription",
   "speech",
   "realtime",
+  "evaluation",
 ]);
 
 const PricingTier = z.object({
@@ -42,7 +43,11 @@ export const VercelModel = z.object({
   released: z.number().optional(),
   context_window: z.number().optional().default(0),
   max_tokens: z.number().optional().default(0),
-  type: ModelType,
+  // Vercel adds new model types without notice ("evaluation" appeared Sep 2026
+  // and broke the sync with a ZodError). The trailing z.string() keeps the
+  // schema forward-compatible so future types fall through to the default
+  // text/text handling in buildVercelModel instead of failing the whole sync.
+  type: KnownModelType.or(z.string()),
   tags: z.array(z.string()).optional().default([]),
   pricing: Pricing.optional(),
 }).passthrough();
@@ -107,12 +112,19 @@ export function buildVercelModel(
     : undefined;
   const cost = buildCost(model.pricing, existing?.cost);
 
+  // Self-heal bogus `family = "o"` stamps left by the old substring matcher
+  // (e.g. cohere rerank, fish-audio, alibaba wan). Same precedent as OpenRouter.
+  const inferredFamily = inferFamily(model.id, model.name);
+  const family = existing?.family === "o" && inferredFamily !== "o"
+    ? inferredFamily
+    : (existing?.family ?? inferredFamily);
+
   const synced: SyncedFullModel = {
     name: existing?.name ?? model.name,
     description: existing?.description ?? describeModel({
       id: model.id,
       name: existing?.name ?? model.name,
-      family: existing?.family ?? inferFamily(model.id, model.name),
+      family,
       reasoning: existing?.reasoning ?? tags.has("reasoning"),
       tool_call: model.type === "language"
         ? existing?.tool_call ?? tags.has("tool-use")
@@ -140,7 +152,7 @@ export function buildVercelModel(
           : ["text"],
       },
     }),
-    family: existing?.family ?? inferFamily(model.id, model.name),
+    family,
     release_date: releaseDate,
     last_updated: existing?.last_updated ?? releaseDate,
     attachment: existing?.attachment ?? (tags.has("vision") || tags.has("file-input")),
@@ -243,18 +255,21 @@ function inferFamily(modelID: string, name: string) {
   const kimiFamily = inferKimiFamily(modelID, name);
   if (kimiFamily !== undefined) return kimiFamily;
 
-  const targets = [modelID, name].map((value) => value.toLowerCase());
-  const families = [...ModelFamilyValues].sort((a, b) => b.length - a.length);
-  return families.find((family) => targets.some((target) => target.includes(family.toLowerCase())))
-    ?? families.find((family) => targets.some((target) => isSubsequence(target, family.toLowerCase())));
-}
-
-function isSubsequence(target: string, value: string) {
-  let index = 0;
-  for (const character of target) {
-    if (character === value[index]) index++;
-  }
-  return index === value.length;
+  // Word-boundary matching like the other gateway syncs. Deliberately no
+  // fuzzy/subsequence fallback: matching a family by scattered letters
+  // produces false positives (e.g. "typesafe-ai/jev" -> "yi"), and plain
+  // substring matching lets single-letter families like "o" match anything
+  // containing that letter (e.g. cohere rerank, fish-audio, alibaba wan).
+  const target = `${modelID} ${name}`.toLowerCase();
+  return [...ModelFamilyValues]
+    .sort((a, b) => b.length - a.length)
+    .find((family) => {
+      const value = family.toLowerCase().replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+      if (family === "o") {
+        return new RegExp(`(^|[^a-z0-9])${value}(?=\\d|$|[^a-z0-9])`).test(target);
+      }
+      return new RegExp(`(^|[^a-z0-9])${value}(?=$|[^a-z0-9])`).test(target);
+    });
 }
 
 function sameVercelModel(current: ExistingModel, desired: SyncedModel) {
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 0e51e85cfe1..3557ba66127 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -4540,6 +4540,83 @@ test("Vercel Claude Opus fast variants factor onto base opus metadata", () => {
   expect(synced).not.toHaveProperty("family");
 });
 
+test("Vercel sync accepts evaluation and unknown future model types", () => {
+  const [evaluation, future] = vercel.parseModels({
+    data: [
+      {
+        id: "typesafe-ai/jev",
+        name: "Jev",
+        created: 1_755_815_280,
+        released: 1_789_430_400,
+        context_window: 0,
+        max_tokens: 0,
+        type: "evaluation",
+        pricing: { input: "0.000000042", output: "0" },
+      },
+      {
+        id: "example/future-model",
+        name: "Future Model",
+        created: 1_755_815_280,
+        context_window: 8_000,
+        max_tokens: 4_000,
+        type: "something-new",
+      },
+    ],
+  });
+
+  expect(evaluation).toBeDefined();
+  expect(future).toBeDefined();
+  expect(buildVercelModel(evaluation!, undefined)).toMatchObject({
+    cost: { input: 0.042, output: 0 },
+    limit: { context: 0, output: 0 },
+    modalities: { input: ["text"], output: ["text"] },
+  });
+  expect(buildVercelModel(future!, undefined)).toMatchObject({
+    limit: { context: 8_000, output: 4_000 },
+    modalities: { input: ["text"], output: ["text"] },
+  });
+});
+
+test("Vercel family inference requires word boundaries", () => {
+  const [jev, rerank, o3] = vercel.parseModels({
+    data: [
+      {
+        id: "typesafe-ai/jev",
+        name: "Jev",
+        created: 1_755_815_280,
+        context_window: 0,
+        max_tokens: 0,
+        type: "evaluation",
+      },
+      {
+        id: "cohere/rerank-v3.5",
+        name: "Cohere Rerank 3.5",
+        created: 1_733_000_000,
+        context_window: 4_096,
+        max_tokens: 4_096,
+        type: "reranking",
+      },
+      {
+        id: "example/o3",
+        name: "o3",
+        created: 1_745_000_000,
+        context_window: 200_000,
+        max_tokens: 100_000,
+        type: "language",
+      },
+    ],
+  });
+
+  // No fuzzy subsequence matches ("yi") or single-letter substring matches ("o").
+  expect(buildVercelModel(jev!, undefined).family).toBeUndefined();
+  expect(buildVercelModel(rerank!, undefined).family).toBeUndefined();
+  // Genuine o-series IDs still match, and keep their stamp when re-synced.
+  expect(buildVercelModel(o3!, undefined).family).toBe("o");
+  expect(buildVercelModel(o3!, { family: "o" }).family).toBe("o");
+  // Existing bogus "o" stamps self-heal on the next sync.
+  expect(buildVercelModel(rerank!, { family: "o" }).family).toBeUndefined();
+});
+
 test("Vercel empty existing reasoning_options falls back to the route base menu", () => {
   const [model] = vercel.parseModels({
     data: [{
diff --git a/providers/vercel/models/alibaba/wan-v2.5-t2v-preview.toml b/providers/vercel/models/alibaba/wan-v2.5-t2v-preview.toml
index 7fca15dd7c7..ac51d13a04c 100644
--- a/providers/vercel/models/alibaba/wan-v2.5-t2v-preview.toml
+++ b/providers/vercel/models/alibaba/wan-v2.5-t2v-preview.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.5 Text-to-Video Preview"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2025-09-24"
 last_updated = "2025-09-24"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.6-i2v-flash.toml b/providers/vercel/models/alibaba/wan-v2.6-i2v-flash.toml
index 3ed6130c721..d69ba8b46d4 100644
--- a/providers/vercel/models/alibaba/wan-v2.6-i2v-flash.toml
+++ b/providers/vercel/models/alibaba/wan-v2.6-i2v-flash.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.6 Image-to-Video Flash"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "o"
 release_date = "2025-12-16"
 last_updated = "2025-12-16"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.6-i2v.toml b/providers/vercel/models/alibaba/wan-v2.6-i2v.toml
index 364138ab961..d1df36cfe39 100644
--- a/providers/vercel/models/alibaba/wan-v2.6-i2v.toml
+++ b/providers/vercel/models/alibaba/wan-v2.6-i2v.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.6 Image-to-Video"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "o"
 release_date = "2025-12-16"
 last_updated = "2025-12-16"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.6-r2v-flash.toml b/providers/vercel/models/alibaba/wan-v2.6-r2v-flash.toml
index a6dc42b25dc..f955c1c452e 100644
--- a/providers/vercel/models/alibaba/wan-v2.6-r2v-flash.toml
+++ b/providers/vercel/models/alibaba/wan-v2.6-r2v-flash.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.6 Reference-to-Video Flash"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2025-12-16"
 last_updated = "2025-12-16"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.6-r2v.toml b/providers/vercel/models/alibaba/wan-v2.6-r2v.toml
index d359c4e4a33..6b5e8edb335 100644
--- a/providers/vercel/models/alibaba/wan-v2.6-r2v.toml
+++ b/providers/vercel/models/alibaba/wan-v2.6-r2v.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.6 Reference-to-Video"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2025-12-16"
 last_updated = "2025-12-16"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.6-t2v.toml b/providers/vercel/models/alibaba/wan-v2.6-t2v.toml
index b6f8c18d623..089929b2dd1 100644
--- a/providers/vercel/models/alibaba/wan-v2.6-t2v.toml
+++ b/providers/vercel/models/alibaba/wan-v2.6-t2v.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.6 Text-to-Video"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2025-12-16"
 last_updated = "2025-12-16"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.7-r2v.toml b/providers/vercel/models/alibaba/wan-v2.7-r2v.toml
index f56f24481a1..bcb46d23d3e 100644
--- a/providers/vercel/models/alibaba/wan-v2.7-r2v.toml
+++ b/providers/vercel/models/alibaba/wan-v2.7-r2v.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.7 Reference-to-Video"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2026-04-07"
 last_updated = "2026-04-07"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v2.7-t2v.toml b/providers/vercel/models/alibaba/wan-v2.7-t2v.toml
index 417a61dd8d4..98d5e074e30 100644
--- a/providers/vercel/models/alibaba/wan-v2.7-t2v.toml
+++ b/providers/vercel/models/alibaba/wan-v2.7-t2v.toml
@@ -1,6 +1,5 @@
 name = "Wan v2.7 Text-to-Video"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2026-04-07"
 last_updated = "2026-04-07"
 attachment = false
diff --git a/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml b/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml
index c61b39ecb77..869095910ba 100644
--- a/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml
+++ b/providers/vercel/models/alibaba/wan-v3.0-video-prime.toml
@@ -1,6 +1,5 @@
 name = "Wan v3.0 Video Prime"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2026-08-28"
 last_updated = "2026-08-28"
 attachment = true
diff --git a/providers/vercel/models/alibaba/wan-v3.0-video.toml b/providers/vercel/models/alibaba/wan-v3.0-video.toml
index 594d5d4a4c2..41c7bed34bb 100644
--- a/providers/vercel/models/alibaba/wan-v3.0-video.toml
+++ b/providers/vercel/models/alibaba/wan-v3.0-video.toml
@@ -1,6 +1,5 @@
 name = "Wan v3.0 Video"
 description = "Video model for prompt-guided generation, editing, and motion workflows"
-family = "o"
 release_date = "2026-08-23"
 last_updated = "2026-08-23"
 attachment = true
diff --git a/providers/vercel/models/cohere/rerank-v3.5.toml b/providers/vercel/models/cohere/rerank-v3.5.toml
index 333260d3e7b..d043fba35e3 100644
--- a/providers/vercel/models/cohere/rerank-v3.5.toml
+++ b/providers/vercel/models/cohere/rerank-v3.5.toml
@@ -1,6 +1,5 @@
 name = "Cohere Rerank 3.5"
 description = "Reranking model for improving retrieval quality in search and recommendation systems"
-family = "o"
 release_date = "2024-12-02"
 last_updated = "2024-12-02"
 attachment = false
diff --git a/providers/vercel/models/cohere/rerank-v4-fast.toml b/providers/vercel/models/cohere/rerank-v4-fast.toml
index d2768d8e5da..6dfc23fb7d9 100644
--- a/providers/vercel/models/cohere/rerank-v4-fast.toml
+++ b/providers/vercel/models/cohere/rerank-v4-fast.toml
@@ -1,6 +1,5 @@
 name = "Cohere Rerank 4 Fast"
 description = "Reranking model for improving retrieval quality in search and recommendation systems"
-family = "o"
 release_date = "2025-12-11"
 last_updated = "2025-12-11"
 attachment = false
diff --git a/providers/vercel/models/cohere/rerank-v4-pro.toml b/providers/vercel/models/cohere/rerank-v4-pro.toml
index e7e41c0c6b6..e2e994e8fa2 100644
--- a/providers/vercel/models/cohere/rerank-v4-pro.toml
+++ b/providers/vercel/models/cohere/rerank-v4-pro.toml
@@ -1,6 +1,5 @@
 name = "Cohere Rerank 4 Pro"
 description = "Reranking model for improving retrieval quality in search and recommendation systems"
-family = "o"
 release_date = "2025-12-11"
 last_updated = "2025-12-11"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s1-free.toml b/providers/vercel/models/fish-audio/s1-free.toml
index 031d08ec2c0..2eb0f456def 100644
--- a/providers/vercel/models/fish-audio/s1-free.toml
+++ b/providers/vercel/models/fish-audio/s1-free.toml
@@ -1,6 +1,5 @@
 name = "S1 (Free)"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2025-10-20"
 last_updated = "2025-10-20"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s1.toml b/providers/vercel/models/fish-audio/s1.toml
index 3b68c1120b6..775ab2b196e 100644
--- a/providers/vercel/models/fish-audio/s1.toml
+++ b/providers/vercel/models/fish-audio/s1.toml
@@ -1,6 +1,5 @@
 name = "S1"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2025-10-20"
 last_updated = "2025-10-20"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s2-pro-free.toml b/providers/vercel/models/fish-audio/s2-pro-free.toml
index 2a65d6f5b70..c2bbbb724c8 100644
--- a/providers/vercel/models/fish-audio/s2-pro-free.toml
+++ b/providers/vercel/models/fish-audio/s2-pro-free.toml
@@ -1,6 +1,5 @@
 name = "S2 Pro (Free)"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2026-03-09"
 last_updated = "2026-03-09"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s2-pro.toml b/providers/vercel/models/fish-audio/s2-pro.toml
index a60fb8fe76b..b154766892a 100644
--- a/providers/vercel/models/fish-audio/s2-pro.toml
+++ b/providers/vercel/models/fish-audio/s2-pro.toml
@@ -1,6 +1,5 @@
 name = "S2 Pro"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2026-03-09"
 last_updated = "2026-03-09"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s2.1-pro-free.toml b/providers/vercel/models/fish-audio/s2.1-pro-free.toml
index d1c95dc0d10..c580e92cd34 100644
--- a/providers/vercel/models/fish-audio/s2.1-pro-free.toml
+++ b/providers/vercel/models/fish-audio/s2.1-pro-free.toml
@@ -1,6 +1,5 @@
 name = "S2.1 Pro (Free)"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2026-07-28"
 last_updated = "2026-07-28"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/s2.1-pro.toml b/providers/vercel/models/fish-audio/s2.1-pro.toml
index b28b42f7465..de3b45e925d 100644
--- a/providers/vercel/models/fish-audio/s2.1-pro.toml
+++ b/providers/vercel/models/fish-audio/s2.1-pro.toml
@@ -1,6 +1,5 @@
 name = "S2.1 Pro"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2026-07-28"
 last_updated = "2026-07-28"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/transcribe-1-free.toml b/providers/vercel/models/fish-audio/transcribe-1-free.toml
index e03d53fa6fa..e21df2298c0 100644
--- a/providers/vercel/models/fish-audio/transcribe-1-free.toml
+++ b/providers/vercel/models/fish-audio/transcribe-1-free.toml
@@ -1,6 +1,5 @@
 name = "Transcribe-1 (Free)"
 description = "Speech transcription model for accurate audio-to-text and captioning workflows"
-family = "o"
 release_date = "2026-03-01"
 last_updated = "2026-03-01"
 attachment = false
diff --git a/providers/vercel/models/fish-audio/transcribe-1.toml b/providers/vercel/models/fish-audio/transcribe-1.toml
index 7020c7432f1..a988499717a 100644
--- a/providers/vercel/models/fish-audio/transcribe-1.toml
+++ b/providers/vercel/models/fish-audio/transcribe-1.toml
@@ -1,6 +1,5 @@
 name = "Transcribe-1"
 description = "Speech transcription model for accurate audio-to-text and captioning workflows"
-family = "o"
 release_date = "2026-03-01"
 last_updated = "2026-03-01"
 attachment = false
diff --git a/providers/vercel/models/inference-net/schematron-v2-small.toml b/providers/vercel/models/inference-net/schematron-v2-small.toml
index 441cd688f11..7a30cc2781e 100644
--- a/providers/vercel/models/inference-net/schematron-v2-small.toml
+++ b/providers/vercel/models/inference-net/schematron-v2-small.toml
@@ -1,6 +1,5 @@
 name = "Schematron V2 Small"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
-family = "o"
 release_date = "2026-04-16"
 last_updated = "2026-04-16"
 attachment = false
diff --git a/providers/vercel/models/inference-net/schematron-v2-turbo.toml b/providers/vercel/models/inference-net/schematron-v2-turbo.toml
index 5ac1c1c9009..ad615547b48 100644
--- a/providers/vercel/models/inference-net/schematron-v2-turbo.toml
+++ b/providers/vercel/models/inference-net/schematron-v2-turbo.toml
@@ -1,6 +1,5 @@
 name = "Schematron V2 Turbo"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
-family = "o"
 release_date = "2026-04-16"
 last_updated = "2026-04-16"
 attachment = false
diff --git a/providers/vercel/models/openai/tts-1-hd.toml b/providers/vercel/models/openai/tts-1-hd.toml
index c18d5c7866e..11a75cdcbca 100644
--- a/providers/vercel/models/openai/tts-1-hd.toml
+++ b/providers/vercel/models/openai/tts-1-hd.toml
@@ -1,6 +1,5 @@
 name = "TTS-1 HD"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2023-11-06"
 last_updated = "2023-11-06"
 attachment = false
diff --git a/providers/vercel/models/openai/tts-1.toml b/providers/vercel/models/openai/tts-1.toml
index 0e0a978b652..9067935181e 100644
--- a/providers/vercel/models/openai/tts-1.toml
+++ b/providers/vercel/models/openai/tts-1.toml
@@ -1,6 +1,5 @@
 name = "TTS-1"
 description = "Speech generation model for controllable voice, narration, and audio delivery"
-family = "o"
 release_date = "2023-11-06"
 last_updated = "2023-11-06"
 attachment = false
diff --git a/providers/vercel/models/quiverai/arrow-1.1.toml b/providers/vercel/models/quiverai/arrow-1.1.toml
index 4221693b8e0..3a02c2c71ed 100644
--- a/providers/vercel/models/quiverai/arrow-1.1.toml
+++ b/providers/vercel/models/quiverai/arrow-1.1.toml
@@ -1,6 +1,5 @@
 name = "Arrow 1.1"
 description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "o"
 release_date = "2026-04-16"
 last_updated = "2026-04-16"
 attachment = false
diff --git a/providers/vercel/models/typesafe-ai/jev.toml b/providers/vercel/models/typesafe-ai/jev.toml
new file mode 100644
index 00000000000..ea15dd157f6
--- /dev/null
+++ b/providers/vercel/models/typesafe-ai/jev.toml
@@ -0,0 +1,20 @@
+name = "Jev"
+description = "Evaluation model for fast, structured decisions in software, returning choices, scores, and boolean probabilities"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.042
+output = 0
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 73ca7787fa0c29c466979b121f6f62ff09ebd53f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 15:26:44 +0000
Subject: [PATCH 1523/2012] chore(sync): update Charm Hyper model catalog
 (#7327)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/gpt-oss-120b.toml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/hyper/models/gpt-oss-120b.toml b/providers/hyper/models/gpt-oss-120b.toml
index 22c22acac3a..bc7e3589b61 100644
--- a/providers/hyper/models/gpt-oss-120b.toml
+++ b/providers/hyper/models/gpt-oss-120b.toml
@@ -4,7 +4,7 @@ last_updated = "2026-07-22"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high"]
 
 [cost]
 input = 0.178
@@ -12,5 +12,4 @@ output = 0.68
 cache_read = 0.089
 
 [limit]
-context = 128_072
 output = 13_107

From dc2dd73e08d847f2e80a8c70d4a6b3cfc7d61f0b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 15:26:49 +0000
Subject: [PATCH 1524/2012] chore(sync): update Kilo model catalog (#7328)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 1 +
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index cf89931c6cd..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
+context = 1_048_576
 output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index bd4537b7471..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 393_216
 
 [modalities]

From 6576f338e897f6e31e0aff5f978e5e23dd8db186 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 15:26:57 +0000
Subject: [PATCH 1525/2012] chore(sync): update Eden AI model catalog (#7330)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index 13d977a7fd0..d5d18a5d5f0 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34611
-output = 0.46148
+input = 0.34443
+output = 0.45924
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 4032a872079..fc91f75adc5 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.749905
-output = 0.749905
+input = 0.746265
+output = 0.746265
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 2d3e6ad6c1a..488fd48dab9 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173055
-output = 0.749905
+input = 0.172215
+output = 0.746265
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 9b9acde4f4d..2236e082a15 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.46148
-output = 0.92296
+input = 0.45924
+output = 0.91848
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 6e9e91ebf7e..c78acf705d7 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.173055
-output = 0.69222
+input = 0.172215
+output = 0.68886
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 6baa7372a71..b6133a15c15 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.038331
-output = 1.038331
+input = 1.03329
+output = 1.03329

From cdf40d2fbf88f6730df802fcd46d48797b8701ac Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 15:27:19 +0000
Subject: [PATCH 1526/2012] chore(sync): update OpenRouter model catalog
 (#7329)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 5b3d39c7244..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e0b3ef90d2e..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576

From 8342b837669d74c555d60ad97beb21418e2fb2c2 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Thu, 17 Sep 2026 10:28:30 -0500
Subject: [PATCH 1527/2012] fix(core): use structured Cloudflare pricing
 (#7326)

* fix(core): parse Cloudflare context pricing

* chore(sync): update Cloudflare AI Gateway catalog

* fix(cloudflare-ai-gateway): verify Astra effort controls
---
 .../sync/providers/cloudflare-ai-gateway.ts   | 159 ++++++++++++++----
 .../core/test/cloudflare-ai-gateway.test.ts   | 102 ++++++++---
 providers/cloudflare-ai-gateway/curation.toml |   9 +
 .../models/openai/gpt-6-astra.toml            |  25 +++
 .../models/xai/grok-4.5.toml                  |   6 +
 .../models/xai/grok-4.6.toml                  |   6 +
 6 files changed, 251 insertions(+), 56 deletions(-)
 create mode 100644 providers/cloudflare-ai-gateway/models/openai/gpt-6-astra.toml

diff --git a/packages/core/src/sync/providers/cloudflare-ai-gateway.ts b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
index 775c8524f36..5e36c029612 100644
--- a/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
+++ b/packages/core/src/sync/providers/cloudflare-ai-gateway.ts
@@ -30,18 +30,61 @@ const NATIVE_NPM: Record<string, string> = {
   openai: "@ai-sdk/openai",
 };
 
+const Price = z.number().nonnegative();
+const ProviderDetail = z.object({
+  id: z.string().min(1),
+  pricing: z.record(Price),
+}).passthrough();
+
 const CatalogEntry = z.object({
   model_id: z.string().refine(isSafeModelID, "model_id must be a safe relative provider/model path"),
   task: z.string(),
   context_length: z.number().int().positive().nullish(),
-  pricing: z.record(z.number().nonnegative()).nullish(),
+  provider_details: z.array(ProviderDetail).nullish(),
 }).passthrough();
 const CatalogModel = CatalogEntry.extend({
   task: z.literal(TEXT_GENERATION),
   context_length: z.number().int().positive().nullish(),
-  pricing: z.record(z.number().nonnegative()),
+  provider_details: z.array(ProviderDetail).length(1),
 });
 
+const FlatPricing = z.object({
+  input_tokens: Price,
+  output_tokens: Price,
+  input_cached_tokens: Price.optional(),
+  input_cache_creation_tokens: Price.optional(),
+}).strict();
+
+const ContextPricing = z.object({
+  long_context_threshold_tokens: Price.int().positive(),
+  short_context_input_tokens: Price,
+  short_context_output_tokens: Price,
+  short_context_input_cached_tokens: Price.optional(),
+  short_context_input_cache_creation_tokens: Price.optional(),
+  long_context_input_tokens: Price,
+  long_context_output_tokens: Price,
+  long_context_input_cached_tokens: Price.optional(),
+  long_context_input_cache_creation_tokens: Price.optional(),
+}).strict();
+
+const Tier200kPricing = z.object({
+  under_200k_per_1m_input_tokens: Price,
+  under_200k_per_1m_output_tokens: Price,
+  under_200k_per_1m_input_cached_tokens: Price.optional(),
+  over_200k_per_1m_input_tokens: Price,
+  over_200k_per_1m_output_tokens: Price,
+  over_200k_per_1m_input_cached_tokens: Price.optional(),
+}).strict();
+
+const Tier512kPricing = z.object({
+  under_512k_per_1m_input_tokens: Price,
+  under_512k_per_1m_output_tokens: Price,
+  under_512k_per_1m_input_cached_tokens: Price.optional(),
+  over_512k_per_1m_input_tokens: Price,
+  over_512k_per_1m_output_tokens: Price,
+  over_512k_per_1m_input_cached_tokens: Price.optional(),
+}).strict();
+
 const CloudflareResponse = z.object({
   success: z.literal(true),
   result: z.array(CatalogEntry),
@@ -150,7 +193,7 @@ export function buildCloudflareAiGatewayModel(
 ): SyncedBaseModel {
   const id = catalog.model_id;
   // Pricing failures must not be hidden by missing reasoning controls.
-  const cost = proxiedCost(catalog.pricing, id);
+  const cost = proxiedCost(catalog.provider_details[0]!.pricing, id);
   const baseModel = curated.base_model ?? resolveBaseModel(id);
   if (baseModel === undefined) {
     throw new Error(`${id}: no lab file and no curated base_model; add it to skip or map it`);
@@ -441,39 +484,91 @@ async function mapLimit<T, R>(items: T[], limit: number, transform: (item: T) =>
   return results;
 }
 
-const FLAT_PRICING_KEYS: Record<string, "input" | "output" | "cache_read" | "cache_write"> = {
-  "Input tokens (per 1M)": "input",
-  "Output tokens (per 1M)": "output",
-  "Cached input tokens (per 1M)": "cache_read",
-  "Cache creation tokens (per 1M)": "cache_write",
-};
-const TIERED_PRICING_KEY = /^(Input|Output|Cached input)\s*(<=?|>=?)\s*(\d+)k\s*\(per 1M\)$/;
-const TIERED_PRICING_FIELDS = {
-  Input: "input",
-  Output: "output",
-  "Cached input": "cache_read",
-} as const;
-
 function proxiedCost(pricing: Record<string, number>, id: string): NonNullable<SyncedBaseModel["cost"]> {
-  const cost: NonNullable<SyncedBaseModel["cost"]> = {};
-  for (const [key, value] of Object.entries(pricing)) {
-    const flatField = FLAT_PRICING_KEYS[key];
-    if (flatField !== undefined) {
-      cost[flatField] = value;
-      continue;
+  try {
+    if ("long_context_threshold_tokens" in pricing) {
+      const value = ContextPricing.parse(pricing);
+      return contextCost(
+        value.long_context_threshold_tokens,
+        tokenCost(
+          value.short_context_input_tokens,
+          value.short_context_output_tokens,
+          value.short_context_input_cached_tokens,
+          value.short_context_input_cache_creation_tokens,
+        ),
+        tokenCost(
+          value.long_context_input_tokens,
+          value.long_context_output_tokens,
+          value.long_context_input_cached_tokens,
+          value.long_context_input_cache_creation_tokens,
+        ),
+      );
     }
-    const tier = TIERED_PRICING_KEY.exec(key);
-    if (tier !== null) {
-      const field = TIERED_PRICING_FIELDS[tier[1] as keyof typeof TIERED_PRICING_FIELDS];
-      if (tier[2]!.startsWith("<")) cost[field] = value;
-      continue;
+    if ("under_200k_per_1m_input_tokens" in pricing) {
+      const value = Tier200kPricing.parse(pricing);
+      return contextCost(
+        200_000,
+        tokenCost(
+          value.under_200k_per_1m_input_tokens,
+          value.under_200k_per_1m_output_tokens,
+          value.under_200k_per_1m_input_cached_tokens,
+        ),
+        tokenCost(
+          value.over_200k_per_1m_input_tokens,
+          value.over_200k_per_1m_output_tokens,
+          value.over_200k_per_1m_input_cached_tokens,
+        ),
+      );
     }
-    throw new Error(`${id}: unmapped pricing key "${key}"`);
-  }
-  if (cost.input === undefined || cost.output === undefined) {
-    throw new Error(`${id}: catalog pricing must include input and output rates`);
+    if ("under_512k_per_1m_input_tokens" in pricing) {
+      const value = Tier512kPricing.parse(pricing);
+      return contextCost(
+        512_000,
+        tokenCost(
+          value.under_512k_per_1m_input_tokens,
+          value.under_512k_per_1m_output_tokens,
+          value.under_512k_per_1m_input_cached_tokens,
+        ),
+        tokenCost(
+          value.over_512k_per_1m_input_tokens,
+          value.over_512k_per_1m_output_tokens,
+          value.over_512k_per_1m_input_cached_tokens,
+        ),
+      );
+    }
+    const value = FlatPricing.parse(pricing);
+    return tokenCost(
+      value.input_tokens,
+      value.output_tokens,
+      value.input_cached_tokens,
+      value.input_cache_creation_tokens,
+    );
+  } catch (error) {
+    if (error instanceof z.ZodError) {
+      throw new Error(`${id}: unsupported structured pricing: ${error.message}`);
+    }
+    throw error;
   }
-  return cost;
+}
+
+function tokenCost(input: number, output: number, cacheRead?: number, cacheWrite?: number) {
+  return {
+    input,
+    output,
+    ...(cacheRead === undefined ? {} : { cache_read: cacheRead }),
+    ...(cacheWrite === undefined ? {} : { cache_write: cacheWrite }),
+  };
+}
+
+function contextCost(
+  size: number,
+  short: ReturnType<typeof tokenCost>,
+  long: ReturnType<typeof tokenCost>,
+): NonNullable<SyncedBaseModel["cost"]> {
+  return {
+    ...short,
+    tiers: [{ ...long, tier: { type: "context", size } }],
+  };
 }
 
 function isSafeModelID(id: string) {
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
index 25db6a6e0fd..dcabdcc7ed5 100644
--- a/packages/core/test/cloudflare-ai-gateway.test.ts
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -13,9 +13,9 @@ import {
 test("missing reasoning controls open issues without deleting existing models or blocking valid ones", async () => {
   const dir = await mkdtemp(path.join(import.meta.dirname, "../../../providers/.reasoning-sync-"));
   const modelsDir = path.join(dir, "models");
-  const ids = ["anthropic/claude-fable-5.1", "anthropic/claude-fable-5-1"];
+  const ids = ["deepreinforce/ornith-1.0-9b", "deepreinforce/ornith-1.0-31b"];
   const file = path.join(modelsDir, `${ids[0]}.toml`);
-  const content = '# Keep authored controls\nbase_model = "anthropic/claude-fable-5-1"\nreasoning_options = [{ type = "effort", values = ["high"] }]\n';
+  const content = '# Keep authored controls\nbase_model = "deepreinforce/ornith-1.0-9b"\nreasoning_options = [{ type = "effort", values = ["high"] }]\n';
   await mkdir(path.dirname(file), { recursive: true });
   await writeFile(file, content);
   const issues = spyOn(missingIssues, "openMissingModelIssues").mockResolvedValue([]);
@@ -23,7 +23,11 @@ test("missing reasoning controls open issues without deleting existing models or
     ...cloudflareAiGateway, modelsDir,
     async fetchModels() {
       return [...ids, "openai/gpt-4.1"].map((model_id) => ({
-        catalog: { model_id, task: "Text Generation", pricing: { "Input tokens (per 1M)": 1, "Output tokens (per 1M)": 2 } },
+        catalog: {
+          model_id,
+          task: "Text Generation",
+          provider_details: providerDetails({ input_tokens: 1, output_tokens: 2 }),
+        },
       }));
     },
   };
@@ -48,12 +52,14 @@ test("builds Cloudflare AI Gateway overrides from catalog metadata", () => {
       model_id: "openai/gpt-5.4",
       task: "Text Generation",
       context_length: 1_050_000,
-      pricing: {
-        "Input <= 200k (per 1M)": 2.5,
-        "Input > 200k (per 1M)": 5,
-        "Output tokens (per 1M)": 15,
-        "Cached input tokens (per 1M)": 0.25,
-      },
+      provider_details: providerDetails({
+        under_200k_per_1m_input_tokens: 2.5,
+        under_200k_per_1m_output_tokens: 15,
+        under_200k_per_1m_input_cached_tokens: 0.25,
+        over_200k_per_1m_input_tokens: 5,
+        over_200k_per_1m_output_tokens: 22.5,
+        over_200k_per_1m_input_cached_tokens: 0.5,
+      }),
     },
     undefined,
     {
@@ -64,12 +70,62 @@ test("builds Cloudflare AI Gateway overrides from catalog metadata", () => {
   expect(model).toEqual({
     base_model: "openai/gpt-5.4",
     reasoning_options: [{ type: "effort", values: ["none", "low", "medium", "high", "xhigh"] }],
-    cost: { input: 2.5, output: 15, cache_read: 0.25 },
+    cost: {
+      input: 2.5,
+      output: 15,
+      cache_read: 0.25,
+      tiers: [{
+        tier: { type: "context", size: 200_000 },
+        input: 5,
+        output: 22.5,
+        cache_read: 0.5,
+      }],
+    },
     limit: { context: 1_050_000 },
     provider: { npm: "@ai-sdk/openai" },
   });
 });
 
+test("maps structured provider pricing instead of display labels", () => {
+  const model = buildCloudflareAiGatewayModel(
+    {
+      model_id: "openai/gpt-6-astra",
+      task: "Text Generation",
+      context_length: 1_050_000,
+      pricing: { "Display label that may change": 999 },
+      provider_details: providerDetails({
+        long_context_threshold_tokens: 272_000,
+        short_context_input_tokens: 10,
+        short_context_input_cached_tokens: 1,
+        short_context_input_cache_creation_tokens: 12,
+        short_context_output_tokens: 50,
+        long_context_input_tokens: 20,
+        long_context_input_cached_tokens: 2,
+        long_context_input_cache_creation_tokens: 25,
+        long_context_output_tokens: 75,
+      }),
+    },
+    undefined,
+    {
+      reasoning_options: [{ type: "effort", values: ["low", "medium", "high"] }],
+    },
+  );
+
+  expect(model.cost).toEqual({
+    input: 10,
+    output: 50,
+    cache_read: 1,
+    cache_write: 12,
+    tiers: [{
+      tier: { type: "context", size: 272_000 },
+      input: 20,
+      output: 75,
+      cache_read: 2,
+      cache_write: 25,
+    }],
+  });
+});
+
 test("derives nested Cloudflare reasoning controls", () => {
   expect(deriveReasoningOptions({
     properties: {
@@ -94,10 +150,7 @@ test("ignores advertised reasoning controls for non-reasoning base models", () =
       model_id: "openai/gpt-4.1",
       task: "Text Generation",
       context_length: 1_047_576,
-      pricing: {
-        "Input tokens (per 1M)": 2,
-        "Output tokens (per 1M)": 8,
-      },
+      provider_details: providerDetails({ input_tokens: 2, output_tokens: 8 }),
     },
     {
       properties: {
@@ -114,12 +167,12 @@ test("fails closed on unknown pricing fields", () => {
     model_id: "openai/gpt-4.1",
     task: "Text Generation",
     context_length: 1_047_576,
-    pricing: {
-      "Input tokens (per 1M)": 2,
-      "Output tokens (per 1M)": 8,
-      "New billing unit": 1,
-    },
-  }, undefined)).toThrow('unmapped pricing key "New billing unit"');
+    provider_details: providerDetails({
+      input_tokens: 2,
+      output_tokens: 8,
+      new_billing_unit: 1,
+    }),
+  }, undefined)).toThrow("unsupported structured pricing");
 });
 
 test("fails closed when Cloudflare pagination is incomplete", async () => {
@@ -230,10 +283,7 @@ test("retries transient Cloudflare responses", async () => {
       model_id: "openai/gpt-4.1",
       task: "Text Generation",
       context_length: 1_047_576,
-      pricing: {
-        "Input tokens (per 1M)": 2,
-        "Output tokens (per 1M)": 8,
-      },
+      provider_details: providerDetails({ input_tokens: 2, output_tokens: 8 }),
     }])));
   };
 
@@ -460,6 +510,10 @@ function catalogPage(
   };
 }
 
+function providerDetails(pricing: Record<string, number>) {
+  return [{ id: "test-provider", pricing }];
+}
+
 function restoreEnv(name: string, value: string | undefined) {
   if (value === undefined) delete process.env[name];
   else process.env[name] = value;
diff --git a/providers/cloudflare-ai-gateway/curation.toml b/providers/cloudflare-ai-gateway/curation.toml
index 916be4ca9a8..40337abd089 100644
--- a/providers/cloudflare-ai-gateway/curation.toml
+++ b/providers/cloudflare-ai-gateway/curation.toml
@@ -34,6 +34,7 @@ skip = [
   "google/gemini-3.6-flash",
   "google/gemini-3.7-flash",
   "google/gemini-3.8-flash",
+  "stealth/union-alpha",
 ]
 
 [models."anthropic/claude-fable-5"]
@@ -130,6 +131,14 @@ reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high
 [models."openai/gpt-5.6-terra"]
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
+# GPT-6 Astra has conflicting validation layers on Cloudflare's Responses route. The catalog
+# schema advertises none|low|medium|high, but `none` reaches OpenAI and is rejected upstream;
+# xhigh|max are rejected by Cloudflare before forwarding. POST /ai/v1/responses verified
+# low end-to-end and those three rejection paths on 2026-09-17.
+[models."openai/gpt-6-astra"]
+note = ["Effort: reasoning.effort = low|medium|high; Cloudflare rejects xhigh|max, and OpenAI rejects none after forwarding (live verified 2026-09-17)."]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
 [models."openai/o3"]
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
diff --git a/providers/cloudflare-ai-gateway/models/openai/gpt-6-astra.toml b/providers/cloudflare-ai-gateway/models/openai/gpt-6-astra.toml
new file mode 100644
index 00000000000..5201b391355
--- /dev/null
+++ b/providers/cloudflare-ai-gateway/models/openai/gpt-6-astra.toml
@@ -0,0 +1,25 @@
+# Effort: reasoning.effort = low|medium|high; Cloudflare rejects xhigh|max, and OpenAI rejects none after forwarding (live verified 2026-09-17).
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[limit]
+context = 1_050_000
+
+[provider]
+npm = "@ai-sdk/openai"
diff --git a/providers/cloudflare-ai-gateway/models/xai/grok-4.5.toml b/providers/cloudflare-ai-gateway/models/xai/grok-4.5.toml
index cc4f3e39b59..94f9069dcc6 100644
--- a/providers/cloudflare-ai-gateway/models/xai/grok-4.5.toml
+++ b/providers/cloudflare-ai-gateway/models/xai/grok-4.5.toml
@@ -9,5 +9,11 @@ input = 2
 output = 6
 cache_read = 0.3
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 0.6
+
 [limit]
 context = 500_000
diff --git a/providers/cloudflare-ai-gateway/models/xai/grok-4.6.toml b/providers/cloudflare-ai-gateway/models/xai/grok-4.6.toml
index bd56d89c029..f5e6683671a 100644
--- a/providers/cloudflare-ai-gateway/models/xai/grok-4.6.toml
+++ b/providers/cloudflare-ai-gateway/models/xai/grok-4.6.toml
@@ -9,5 +9,11 @@ input = 2
 output = 6
 cache_read = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+
 [limit]
 context = 500_000

From 8830b58a133e66eb8d9d2570828d0c41b8f5cecd Mon Sep 17 00:00:00 2001
From: eatmoreduck <yuxianice@163.com>
Date: Thu, 17 Sep 2026 23:30:55 +0800
Subject: [PATCH 1528/2012] fix(providers): drop stale zhipuai-coding-plan
 entries after GLM-5.3 consolidation (#7331)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Zhipu's GLM Coding Plan consolidated to GLM-5.3 and GLM-5.3-Flash
(official announcement: docs.bigmodel.cn/cn/coding-plan/overview,
"可用模型": all plans serve GLM-5.3 and GLM-5.3-Flash only).

First-hand verification against the live coding-plan endpoint
(open.bigmodel.cn/api/coding/paas/v4, current subscription):

| catalog id         | server answers            | action |
|--------------------|---------------------------|--------|
| glm-5.3            | glm-5.3                   | keep   |
| glm-5.3-flash      | glm-5.3-flash             | keep   |
| glm-4.6v           | glm-4.6v (still served!)  | keep   |
| glm-5.2            | glm-5.3                   | drop   |
| glm-5.1            | glm-5.3                   | drop   |
| glm-5-turbo        | glm-5.3-flash             | drop   |
| glm-4.7            | glm-5.3-flash             | drop   |
| glm-4.5-air        | glm-5.3-flash             | drop   |
| glm-5.2-highspeed  | 403 not in plan           | drop   |
| glm-5.3-highspeed  | 403 not in plan           | drop   |
| glm-5v-turbo       | 403 not in plan           | drop   |

Keeping rerouted entries means a user picking e.g. glm-5-turbo talks
to GLM-5.3-Flash while the catalog applies stale context/maxTokens/
cost/reasoning metadata to the session.

Note: glm-4.6v is not mentioned in the announcement but is still
served as-is on coding plans, so it stays — it would have been
wrongly dropped by announcement-only triage.

Mirrors the earlier deepseek consolidation. Refs earendil-works/pi#9616
(pi mirrors this provider; maintainer asked for a volunteer to sync
the catalog upstream). bun validate passes; the assembled provider
catalog now exposes exactly glm-5.3, glm-5.3-flash, glm-4.6v.
---
 .../models/glm-4.5-air.toml                   |  1 -
 .../zhipuai-coding-plan/models/glm-4.7.toml   |  1 -
 .../models/glm-5-turbo.toml                   |  1 -
 .../zhipuai-coding-plan/models/glm-5.1.toml   | 31 -------------------
 .../models/glm-5.2-highspeed.toml             |  1 -
 .../zhipuai-coding-plan/models/glm-5.2.toml   |  1 -
 .../models/glm-5.3-highspeed.toml             | 17 ----------
 .../models/glm-5v-turbo.toml                  | 30 ------------------
 8 files changed, 83 deletions(-)
 delete mode 120000 providers/zhipuai-coding-plan/models/glm-4.5-air.toml
 delete mode 120000 providers/zhipuai-coding-plan/models/glm-4.7.toml
 delete mode 120000 providers/zhipuai-coding-plan/models/glm-5-turbo.toml
 delete mode 100644 providers/zhipuai-coding-plan/models/glm-5.1.toml
 delete mode 120000 providers/zhipuai-coding-plan/models/glm-5.2-highspeed.toml
 delete mode 120000 providers/zhipuai-coding-plan/models/glm-5.2.toml
 delete mode 100644 providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml
 delete mode 100644 providers/zhipuai-coding-plan/models/glm-5v-turbo.toml

diff --git a/providers/zhipuai-coding-plan/models/glm-4.5-air.toml b/providers/zhipuai-coding-plan/models/glm-4.5-air.toml
deleted file mode 120000
index 18c5af5fb47..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-4.5-air.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../zai-coding-plan/models/glm-4.5-air.toml
\ No newline at end of file
diff --git a/providers/zhipuai-coding-plan/models/glm-4.7.toml b/providers/zhipuai-coding-plan/models/glm-4.7.toml
deleted file mode 120000
index aa6a452eb32..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-4.7.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../zai-coding-plan/models/glm-4.7.toml
\ No newline at end of file
diff --git a/providers/zhipuai-coding-plan/models/glm-5-turbo.toml b/providers/zhipuai-coding-plan/models/glm-5-turbo.toml
deleted file mode 120000
index ad28ebce0c4..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5-turbo.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../zai-coding-plan/models/glm-5-turbo.toml
\ No newline at end of file
diff --git a/providers/zhipuai-coding-plan/models/glm-5.1.toml b/providers/zhipuai-coding-plan/models/glm-5.1.toml
deleted file mode 100644
index 8c55069b0a1..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5.1.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-name = "GLM-5.1"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-family = "glm"
-release_date = "2026-03-27"
-last_updated = "2026-03-27"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
-
-[limit]
-context = 200000
-output = 131072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/zhipuai-coding-plan/models/glm-5.2-highspeed.toml b/providers/zhipuai-coding-plan/models/glm-5.2-highspeed.toml
deleted file mode 120000
index a4242b49c96..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5.2-highspeed.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../zai-coding-plan/models/glm-5.2-highspeed.toml
\ No newline at end of file
diff --git a/providers/zhipuai-coding-plan/models/glm-5.2.toml b/providers/zhipuai-coding-plan/models/glm-5.2.toml
deleted file mode 120000
index 539459c44b6..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5.2.toml
+++ /dev/null
@@ -1 +0,0 @@
-../../zai-coding-plan/models/glm-5.2.toml
\ No newline at end of file
diff --git a/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml b/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml
deleted file mode 100644
index 2f36aacb151..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "zhipuai/glm-5.3"
-name = "GLM-5.3 Highspeed"
-
-# Effort levels mirror the base GLM-5.3 (low, high, max).
-# https://docs.bigmodel.cn/cn/coding-plan/latest-model
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
diff --git a/providers/zhipuai-coding-plan/models/glm-5v-turbo.toml b/providers/zhipuai-coding-plan/models/glm-5v-turbo.toml
deleted file mode 100644
index a451c68fd2f..00000000000
--- a/providers/zhipuai-coding-plan/models/glm-5v-turbo.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "GLM-5V-Turbo"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
-release_date = "2026-04-01"
-last_updated = "2026-04-01"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
-
-[limit]
-context = 200000
-output = 131072
-
-[modalities]
-input = ["text", "image", "video", "pdf"]
-output = ["text"]

From 6f0026db1f496299da7bc4f3d75c15470c9590c0 Mon Sep 17 00:00:00 2001
From: heimoshuiyu <heimoshuiyu@gmail.com>
Date: Thu, 17 Sep 2026 23:55:04 +0800
Subject: [PATCH 1529/2012] fix(providers): restore zhipuai-coding-plan
 glm-5.3-highspeed

#7331 classified glm-5.3-highspeed as stale from a single-plan 403
'not available on plan'. That is tier access control, not retirement:
the coding endpoint still serves the id as-is on eligible plan tiers
(first-hand verified 2026-09-17). The same PR kept the byte-identical
zai-coding-plan twin, and tier-gated highspeed variants are established
in this catalog (kimi-for-coding, minimax-coding-plan).

Only glm-5.3-highspeed is restored; the other entries dropped in #7331
have no evidence of being served on any tier. Leading comment now
documents the tier gating.
---
 .../models/glm-5.3-highspeed.toml             | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml

diff --git a/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml b/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml
new file mode 100644
index 00000000000..90c28cdec13
--- /dev/null
+++ b/providers/zhipuai-coding-plan/models/glm-5.3-highspeed.toml
@@ -0,0 +1,21 @@
+# GLM-5.3 Highspeed is tier-gated on the coding endpoint: served as-is on
+# eligible (higher) plan tiers, 403 "not available on plan" on lower tiers
+# (#7331 tested a single lower tier). First-hand verified served via
+# open.bigmodel.cn/api/coding/paas/v4 on 2026-09-17.
+# https://docs.bigmodel.cn/cn/coding-plan/latest-model
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 Highspeed"
+
+# Effort levels mirror the base GLM-5.3 (low, high, max).
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 0fa351dc22f128d70c5cbda0c862b027d8a4a971 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 16:29:03 +0000
Subject: [PATCH 1530/2012] chore(sync): update Kilo model catalog (#7336)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 8b9c29ab7a37e4a37901b371791731c0354de07c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 16:29:08 +0000
Subject: [PATCH 1531/2012] chore(sync): update OpenRouter model catalog
 (#7337)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/tencent/hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From e5c7df1e3ee4b733164c09ab55b5dea7ac696bdb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 17:22:57 +0000
Subject: [PATCH 1532/2012] chore(sync): update Kilo model catalog (#7338)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f09092fb568 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From 5e9fdbc3d88cb831627a93dc1362cc10773c176c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 17:23:04 +0000
Subject: [PATCH 1533/2012] chore(sync): update OpenRouter model catalog
 (#7339)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..ec32a69e6bd 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From 49881100d13bd8c96cd87d010a97580f3e901f0f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 18:32:01 +0000
Subject: [PATCH 1534/2012] chore(sync): update Kilo model catalog (#7341)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-27b:free.toml   | 14 ++++++++++++++
 providers/kilo/models/~moonshotai/kimi-latest.toml |  6 +++---
 2 files changed, 17 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/qwen/qwen3.8-27b:free.toml

diff --git a/providers/kilo/models/qwen/qwen3.8-27b:free.toml b/providers/kilo/models/qwen/qwen3.8-27b:free.toml
new file mode 100644
index 00000000000..0b7df919a8f
--- /dev/null
+++ b/providers/kilo/models/qwen/qwen3.8-27b:free.toml
@@ -0,0 +1,14 @@
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen: Qwen3.8 27B (free)"
+description = "Qwen3.8 27B is an open-weight dense vision-language model from Qwen. It is suited for coding, professional workflows, research, multimodal interaction, and long-running agent tasks, with flexible thinking that can be..."
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+output = 235_929
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f09092fb568..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 35961758828f9d1fc9dbd7c36046c34e924ee01a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 18:32:25 +0000
Subject: [PATCH 1535/2012] chore(sync): update OpenRouter model catalog
 (#7340)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/qwen/qwen3.8-27b:free.toml         | 19 +++++++++++++++++++
 .../models/~moonshotai/kimi-latest.toml       |  6 +++---
 2 files changed, 22 insertions(+), 3 deletions(-)
 create mode 100644 providers/openrouter/models/qwen/qwen3.8-27b:free.toml

diff --git a/providers/openrouter/models/qwen/qwen3.8-27b:free.toml b/providers/openrouter/models/qwen/qwen3.8-27b:free.toml
new file mode 100644
index 00000000000..51491fc303f
--- /dev/null
+++ b/providers/openrouter/models/qwen/qwen3.8-27b:free.toml
@@ -0,0 +1,19 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8 27B (free)"
+description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+output = 235_929
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index ec32a69e6bd..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 9723c50f43d2eac04defffad10a04a129ed32d86 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 19:23:18 +0000
Subject: [PATCH 1536/2012] chore(sync): update Kilo model catalog (#7344)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-chat.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-chat.toml b/providers/kilo/models/deepseek/deepseek-chat.toml
index f67bbe80e1f..42fda83c47b 100644
--- a/providers/kilo/models/deepseek/deepseek-chat.toml
+++ b/providers/kilo/models/deepseek/deepseek-chat.toml
@@ -4,9 +4,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.2574
-output = 1.0287
+input = 0.32
+output = 0.89
 
 [limit]
-context = 128_000
-output = 16_000
+context = 163_840
+output = 16_384

From 4e164d30bf81c37054506e1ad4e0ac04b5a09f81 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 19:23:53 +0000
Subject: [PATCH 1537/2012] chore(sync): update OpenRouter model catalog
 (#7343)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-chat.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-chat.toml b/providers/openrouter/models/deepseek/deepseek-chat.toml
index a01fc1ca488..7768aadee98 100644
--- a/providers/openrouter/models/deepseek/deepseek-chat.toml
+++ b/providers/openrouter/models/deepseek/deepseek-chat.toml
@@ -3,9 +3,9 @@ attachment = false
 structured_output = true
 
 [cost]
-input = 0.2574
-output = 1.0287
+input = 0.32
+output = 0.89
 
 [limit]
 context = 163_840
-output = 16_000
+output = 16_384

From 99dea1579a300a8f6f60c2f06acf8c135b54f0e1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 20:27:34 +0000
Subject: [PATCH 1538/2012] chore(sync): update NanoGPT model catalog (#7345)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/TEE/gpt-oss-20b.toml      | 14 -----------
 .../meta-llama/llama-3.1-405b-instruct.toml   | 23 ------------------
 .../slowburn/gemma4-31b-splituntied.toml      | 24 +++++++++++++++++++
 3 files changed, 24 insertions(+), 37 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/gpt-oss-20b.toml
 delete mode 100644 providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
 create mode 100644 providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml

diff --git a/providers/nano-gpt/models/TEE/gpt-oss-20b.toml b/providers/nano-gpt/models/TEE/gpt-oss-20b.toml
deleted file mode 100644
index 667c4defe72..00000000000
--- a/providers/nano-gpt/models/TEE/gpt-oss-20b.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "openai/gpt-oss-20b"
-name = "GPT-OSS 20B TEE"
-reasoning = false
-tool_call = false
-structured_output = false
-
-[cost]
-input = 0.04
-output = 0.15
-cache_read = 0.02
-
-[limit]
-input = 131_072
-output = 8_192
diff --git a/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml b/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
deleted file mode 100644
index 7d947439c26..00000000000
--- a/providers/nano-gpt/models/meta-llama/llama-3.1-405b-instruct.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Llama 3.1 405B"
-description = "Llama 3.1 405B hosted by IONOS in Berlin, Germany. Zero data retention."
-family = "llama"
-release_date = "2026-09-10"
-last_updated = "2026-09-10"
-attachment = false
-reasoning = false
-tool_call = true
-structured_output = false
-open_weights = true
-
-[cost]
-input = 2.0265
-output = 2.0265
-
-[limit]
-context = 131_072
-input = 131_072
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml b/providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml
new file mode 100644
index 00000000000..60615a27959
--- /dev/null
+++ b/providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml
@@ -0,0 +1,24 @@
+name = "Gemma 4 31B Split-Untied"
+description = "Slowburn's Split-Untied is a text-only Gemma 4 31B community finetune with an untied BF16 output head, built for creative writing, roleplay, expressive dialogue, and tool use."
+family = "gemma"
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.1
+output = 0.3
+cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 1eaddf4906b77fcd77dc3f3c13267fe9a10fe18f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 21:24:39 +0000
Subject: [PATCH 1539/2012] chore(sync): update Kilo model catalog (#7348)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 473252b6895..b586ef3f991 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.13
+input = 0.04
+output = 0.1
 cache_read = 0.01
 
 [limit]

From b07185f616a15b0f64d1b9bee986b2d32c869fd2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 21:24:44 +0000
Subject: [PATCH 1540/2012] chore(sync): update OpenRouter model catalog
 (#7349)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-v4-flash-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index f6dda11044a..12c99d28b38 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.13
+input = 0.04
+output = 0.1
 cache_read = 0.01
 
 [limit]

From 8d5528980057dd6ecde5be9d0d1e84994bb8a5ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 21:25:07 +0000
Subject: [PATCH 1541/2012] chore(sync): update NanoGPT model catalog (#7350)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../TEE/qwen3.6-35b-a3b-uncensored.toml       | 33 -------------------
 .../models/gemma-4-26b-a4b-uncensored.toml    | 25 --------------
 .../gemma-4-26b-a4b-uncensored:thinking.toml  | 25 --------------
 .../models/qwen/qwen3.8-omni-flash.toml       | 28 ++++++++++++++++
 4 files changed, 28 insertions(+), 83 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/qwen3.6-35b-a3b-uncensored.toml
 delete mode 100644 providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
 delete mode 100644 providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml

diff --git a/providers/nano-gpt/models/TEE/qwen3.6-35b-a3b-uncensored.toml b/providers/nano-gpt/models/TEE/qwen3.6-35b-a3b-uncensored.toml
deleted file mode 100644
index f370a8ecce8..00000000000
--- a/providers/nano-gpt/models/TEE/qwen3.6-35b-a3b-uncensored.toml
+++ /dev/null
@@ -1,33 +0,0 @@
-# Not included in subscription
-name = "Qwen3.6 35B A3B Uncensored TEE"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "qwen3.6"
-release_date = "2026-05-23"
-last_updated = "2026-05-23"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = 1_024
-max = 131_072
-
-[cost]
-input = 0.3
-output = 1.5
-cache_read = 0.15
-
-[limit]
-context = 131_072
-input = 131_072
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
deleted file mode 100644
index e6d38df0eb8..00000000000
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Gemma 4 26B A4B Uncensored"
-description = "Gemma 4 26B A4B Uncensored is an FP8 open-weight multimodal mixture-of-experts model LoRA-tuned for fewer refusals across chat, coding, tool use, and long-context work."
-family = "gemma"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.12
-output = 0.38
-cache_read = 0.06
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml b/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
deleted file mode 100644
index f1ac4a30667..00000000000
--- a/providers/nano-gpt/models/gemma-4-26b-a4b-uncensored:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Gemma 4 26B A4B Uncensored Thinking"
-description = "Gemma 4 26B A4B Uncensored with thinking enabled for more deliberate coding, multimodal analysis, tool use, and long-context problem solving."
-family = "gemma"
-release_date = "2026-07-29"
-last_updated = "2026-07-29"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.12
-output = 0.38
-cache_read = 0.06
-
-[limit]
-context = 262_144
-input = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml b/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..5e043e47421
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml
@@ -0,0 +1,28 @@
+name = "Qwen3.8 Omni Flash"
+description = "Qwen3.8 Omni Flash is a fast omni-modal reasoning model for understanding text, images, audio, and video. It is especially suited to meeting summaries, transcripts and subtitles, speaker-aware audiovisual analysis, and long-form content review. It returns text and supports a nearly one-million-token context window, tool calling, and structured output."
+family = "qwen"
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+
+[limit]
+context = 991_808
+input = 991_808
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "audio"]
+output = ["text"]

From a9a3139f8d59ae0491f1807b1450e838ab9207a9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 22:24:07 +0000
Subject: [PATCH 1542/2012] chore(sync): update Merge Gateway model catalog
 (#7351)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-4-1-20250805.toml         | 3 ---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml   | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 +++----
 4 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/providers/merge-gateway/models/anthropic/claude-opus-4-1-20250805.toml b/providers/merge-gateway/models/anthropic/claude-opus-4-1-20250805.toml
index a062e98b15c..d2f26196362 100644
--- a/providers/merge-gateway/models/anthropic/claude-opus-4-1-20250805.toml
+++ b/providers/merge-gateway/models/anthropic/claude-opus-4-1-20250805.toml
@@ -16,6 +16,3 @@ input = 15
 output = 75
 cache_read = 1.5
 cache_write = 18.75
-
-[modalities]
-input = ["text", "image"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index b86a0e589cd..5ef98e7a3a0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
index 42aa3ae2951..f966f69d01c 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-35b-a3b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-35b-a3b"
 name = "Qwen3.5 35B A3B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 0.459
 cache_read = 0.020357
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 84592d1b8dd..0264c6bc733 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,7 +2,6 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,8 +12,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From be7915bdab0ed1ae874c9dfc03f85f9a20a72742 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 22:24:10 +0000
Subject: [PATCH 1543/2012] chore(sync): update OpenRouter model catalog
 (#7353)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~deepseek/deepseek-flash-latest.toml           | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 98fd23a037b..bdcd94e72e3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text", "image"]

From 627d783376cc9fa1acee6d3c00825d120cb88baf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:21:59 +0000
Subject: [PATCH 1544/2012] chore(sync): update Eden AI model catalog (#7358)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/flexai/Nemotron-3-Super-120B-A12B.toml | 12 ------------
 .../nebius/meta-llama/Llama-3.3-70B-Instruct.toml | 11 -----------
 .../models/together_ai/openai/gpt-oss-20b.toml    | 10 ----------
 .../thinkingmachines/Inkling-Small.toml           | 15 ---------------
 4 files changed, 48 deletions(-)
 delete mode 100644 providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
 delete mode 100644 providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
 delete mode 100644 providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
 delete mode 100644 providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml

diff --git a/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml b/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
deleted file mode 100644
index 310bb953a9d..00000000000
--- a/providers/edenai/models/flexai/Nemotron-3-Super-120B-A12B.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "nvidia/nemotron-3-super-120b-a12b"
-name = "Nemotron 3 Super 120B A12B (FlexAI)"
-tool_call = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium"]
-
-[cost]
-input = 0.085
-output = 0.4
diff --git a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
deleted file mode 100644
index 808d6b6ba5f..00000000000
--- a/providers/edenai/models/nebius/meta-llama/Llama-3.3-70B-Instruct.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "meta/llama-3.3-70b-instruct"
-name = "Llama-3.3-70B-Instruct (Nebius)"
-structured_output = true
-
-[cost]
-input = 0.13
-output = 0.4
-cache_read = 0.13
-
-[limit]
-context = 131_072
diff --git a/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml b/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
deleted file mode 100644
index 4ac31b02dbc..00000000000
--- a/providers/edenai/models/together_ai/openai/gpt-oss-20b.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "openai/gpt-oss-20b"
-name = "GPT OSS 20B (Together AI)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.05
-output = 0.2
diff --git a/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml b/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml
deleted file mode 100644
index 9627b70053f..00000000000
--- a/providers/edenai/models/together_ai/thinkingmachines/Inkling-Small.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "thinkingmachines/inkling-small"
-name = "Inkling Small (Together AI)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "max"]
-
-[cost]
-input = 0.5
-output = 1.2
-cache_read = 0.1
-
-[limit]
-context = 524_288

From 0baada4f61b7a08295bcce894161e08ff90d1475 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:22:02 +0000
Subject: [PATCH 1545/2012] chore(sync): update OpenRouter model catalog
 (#7355)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../union-alpha.toml => unbiased/pareto.toml}        | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
 rename providers/openrouter/models/{stealth/union-alpha.toml => unbiased/pareto.toml} (73%)

diff --git a/providers/openrouter/models/stealth/union-alpha.toml b/providers/openrouter/models/unbiased/pareto.toml
similarity index 73%
rename from providers/openrouter/models/stealth/union-alpha.toml
rename to providers/openrouter/models/unbiased/pareto.toml
index fed182b70f6..627dcaf474b 100644
--- a/providers/openrouter/models/stealth/union-alpha.toml
+++ b/providers/openrouter/models/unbiased/pareto.toml
@@ -1,8 +1,7 @@
-name = "Union Alpha"
+name = "Pareto"
 description = "Multimodal model for analyzing text, images, documents, and rich media"
-family = "alpha"
-release_date = "2026-09-16"
-last_updated = "2026-09-16"
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
 attachment = true
 reasoning = false
 temperature = true
@@ -11,8 +10,9 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 2.5
+output = 7.5
+cache_read = 0.25
 
 [limit]
 context = 262_144

From 87dbc7e3c90b733a43aa7fe71be6bc79529990b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:22:06 +0000
Subject: [PATCH 1546/2012] chore(sync): update NanoGPT model catalog (#7357)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml         | 4 ++--
 providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml   | 4 ++--
 .../models/qwen/qwen3.8-27b-obliterated:thinking.toml         | 4 ++--
 providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml         | 4 ++--
 providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml    | 4 ++--
 .../nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
index 64b901ef6c3..afff05408b8 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-fable.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.125
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
index 4a735ca88fa..a9ba3297827 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.2
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
index 00bcf527067..2913df38ed3 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-obliterated:thinking.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.2
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml
index 19c33f74ea6..c8df4c54ae7 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-queen.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.125
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index 88a6536a1da..516d35fc982 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.125
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 8dfeb5f15a4..169511f2187 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -16,8 +16,8 @@ output = 1.5
 cache_read = 0.125
 
 [limit]
-context = 262_144
-input = 262_144
+context = 524_288
+input = 524_288
 output = 32_768
 
 [modalities]

From a4dd2b42be3558bf514a8a56011a2257a25522ea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:22:14 +0000
Subject: [PATCH 1547/2012] chore(sync): update Kilo model catalog (#7356)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/stealth/union-alpha.toml      | 23 -------------------
 providers/kilo/models/unbiased/pareto.toml    | 23 +++++++++++++++++++
 .../~deepseek/deepseek-flash-latest.toml      | 10 ++++----
 3 files changed, 28 insertions(+), 28 deletions(-)
 delete mode 100644 providers/kilo/models/stealth/union-alpha.toml
 create mode 100644 providers/kilo/models/unbiased/pareto.toml

diff --git a/providers/kilo/models/stealth/union-alpha.toml b/providers/kilo/models/stealth/union-alpha.toml
deleted file mode 100644
index cc5ad3f5672..00000000000
--- a/providers/kilo/models/stealth/union-alpha.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Union Alpha"
-description = "Union Alpha is a multimodal model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks. Union Alpha is a stealth model...."
-family = "alpha"
-release_date = "2026-09-16"
-last_updated = "2026-09-16"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/kilo/models/unbiased/pareto.toml b/providers/kilo/models/unbiased/pareto.toml
new file mode 100644
index 00000000000..7e1b27a45c1
--- /dev/null
+++ b/providers/kilo/models/unbiased/pareto.toml
@@ -0,0 +1,23 @@
+name = "Pareto"
+description = "Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks."
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[cost]
+input = 2.5
+output = 7.5
+cache_read = 0.25
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index f491c2853c8..0b2d5078274 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
-context = 1_000_000
-output = 393_216
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text", "image"]

From 14f892c0a44c0a0bcdf9805e7fd695402b4d4218 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 19:55:46 -0400
Subject: [PATCH 1548/2012] chore: remove expired Union Alpha model (#7354)

Co-authored-by: thdxr <826656+thdxr@users.noreply.github.com>
---
 providers/opencode-go/models/union-alpha.toml | 25 -------------------
 providers/opencode/models/union-alpha.toml    | 25 -------------------
 2 files changed, 50 deletions(-)
 delete mode 100644 providers/opencode-go/models/union-alpha.toml
 delete mode 100644 providers/opencode/models/union-alpha.toml

diff --git a/providers/opencode-go/models/union-alpha.toml b/providers/opencode-go/models/union-alpha.toml
deleted file mode 100644
index 25794670176..00000000000
--- a/providers/opencode-go/models/union-alpha.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Union Alpha Free"
-description = "Stealth model built for agentic coding"
-release_date = "2026-09-16"
-last_updated = "2026-09-16"
-attachment = true
-reasoning = true
-reasoning_options = []
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[provider]
-npm = "@ai-sdk/anthropic"
\ No newline at end of file
diff --git a/providers/opencode/models/union-alpha.toml b/providers/opencode/models/union-alpha.toml
deleted file mode 100644
index 25794670176..00000000000
--- a/providers/opencode/models/union-alpha.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Union Alpha Free"
-description = "Stealth model built for agentic coding"
-release_date = "2026-09-16"
-last_updated = "2026-09-16"
-attachment = true
-reasoning = true
-reasoning_options = []
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[provider]
-npm = "@ai-sdk/anthropic"
\ No newline at end of file

From 6a40f26b18a21a0fe07649d6ff07b2b4e7fd2363 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 00:59:26 +0000
Subject: [PATCH 1549/2012] chore(sync): update Kilo model catalog (#7360)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 2 +-
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 1 -
 providers/kilo/models/openai/gpt-oss-120b.toml           | 2 +-
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml                  | 9 ++++-----
 .../kilo/models/~deepseek/deepseek-flash-latest.toml     | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml  | 8 ++++----
 8 files changed, 20 insertions(+), 22 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a280435f49f..41e3e4d74b4 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,4 +12,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 262_144
diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/openai/gpt-oss-120b.toml b/providers/kilo/models/openai/gpt-oss-120b.toml
index 2b409022a13..4e8c1b4eab6 100644
--- a/providers/kilo/models/openai/gpt-oss-120b.toml
+++ b/providers/kilo/models/openai/gpt-oss-120b.toml
@@ -11,4 +11,4 @@ output = 0.17
 cache_read = 0.03
 
 [limit]
-output = 117_964
+output = 65_536
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..8ea8b76bcff 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.14
+output = 0.58
+cache_read = 0.035
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fd68fe0c68b 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
+input = 0.5
+output = 2
+cache_read = 0.1
 
 [limit]
-context = 198_000
-output = 16_384
+context = 202_752
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 0b2d5078274..673b730eb39 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.15
+output = 0.6
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index b586ef3f991..cb2fad7d882 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.1
-cache_read = 0.01
+input = 0.0549
+output = 0.1734
+cache_read = 0.0087
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From fcfc5ad21c4635c7822ed75791e5f1aa73ee3cea Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 00:59:33 +0000
Subject: [PATCH 1550/2012] chore(sync): update Venice model catalog (#7361)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/mistral-small-2603.toml | 17 -----------------
 1 file changed, 17 deletions(-)
 delete mode 100644 providers/venice/models/mistral-small-2603.toml

diff --git a/providers/venice/models/mistral-small-2603.toml b/providers/venice/models/mistral-small-2603.toml
deleted file mode 100644
index 0b9b14a6971..00000000000
--- a/providers/venice/models/mistral-small-2603.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-# Live /models (2026-06-25): none|high only; no dedicated reasoning budget.
-# https://api.venice.ai/api/v1/models?type=text (accessed 2026-06-25)
-
-base_model = "mistral/mistral-small-2603"
-last_updated = "2026-06-11"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.1875
-output = 0.75
-
-[limit]
-output = 65_536

From cc60a2eb35239c066efd6b7d79489e72bceab374 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 00:59:42 +0000
Subject: [PATCH 1551/2012] chore(sync): update DigitalOcean model catalog
 (#7362)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/llama-4-maverick.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/digitalocean/models/llama-4-maverick.toml b/providers/digitalocean/models/llama-4-maverick.toml
index a72af75a264..095b2da6674 100644
--- a/providers/digitalocean/models/llama-4-maverick.toml
+++ b/providers/digitalocean/models/llama-4-maverick.toml
@@ -12,8 +12,8 @@ knowledge = "2024-08"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.696
+input = 0.25
+output = 0.87
 
 [limit]
 context = 128_000

From 52785aa3ff32ba272c17ac1cf84afe7a0b30d0dd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 00:59:50 +0000
Subject: [PATCH 1552/2012] chore(sync): update OpenRouter model catalog
 (#7359)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml | 7 +++----
 providers/openrouter/models/moonshotai/kimi-k3.toml      | 6 +++---
 providers/openrouter/models/openai/gpt-oss-120b.toml     | 7 ++++---
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 +++------
 .../models/~deepseek/deepseek-flash-latest.toml          | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml       | 8 ++++----
 10 files changed, 34 insertions(+), 37 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 2186c60b6fb..0417e3027d5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,10 +11,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.2156
+output = 0.6468
+cache_read = 0.00686
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 262_144
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 3eee6643f71..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.07
-output = 0.14
-cache_read = 0.014
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index c11eb6c3dae..618c6fb2bbd 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/openai/gpt-oss-120b.toml b/providers/openrouter/models/openai/gpt-oss-120b.toml
index e888deb71f4..16d06df52b1 100644
--- a/providers/openrouter/models/openai/gpt-oss-120b.toml
+++ b/providers/openrouter/models/openai/gpt-oss-120b.toml
@@ -5,8 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.037
-output = 0.17
+input = 0.15
+output = 0.6
+cache_read = 0.075
 
 [limit]
-output = 117_964
+output = 65_536
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.5
+output = 2
+cache_read = 0.1
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index bdcd94e72e3..dc34ba0ccf7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.15
+output = 0.6
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 12c99d28b38..e8b1b0581ee 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.1
-cache_read = 0.01
+input = 0.0549
+output = 0.1734
+cache_read = 0.0087
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From 970ca99f25e8ebdd30c08d36933aa7921669b59d Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 20:25:42 -0500
Subject: [PATCH 1553/2012] fix: [missing-model] tinfoil: glm-5-3 (#6153)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/tinfoil/models/glm-5-3.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/tinfoil/models/glm-5-3.toml

diff --git a/providers/tinfoil/models/glm-5-3.toml b/providers/tinfoil/models/glm-5-3.toml
new file mode 100644
index 00000000000..5452c715c46
--- /dev/null
+++ b/providers/tinfoil/models/glm-5-3.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-03):
+# - https://inference.tinfoil.sh/v1/models (id glm-5-3; context 1048576; pricing input 1.8 / output 5.75 / cache_read 0.45 per 1M)
+# - https://docs.tinfoil.sh/models/chat.md (GLM-5.3; 1M context; weights Inferact/GLM-5.3-NVFP4)
+# - https://docs.tinfoil.sh/guides/reasoning.md (glm-5-3 effort low|high|max; reasoning always on)
+# Effort: reasoning_effort = low|high|max on Tinfoil; reasoning is always on (no none; enable_thinking not honoured).
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.8
+output = 5.75
+cache_read = 0.45
+
+[limit]
+context = 1_048_576

From 799ca9e1fbb66614e4f8eabead12e5772ca0cbac Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 01:34:50 +0000
Subject: [PATCH 1554/2012] chore(sync): update Kilo model catalog (#7364)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731:free.toml     | 15 +++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml     |  6 +++---
 2 files changed, 18 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml
new file mode 100644
index 00000000000..73a49138fbe
--- /dev/null
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek: DeepSeek V4 Flash 0731 (free)"
+description = "DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows...."
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..283ade8c995 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From cb22b61f357dcdce1ae31875d5d6760563136dce Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 01:34:54 +0000
Subject: [PATCH 1555/2012] chore(sync): update OpenRouter model catalog
 (#7363)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731:free.toml | 20 +++++++++++++++++++
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 +++---
 .../models/deepseek/deepseek-v4.1-flash.toml  |  6 +++---
 .../models/~deepseek/deepseek-pro-latest.toml |  6 +++---
 4 files changed, 29 insertions(+), 9 deletions(-)
 create mode 100644 providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml
new file mode 100644
index 00000000000..1f431a902ab
--- /dev/null
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml
@@ -0,0 +1,20 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 (free)"
+description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..75641321a8f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576

From 3e5c484bb74edc3533a5ea17120f340ccd76c801 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 01:34:57 +0000
Subject: [PATCH 1556/2012] chore(sync): update NanoGPT model catalog (#7365)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From c314cc4074d459d56b499d91bec094ecb9847c1a Mon Sep 17 00:00:00 2001
From: Joost <15650945+jope35@users.noreply.github.com>
Date: Fri, 18 Sep 2026 03:49:37 +0200
Subject: [PATCH 1557/2012] feat(mistral): add Z.ai GLM-5.3 and fix GLM-5.2
 reasoning efforts (#7334)

Add zai-glm-5-3 hosted by Mistral ($1.40/$4.40 per MTok, $0.14 cached
input), base_model zhipuai/glm-5.3, effort low|high|max.

Fix zai-glm-5-2 reasoning_options: Mistral accepts reasoning_effort=none
(thinking fully disabled, verified live), so values are none|high|max.
low|medium map to high and xhigh maps to max per the GLM-5.2 native
mapping, so aliases stay excluded.
---
 providers/mistral/models/zai-glm-5-2.toml |  5 ++++-
 providers/mistral/models/zai-glm-5-3.toml | 13 +++++++++++++
 2 files changed, 17 insertions(+), 1 deletion(-)
 create mode 100644 providers/mistral/models/zai-glm-5-3.toml

diff --git a/providers/mistral/models/zai-glm-5-2.toml b/providers/mistral/models/zai-glm-5-2.toml
index 3123d2e4883..2d3bc8af322 100644
--- a/providers/mistral/models/zai-glm-5-2.toml
+++ b/providers/mistral/models/zai-glm-5-2.toml
@@ -1,10 +1,13 @@
 # Source: https://docs.mistral.ai/models/zai-glm-5-2 (accessed 2026-08-26)
+# Effort: reasoning_effort = none|high|max. Verified live 2026-09-17: none
+# disables thinking; low|medium map to high, xhigh maps to max (GLM-5.2 native
+# mapping). Model reasons by default (default effort: max).
 base_model = "zhipuai/glm-5.2"
 status = "beta"
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["none", "high", "max"]
 
 [cost]
 input = 1.40
diff --git a/providers/mistral/models/zai-glm-5-3.toml b/providers/mistral/models/zai-glm-5-3.toml
new file mode 100644
index 00000000000..46a9c565458
--- /dev/null
+++ b/providers/mistral/models/zai-glm-5-3.toml
@@ -0,0 +1,13 @@
+# Source: https://docs.mistral.ai/models/zai-glm-5-3 (accessed 2026-09-17)
+# Effort: reasoning_effort = low|high|max; GLM-5.3 native effective levels
+# (unlike 5.2, low is a real level), served unmodified by Mistral.
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.14

From 596b55c1c9b4e752fba3d772db03060fac283e50 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 20:54:17 -0500
Subject: [PATCH 1558/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7352)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): correct Qwen 3.8 Omni metadata

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/alibaba/qwen3.8-omni-flash.toml        | 20 +++++++++++++++++++
 .../models/alibaba/qwen3.8-omni-flash.toml    | 15 ++++++++++++++
 2 files changed, 35 insertions(+)
 create mode 100644 models/alibaba/qwen3.8-omni-flash.toml
 create mode 100644 providers/vercel/models/alibaba/qwen3.8-omni-flash.toml

diff --git a/models/alibaba/qwen3.8-omni-flash.toml b/models/alibaba/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..9308b2b7d6e
--- /dev/null
+++ b/models/alibaba/qwen3.8-omni-flash.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-18):
+# https://www.alibabacloud.com/help/en/model-studio/qwen3-8-omni-flash
+# https://www.alibabacloud.com/help/en/model-studio/qwen-omni
+name = "Qwen3.8 Omni Flash"
+description = "Qwen omni model for text, vision, audio, and multimodal agent tasks"
+family = "qwen"
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text", "image", "audio", "video"]
+output = ["text"]
diff --git a/providers/vercel/models/alibaba/qwen3.8-omni-flash.toml b/providers/vercel/models/alibaba/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..f8c34493ce1
--- /dev/null
+++ b/providers/vercel/models/alibaba/qwen3.8-omni-flash.toml
@@ -0,0 +1,15 @@
+# Effort: reasoning_effort = none|minimal|low|medium|high|xhigh|max
+# Alibaba maps minimal to low and high|max to xhigh; none disables thinking.
+# https://www.alibabacloud.com/help/en/model-studio/qwen-omni
+base_model = "alibaba/qwen3.8-omni-flash"
+name = "Qwen 3.8 Omni Flash"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From f73b04b40eecc0676262c4b6af38b3c2da7df64f Mon Sep 17 00:00:00 2001
From: charles <crosenblum@coreweave.com>
Date: Thu, 17 Sep 2026 21:54:33 -0400
Subject: [PATCH 1559/2012] fix(wandb): use CoreWeave provider logo (#7335)

---
 providers/wandb/logo.svg | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/providers/wandb/logo.svg b/providers/wandb/logo.svg
index 6faa9e9ae5b..57144021e2c 100644
--- a/providers/wandb/logo.svg
+++ b/providers/wandb/logo.svg
@@ -1,3 +1,6 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
-  <path d="M2.48 0a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm19.04 0a1.55 1.55 0 1 0 0 3.101 1.55 1.55 0 0 0 0-3.101zM12 2.295a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zM2.48 5.272a2.48 2.48 0 1 0 0 4.96 2.48 2.48 0 0 0 0-4.96zm19.04 0a2.48 2.48 0 1 0 0 4.96 2.48 2.48 0 0 0 0-4.96zM12 8.496a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm-9.52 3.907a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1zm19.04 0a1.55 1.55 0 1 0 0 3.102 1.55 1.55 0 0 0 0-3.102zM12 13.767a2.48 2.48 0 1 0 0 4.962 2.48 2.48 0 0 0 0-4.962zm-9.52 3.907a2.48 2.48 0 1 0 .001 4.962 2.48 2.48 0 0 0 0-4.962zm19.04.93a1.55 1.55 0 1 0 0 3.102 1.55 1.55 0 0 0 0-3.101zM12 20.9a1.55 1.55 0 1 0 0 3.1 1.55 1.55 0 0 0 0-3.1Z" />
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="15 -7 160 160" fill="currentColor">
+  <path fill-rule="evenodd" d="M80.37 94.96H59.71c-1.26 0-2.52-.79-3.28-2.1l-9.57-17.83c-.5-1.31-.5-2.62 0-3.67l9.57-17.83c.76-1.31 2.02-2.1 3.28-2.1h11.84c1.26 0 2.52-.79 3.28-2.1l8.57-16c.76-1.31-.25-2.88-1.51-2.88H50.14c-2.77 0-5.04 1.57-6.55 3.93l-18.9 35.14a8.113 8.113 0 0 0 0 7.6L43.59 112a7.47 7.47 0 0 0 6.55 3.93h40.57c1.51 0 2.27-1.57 1.51-2.88L83.4 96.79c-.25-1.31-1.51-1.84-3.02-1.84Z"/>
+  <path d="m136.31 45.92-8.06-14.16c-.76-1.31-2.52-1.31-3.28 0L90.95 94.69c-.5 1.31-.5 2.62 0 3.67l8.82 16.52c.76 1.31 2.52 1.31 3.28 0l33.26-61.36a8.113 8.113 0 0 0 0-7.6Z"/>
+  <path fill-rule="evenodd" d="m95.24 74.77 10.08-18.62c.5-1.05.5-2.62 0-3.93L93.73 31.5c-.76-1.31-2.52-1.31-3.28 0l-9.83 18.09c-.5 1.05-.5 2.62 0 3.67L91.96 74.5c.76 1.57 2.77 1.57 3.28.26Z"/>
+  <path d="m166.04 45.92-8.06-14.16c-.76-1.31-2.52-1.31-3.28 0l-34.02 62.93c-.5 1.31-.5 2.62 0 3.67l9.07 16.52c.76 1.31 2.52 1.31 3.28 0l33.26-61.36c1.26-2.36 1.26-5.24-.25-7.6Z"/>
 </svg>

From e6ffd6160fd9b1b9374a5dc30473d7dd4c4e6d12 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 02:35:34 +0000
Subject: [PATCH 1560/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7366)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../empiriolabs/models/qwen3-8-omni-flash.toml  | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/empiriolabs/models/qwen3-8-omni-flash.toml

diff --git a/providers/empiriolabs/models/qwen3-8-omni-flash.toml b/providers/empiriolabs/models/qwen3-8-omni-flash.toml
new file mode 100644
index 00000000000..3f74a9fcb03
--- /dev/null
+++ b/providers/empiriolabs/models/qwen3-8-omni-flash.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-omni-flash"
+temperature = true
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 262_144
+
+[cost]
+input = 0.3
+output = 0.94
+cache_read = 0.3

From 26649e24fae2a8be555e0aef4d159d25e3c82e09 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 02:35:37 +0000
Subject: [PATCH 1561/2012] chore(sync): update NanoGPT model catalog (#7367)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/qwen/qwen3.8-omni-flash.toml  | 15 +--------------
 1 file changed, 1 insertion(+), 14 deletions(-)

diff --git a/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml b/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml
index 5e043e47421..9c6c8edf86a 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-omni-flash.toml
@@ -1,13 +1,5 @@
-name = "Qwen3.8 Omni Flash"
-description = "Qwen3.8 Omni Flash is a fast omni-modal reasoning model for understanding text, images, audio, and video. It is especially suited to meeting summaries, transcripts and subtitles, speaker-aware audiovisual analysis, and long-form content review. It returns text and supports a nearly one-million-token context window, tool calling, and structured output."
-family = "qwen"
-release_date = "2026-09-17"
-last_updated = "2026-09-17"
-attachment = true
-reasoning = true
-tool_call = true
+base_model = "alibaba/qwen3.8-omni-flash"
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -21,8 +13,3 @@ cache_read = 0.016
 [limit]
 context = 991_808
 input = 991_808
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "audio"]
-output = ["text"]

From 12d6e1f99ddfd41d7fb64947208f9dc78544492d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 02:35:47 +0000
Subject: [PATCH 1562/2012] chore(sync): update Kilo model catalog (#7368)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 38aa9fed430..b6ea0fea56d 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.7735
+output = 2.431
+cache_read = 0.127075
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 19d59cf311b7a9894dd2cfd3a50d252e8d24a794 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 02:36:01 +0000
Subject: [PATCH 1563/2012] chore(sync): update OpenRouter model catalog
 (#7369)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~z-ai/glm-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3e3ce108967..25e0814a971 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.7735
+output = 2.431
+cache_read = 0.127075
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 943_718
 
 [modalities]
 input = ["text"]

From 666aec20bc50aa8c2c8c938b63f021703d6a92e3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 03:30:03 +0000
Subject: [PATCH 1564/2012] chore(sync): update Kilo model catalog (#7371)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml                  | 9 +++++----
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml  | 6 +++---
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
 
 [limit]
-context = 202_752
+context = 198_000
+output = 16_384
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index cb2fad7d882..396a029912a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.0549
-output = 0.1734
-cache_read = 0.0087
+input = 0.0558
+output = 0.1767
+cache_read = 0.0088
 
 [limit]
 context = 1_048_576

From 285557390e8b62bf7a3e17ef50b7fd0374d0fd14 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 03:30:10 +0000
Subject: [PATCH 1565/2012] chore(sync): update OpenRouter model catalog
 (#7370)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 ++++++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml       | 6 +++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index e8b1b0581ee..7356cb84c22 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.0549
-output = 0.1734
-cache_read = 0.0087
+input = 0.0558
+output = 0.1767
+cache_read = 0.0088
 
 [limit]
 context = 1_310_720

From c69ea5ca073c78d55c91922b0b4d61360005ca18 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:01:19 -0500
Subject: [PATCH 1566/2012] fix(sync): handle Meta headings and unsupported Jev
 (#7372)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/meta.ts      |  9 ++++-
 .../core/test/cloudflare-ai-gateway.test.ts   | 34 +++++++++++++++++++
 packages/core/test/meta.test.ts               | 10 ++++--
 providers/cloudflare-ai-gateway/curation.toml |  1 +
 4 files changed, 51 insertions(+), 3 deletions(-)

diff --git a/packages/core/src/sync/providers/meta.ts b/packages/core/src/sync/providers/meta.ts
index 7c56b9be2ca..3785e620010 100644
--- a/packages/core/src/sync/providers/meta.ts
+++ b/packages/core/src/sync/providers/meta.ts
@@ -42,7 +42,7 @@ function table(markdown: string, header: string[]) {
 }
 
 function tierCost(markdown: string, tier: string) {
-  const sections = markdown.split(/^### /m).filter((section) => section.split("\n")[0]?.includes(`{#${tier}}`));
+  const sections = markdown.split(/^### /m).filter((section) => sectionID(section) === tier);
   if (sections.length !== 1) throw new Error(`Meta docs need exactly one pricing section for ${tier}`);
   const prices = new Map<string, number>();
   for (const [usage, price] of table(sections[0]!, ["Usage", "Price per 1M tokens"])) {
@@ -58,6 +58,13 @@ function tierCost(markdown: string, tier: string) {
   };
 }
 
+function sectionID(section: string) {
+  const heading = section.split("\n")[0]?.trim() ?? "";
+  const explicit = heading.match(/\{#([a-z0-9-]+)\}$/)?.[1];
+  if (explicit !== undefined) return explicit;
+  return heading.toLowerCase().replace(/[^a-z0-9]+/g, "-").replace(/^-+|-+$/g, "");
+}
+
 export function parseMetaModels(raw: unknown): MetaModel[] {
   const response = MetaResponse.parse(raw);
   const rows = table(response.models, ["Model ID", "Tier", "Input modalities", "Output modalities", "Context window"]);
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
index dcabdcc7ed5..b102f7c3d7f 100644
--- a/packages/core/test/cloudflare-ai-gateway.test.ts
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -251,6 +251,40 @@ test("rejects a catalog with no eligible proxied models", async () => {
   }
 });
 
+test("skips curated catalog models without compatible lab metadata", async () => {
+  const originalFetch = globalThis.fetch;
+  const originalToken = process.env.CLOUDFLARE_API_TOKEN;
+  const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
+  process.env.CLOUDFLARE_API_TOKEN = "test";
+  process.env.CLOUDFLARE_ACCOUNT_ID = "test";
+  globalThis.fetch = async (input) => {
+    if (String(input).endsWith("/schema")) return new Response(null, { status: 404 });
+    return new Response(JSON.stringify(catalogPage([
+      {
+        model_id: "typesafe/jev",
+        task: "Text Generation",
+        context_length: 32_000,
+        provider_details: providerDetails({ input_tokens: 0.042, output_tokens: 0 }),
+      },
+      {
+        model_id: "openai/gpt-4.1",
+        task: "Text Generation",
+        context_length: 1_047_576,
+        provider_details: providerDetails({ input_tokens: 2, output_tokens: 8 }),
+      },
+    ])));
+  };
+
+  try {
+    const models = await cloudflareAiGateway.fetchModels();
+    expect(models.map((model) => model.catalog.model_id)).toEqual(["openai/gpt-4.1"]);
+  } finally {
+    globalThis.fetch = originalFetch;
+    restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
+    restoreEnv("CLOUDFLARE_ACCOUNT_ID", originalAccount);
+  }
+});
+
 test("validates Cloudflare page metadata", async () => {
   const originalFetch = globalThis.fetch;
   const originalToken = process.env.CLOUDFLARE_API_TOKEN;
diff --git a/packages/core/test/meta.test.ts b/packages/core/test/meta.test.ts
index 7163c343f54..dd91aa1345b 100644
--- a/packages/core/test/meta.test.ts
+++ b/packages/core/test/meta.test.ts
@@ -23,7 +23,7 @@ const models = `
 `;
 
 const pricing = `
-### Standard tier {#standard-tier}
+### Standard tier
 
 | Usage | Price per 1M tokens |
 | :---- | :---- |
@@ -31,7 +31,7 @@ const pricing = `
 | Input | $1.25 |
 | Output | $4.25 |
 
-### Contributor tier {#contributor-tier}
+### Contributor tier
 
 | Usage | Price per 1M tokens |
 | :---- | :---- |
@@ -60,6 +60,12 @@ test("parses public text models, tier prices in USD/MTok, and context windows",
   ]);
   expect(parseMetaModels({ ...source, pricing: pricing.replace("$0.002", "$0") })[1]?.cost.cache_read).toBe(0);
   expect(parseMetaModels({ ...source, models: models.replaceAll("1,048,576", "1048576") })[0]?.context).toBe(1_048_576);
+  expect(parseMetaModels({
+    ...source,
+    pricing: pricing
+      .replace("### Standard tier", "### Standard tier {#standard-tier}")
+      .replace("### Contributor tier", "### Contributor tier {#contributor-tier}"),
+  })).toHaveLength(2);
 });
 
 test("rejects incomplete or changed docs rather than guessing prices or limits", () => {
diff --git a/providers/cloudflare-ai-gateway/curation.toml b/providers/cloudflare-ai-gateway/curation.toml
index 40337abd089..8ec54100a76 100644
--- a/providers/cloudflare-ai-gateway/curation.toml
+++ b/providers/cloudflare-ai-gateway/curation.toml
@@ -35,6 +35,7 @@ skip = [
   "google/gemini-3.7-flash",
   "google/gemini-3.8-flash",
   "stealth/union-alpha",
+  "typesafe/jev",
 ]
 
 [models."anthropic/claude-fable-5"]

From 726ed92549e3d383c911a60c33e67f7fa30742d2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:09:36 -0500
Subject: [PATCH 1567/2012] feat: add Jev to Cloudflare AI Gateway (#7373)

* feat: add TypeSafe Jev

* feat(cloudflare-ai-gateway): add Jev

---------

Co-authored-by: Dax Raad <d@ironbay.co>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/typesafe/jev-latest.toml               | 23 +++++++++++++++++++
 .../core/test/cloudflare-ai-gateway.test.ts   | 12 ++++++++--
 providers/cloudflare-ai-gateway/curation.toml |  6 ++++-
 .../models/typesafe/jev.toml                  |  8 +++++++
 providers/opencode/models/jev-latest.toml     |  6 +++++
 providers/vercel/models/typesafe-ai/jev.toml  | 17 +-------------
 6 files changed, 53 insertions(+), 19 deletions(-)
 create mode 100644 models/typesafe/jev-latest.toml
 create mode 100644 providers/cloudflare-ai-gateway/models/typesafe/jev.toml
 create mode 100644 providers/opencode/models/jev-latest.toml

diff --git a/models/typesafe/jev-latest.toml b/models/typesafe/jev-latest.toml
new file mode 100644
index 00000000000..9b2f3a75896
--- /dev/null
+++ b/models/typesafe/jev-latest.toml
@@ -0,0 +1,23 @@
+# Sources:
+# - https://docs.typesafe.ai/models.md
+# - https://docs.typesafe.ai/api.md
+# - https://typesafe.ai/blog/introducing-system-one-models-and-jev
+# The 64k context limit covers the state plus all questions in one request.
+name = "Jev"
+description = "System One model for fast, typed probabilistic decisions over text or structured state"
+release_date = "2026-09-15"
+last_updated = "2026-09-15"
+attachment = false
+reasoning = false
+temperature = false
+tool_call = false
+structured_output = true
+open_weights = false
+
+[limit]
+context = 64000
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
index b102f7c3d7f..316466554dc 100644
--- a/packages/core/test/cloudflare-ai-gateway.test.ts
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -251,7 +251,7 @@ test("rejects a catalog with no eligible proxied models", async () => {
   }
 });
 
-test("skips curated catalog models without compatible lab metadata", async () => {
+test("syncs Jev through shared TypeSafe metadata without inventing tool support", async () => {
   const originalFetch = globalThis.fetch;
   const originalToken = process.env.CLOUDFLARE_API_TOKEN;
   const originalAccount = process.env.CLOUDFLARE_ACCOUNT_ID;
@@ -277,7 +277,15 @@ test("skips curated catalog models without compatible lab metadata", async () =>
 
   try {
     const models = await cloudflareAiGateway.fetchModels();
-    expect(models.map((model) => model.catalog.model_id)).toEqual(["openai/gpt-4.1"]);
+    expect(models.map((model) => model.catalog.model_id)).toEqual(["typesafe/jev", "openai/gpt-4.1"]);
+    expect(cloudflareAiGateway.translateModel(models[0]!, {
+      existing: () => undefined,
+      authored: () => undefined,
+    }).model).toEqual({
+      base_model: "typesafe/jev-latest",
+      cost: { input: 0.042, output: 0 },
+      limit: { context: 32_000 },
+    });
   } finally {
     globalThis.fetch = originalFetch;
     restoreEnv("CLOUDFLARE_API_TOKEN", originalToken);
diff --git a/providers/cloudflare-ai-gateway/curation.toml b/providers/cloudflare-ai-gateway/curation.toml
index 8ec54100a76..b49e5a8d3dd 100644
--- a/providers/cloudflare-ai-gateway/curation.toml
+++ b/providers/cloudflare-ai-gateway/curation.toml
@@ -35,9 +35,13 @@ skip = [
   "google/gemini-3.7-flash",
   "google/gemini-3.8-flash",
   "stealth/union-alpha",
-  "typesafe/jev",
 ]
 
+# Cloudflare exposes TypeSafe's evaluation-only Jev alias over /ai/run. The shared
+# metadata marks its structured output support and lack of tool calling.
+[models."typesafe/jev"]
+base_model = "typesafe/jev-latest"
+
 [models."anthropic/claude-fable-5"]
 structured_output = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
diff --git a/providers/cloudflare-ai-gateway/models/typesafe/jev.toml b/providers/cloudflare-ai-gateway/models/typesafe/jev.toml
new file mode 100644
index 00000000000..b4666c376d8
--- /dev/null
+++ b/providers/cloudflare-ai-gateway/models/typesafe/jev.toml
@@ -0,0 +1,8 @@
+base_model = "typesafe/jev-latest"
+
+[cost]
+input = 0.042
+output = 0
+
+[limit]
+context = 32000
diff --git a/providers/opencode/models/jev-latest.toml b/providers/opencode/models/jev-latest.toml
new file mode 100644
index 00000000000..409cf1a8931
--- /dev/null
+++ b/providers/opencode/models/jev-latest.toml
@@ -0,0 +1,6 @@
+# Pricing source: https://typesafe.ai/blog/introducing-system-one-models-and-jev
+base_model = "typesafe/jev-latest"
+
+[cost]
+input = 0.042
+output = 0
diff --git a/providers/vercel/models/typesafe-ai/jev.toml b/providers/vercel/models/typesafe-ai/jev.toml
index ea15dd157f6..e3c1d0652ad 100644
--- a/providers/vercel/models/typesafe-ai/jev.toml
+++ b/providers/vercel/models/typesafe-ai/jev.toml
@@ -1,20 +1,5 @@
-name = "Jev"
-description = "Evaluation model for fast, structured decisions in software, returning choices, scores, and boolean probabilities"
-release_date = "2026-09-15"
-last_updated = "2026-09-15"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
+base_model = "typesafe/jev-latest"
 
 [cost]
 input = 0.042
 output = 0
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 05d80192847e0379a850bece68b030cd09863f41 Mon Sep 17 00:00:00 2001
From: Andre Landgraf <andre.timo.landgraf@gmail.com>
Date: Thu, 17 Sep 2026 21:09:50 -0700
Subject: [PATCH 1568/2012] Neon: add grok-4-6 cost (#7332)

The Neon overlay listed Grok 4.6 without a price after the public catalog published one.
---
 providers/neon/models/grok-4-6.toml | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/providers/neon/models/grok-4-6.toml b/providers/neon/models/grok-4-6.toml
index b17b6990ea0..e611f3d8cdd 100644
--- a/providers/neon/models/grok-4-6.toml
+++ b/providers/neon/models/grok-4-6.toml
@@ -5,12 +5,16 @@
 # `none` and `minimal` return 200 but do not expose a reasoning channel to
 # compare against, and first-party xAI does not treat this model as disableable.
 # Do not advertise an off switch we cannot verify.
+# Cost: https://neon.com/models.json grok-4-6 (USD / 1M tokens).
 base_model = "xai/grok-4.6"
 reasoning_options = [
   { type = "effort", values = ["low", "medium", "high", "xhigh"] },
 ]
 
-# Neon rate is not published yet. Do not copy first-party xAI prices.
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
 
 # Lab output is 500_000. The gateway names 524_288:
 # 'max_tokens' (100000000) exceeds model maximum (524288). Measured 2026-09-04.

From eba7e85ebe20edbb1a2de528081a63db52012ae7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 04:29:47 +0000
Subject: [PATCH 1569/2012] chore(sync): update OpenRouter model catalog
 (#7374)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml    | 6 +++---
 .../openrouter/models/mistralai/mistral-large-2512.toml    | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml              | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml   | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml          | 6 +++---
 7 files changed, 22 insertions(+), 21 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 95460b6f5f3..86c8922eef2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.6
-output = 3.2
-cache_read = 0.135
+input = 0.94336
+output = 1.88672
+cache_read = 0.079596
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/mistralai/mistral-large-2512.toml b/providers/openrouter/models/mistralai/mistral-large-2512.toml
index 8dcd57cfa35..1d8069e28ff 100644
--- a/providers/openrouter/models/mistralai/mistral-large-2512.toml
+++ b/providers/openrouter/models/mistralai/mistral-large-2512.toml
@@ -2,9 +2,9 @@ base_model = "mistral/mistral-large-2512"
 structured_output = true
 
 [cost]
-input = 0.5
-output = 1.5
-cache_read = 0.05
+input = 0.55
+output = 1.65
+cache_read = 0.055
 
 [limit]
 output = 209_715
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 557dd11171f..95694d19041 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,10 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.14
+input = 0.4875
+output = 1.56
+cache_read = 0.091
 
 [limit]
 context = 1_048_576
+output = 163_840
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 75641321a8f..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 25e0814a971..9d68ba72961 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7735
-output = 2.431
-cache_read = 0.127075
+input = 0.833
+output = 2.618
+cache_read = 0.13685
 
 [limit]
 context = 1_310_720

From 93e3bb53dc29b60dac405a5ae31d2409e1db5584 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 04:29:50 +0000
Subject: [PATCH 1570/2012] chore(sync): update NanoGPT model catalog (#7375)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000

From bd742bfc971e7589074720ebf1c9a4d655f0e4e1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 04:29:55 +0000
Subject: [PATCH 1571/2012] chore(sync): update Ofox model catalog (#7377)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml     |  6 +++---
 .../ofox/models/deepseek/deepseek-v4-pro-0813.toml  |  6 +++---
 .../ofox/models/deepseek/deepseek-v4.1-flash.toml   |  6 +++---
 providers/ofox/models/openai/gpt-4.1-mini.toml      |  6 +++---
 providers/ofox/models/openai/gpt-4.1.toml           |  6 +++---
 providers/ofox/models/openai/gpt-4o-mini.toml       |  6 +++---
 providers/ofox/models/openai/gpt-4o.toml            |  6 +++---
 providers/ofox/models/openai/gpt-5-mini.toml        | 11 +++++++----
 providers/ofox/models/openai/gpt-5-nano.toml        |  6 +++---
 providers/ofox/models/openai/gpt-5.1-codex-max.toml | 11 +++++++----
 .../ofox/models/openai/gpt-5.1-codex-mini.toml      | 11 +++++++----
 providers/ofox/models/openai/gpt-5.1.toml           | 11 +++++++----
 providers/ofox/models/openai/gpt-5.2-codex.toml     | 11 +++++++----
 providers/ofox/models/openai/gpt-5.2.toml           | 11 +++++++----
 providers/ofox/models/openai/gpt-5.3-codex.toml     | 11 +++++++----
 providers/ofox/models/openai/gpt-5.4-mini.toml      | 11 +++++++----
 providers/ofox/models/openai/gpt-5.4-nano.toml      | 11 +++++++----
 providers/ofox/models/openai/gpt-5.4-pro.toml       |  9 ++++++---
 providers/ofox/models/openai/gpt-5.4.toml           | 11 +++++++----
 providers/ofox/models/openai/gpt-5.5.toml           | 11 +++++++----
 providers/ofox/models/openai/gpt-5.toml             | 11 +++++++----
 providers/ofox/models/openai/gpt-6-astra.toml       | 13 ++++++++-----
 22 files changed, 122 insertions(+), 80 deletions(-)

diff --git a/providers/ofox/models/deepseek/deepseek-v4-flash-0731.toml b/providers/ofox/models/deepseek/deepseek-v4-flash-0731.toml
index 846d1a51acf..9bbcfa17ecf 100644
--- a/providers/ofox/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/ofox/models/deepseek/deepseek-v4-flash-0731.toml
@@ -12,6 +12,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.308
+output = 0.924
+cache_read = 0.0098
diff --git a/providers/ofox/models/deepseek/deepseek-v4-pro-0813.toml b/providers/ofox/models/deepseek/deepseek-v4-pro-0813.toml
index 28bd1ad9388..be532dcc24e 100644
--- a/providers/ofox/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/ofox/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,6 +12,6 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.924
+output = 2.772
+cache_read = 0.0308
diff --git a/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml b/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml
index 625c4e0903d..ab0f22f5c0c 100644
--- a/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/ofox/models/deepseek/deepseek-v4.1-flash.toml
@@ -13,6 +13,6 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.21
+output = 0.84
+cache_read = 0.0042
diff --git a/providers/ofox/models/openai/gpt-4.1-mini.toml b/providers/ofox/models/openai/gpt-4.1-mini.toml
index fed0af7b191..7db37cb4a30 100644
--- a/providers/ofox/models/openai/gpt-4.1-mini.toml
+++ b/providers/ofox/models/openai/gpt-4.1-mini.toml
@@ -3,9 +3,9 @@
 base_model = "openai/gpt-4.1-mini"
 
 [cost]
-input = 0.4
-output = 1.6
-cache_read = 0.1
+input = 0.32
+output = 1.28
+cache_read = 0.08
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-4.1.toml b/providers/ofox/models/openai/gpt-4.1.toml
index 89ea7a1baef..873c885bb86 100644
--- a/providers/ofox/models/openai/gpt-4.1.toml
+++ b/providers/ofox/models/openai/gpt-4.1.toml
@@ -4,9 +4,9 @@
 base_model = "openai/gpt-4.1"
 
 [cost]
-input = 2
-output = 8
-cache_read = 0.5
+input = 1.6
+output = 6.4
+cache_read = 0.4
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-4o-mini.toml b/providers/ofox/models/openai/gpt-4o-mini.toml
index 1a04244651f..5c2ab7b62d3 100644
--- a/providers/ofox/models/openai/gpt-4o-mini.toml
+++ b/providers/ofox/models/openai/gpt-4o-mini.toml
@@ -4,9 +4,9 @@
 base_model = "openai/gpt-4o-mini"
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.075
+input = 0.12
+output = 0.48
+cache_read = 0.06
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-4o.toml b/providers/ofox/models/openai/gpt-4o.toml
index 5375b57603f..abe97a2e139 100644
--- a/providers/ofox/models/openai/gpt-4o.toml
+++ b/providers/ofox/models/openai/gpt-4o.toml
@@ -3,9 +3,9 @@
 base_model = "openai/gpt-4o"
 
 [cost]
-input = 2.5
-output = 10
-cache_read = 1.25
+input = 2
+output = 8
+cache_read = 1
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5-mini.toml b/providers/ofox/models/openai/gpt-5-mini.toml
index c69ea0d1474..1a91f1921b9 100644
--- a/providers/ofox/models/openai/gpt-5-mini.toml
+++ b/providers/ofox/models/openai/gpt-5-mini.toml
@@ -6,12 +6,15 @@
 # - peers: providers/openai/models/gpt-5-mini.toml, providers/ofox/models/openai/gpt-5.5.toml
 base_model = "openai/gpt-5-mini"
 base_model_omit = ["limit.input"]
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 2
-cache_read = 0.03
+input = 0.2
+output = 1.6
+cache_read = 0.024
 
 [limit]
 context = 256_000
diff --git a/providers/ofox/models/openai/gpt-5-nano.toml b/providers/ofox/models/openai/gpt-5-nano.toml
index 4db26dd6594..144156961a1 100644
--- a/providers/ofox/models/openai/gpt-5-nano.toml
+++ b/providers/ofox/models/openai/gpt-5-nano.toml
@@ -4,9 +4,9 @@ base_model = "openai/gpt-5-nano"
 reasoning = false
 
 [cost]
-input = 0.05
-output = 0.4
-cache_read = 0.01
+input = 0.04
+output = 0.32
+cache_read = 0.008
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.1-codex-max.toml b/providers/ofox/models/openai/gpt-5.1-codex-max.toml
index 91408197ffe..7d613b82a99 100644
--- a/providers/ofox/models/openai/gpt-5.1-codex-max.toml
+++ b/providers/ofox/models/openai/gpt-5.1-codex-max.toml
@@ -5,12 +5,15 @@
 # Effort: reasoning_effort = low|medium|high|xhigh (OpenAI Codex Max / established relay peers)
 base_model = "openai/gpt-5.1-codex-max"
 base_model_omit = ["limit.input"]
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.25
-output = 10
-cache_read = 0.13
+input = 1
+output = 8
+cache_read = 0.104
 
 [limit]
 context = 256_000
diff --git a/providers/ofox/models/openai/gpt-5.1-codex-mini.toml b/providers/ofox/models/openai/gpt-5.1-codex-mini.toml
index 7fbf643e70a..8753622a4f9 100644
--- a/providers/ofox/models/openai/gpt-5.1-codex-mini.toml
+++ b/providers/ofox/models/openai/gpt-5.1-codex-mini.toml
@@ -5,12 +5,15 @@
 # https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/reasoning
 base_model = "openai/gpt-5.1-codex-mini"
 base_model_omit = ["limit.input"]
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 2
-cache_read = 0.03
+input = 0.2
+output = 1.6
+cache_read = 0.024
 
 [limit]
 context = 256_000
diff --git a/providers/ofox/models/openai/gpt-5.1.toml b/providers/ofox/models/openai/gpt-5.1.toml
index f8cf72b22eb..9dd3ad82c15 100644
--- a/providers/ofox/models/openai/gpt-5.1.toml
+++ b/providers/ofox/models/openai/gpt-5.1.toml
@@ -1,11 +1,14 @@
 # https://ofox.ai/models/openai/gpt-5.1 — input $1.25/M, output $10/M, cache_read $0.13/M
 base_model = "openai/gpt-5.1"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 1.25
-output = 10
-cache_read = 0.13
+input = 1
+output = 8
+cache_read = 0.104
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.2-codex.toml b/providers/ofox/models/openai/gpt-5.2-codex.toml
index 456524152de..770e2369a97 100644
--- a/providers/ofox/models/openai/gpt-5.2-codex.toml
+++ b/providers/ofox/models/openai/gpt-5.2-codex.toml
@@ -3,12 +3,15 @@
 # - https://api.ofox.ai/v1/models/openai/gpt-5.2-codex (prompt/completion/cache_read per-token)
 # - https://developers.openai.com/api/docs/models/gpt-5.2-codex (effort: low|medium|high|xhigh)
 base_model = "openai/gpt-5.2-codex"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.75
-output = 14
-cache_read = 0.18
+input = 1.4
+output = 11.2
+cache_read = 0.144
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.2.toml b/providers/ofox/models/openai/gpt-5.2.toml
index 9c764368464..6ab0d2b9bdc 100644
--- a/providers/ofox/models/openai/gpt-5.2.toml
+++ b/providers/ofox/models/openai/gpt-5.2.toml
@@ -3,12 +3,15 @@
 # https://api.ofox.ai/v2/models/catalog (remote model id openai/gpt-5.2)
 # Reasoning: OpenAI first-party gpt-5.2 + ofox openai/gpt-5.5 peer
 base_model = "openai/gpt-5.2"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.75
-output = 14
-cache_read = 0.18
+input = 1.4
+output = 11.2
+cache_read = 0.144
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.3-codex.toml b/providers/ofox/models/openai/gpt-5.3-codex.toml
index 192d650e1df..f331b887502 100644
--- a/providers/ofox/models/openai/gpt-5.3-codex.toml
+++ b/providers/ofox/models/openai/gpt-5.3-codex.toml
@@ -1,11 +1,14 @@
 # Source: https://api.ofox.ai/v2/models/catalog?include=provider_price&limit=500
 base_model = "openai/gpt-5.3-codex"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 1.75
-output = 14
-cache_read = 0.18
+input = 1.4
+output = 11.2
+cache_read = 0.144
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.4-mini.toml b/providers/ofox/models/openai/gpt-5.4-mini.toml
index 1f3a57db295..b3ab0f891fd 100644
--- a/providers/ofox/models/openai/gpt-5.4-mini.toml
+++ b/providers/ofox/models/openai/gpt-5.4-mini.toml
@@ -3,12 +3,15 @@
 # https://ofox.ai/en/models
 # https://ofox.ai/docs (example model id openai/gpt-5.4-mini)
 base_model = "openai/gpt-5.4-mini"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.75
-output = 4.5
-cache_read = 0.075
+input = 0.6
+output = 3.6
+cache_read = 0.06
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.4-nano.toml b/providers/ofox/models/openai/gpt-5.4-nano.toml
index 734225204ad..4dc4fd65cdf 100644
--- a/providers/ofox/models/openai/gpt-5.4-nano.toml
+++ b/providers/ofox/models/openai/gpt-5.4-nano.toml
@@ -2,12 +2,15 @@
 # https://ofox.ai/en/models/openai/gpt-5.4-nano (Ofox list price: $0.20/$1.25, cache_read $0.02)
 # https://developers.openai.com/api/docs/models/gpt-5.4-nano (lab pricing + reasoning effort)
 base_model = "openai/gpt-5.4-nano"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.2
-output = 1.25
-cache_read = 0.02
+input = 0.16
+output = 1
+cache_read = 0.016
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.4-pro.toml b/providers/ofox/models/openai/gpt-5.4-pro.toml
index c0214b5c75c..a08a8a55eba 100644
--- a/providers/ofox/models/openai/gpt-5.4-pro.toml
+++ b/providers/ofox/models/openai/gpt-5.4-pro.toml
@@ -1,11 +1,14 @@
 # Ofox bills a flat 30/180 across the full advertised context window; the Ofox
 # catalog publishes no >=272k context tier, so no [[cost.tiers]] applies here.
 base_model = "openai/gpt-5.4-pro"
-reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["medium", "high", "xhigh"]
 
 [cost]
-input = 30
-output = 180
+input = 24
+output = 144
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.4.toml b/providers/ofox/models/openai/gpt-5.4.toml
index b2dcf2dc973..c48c7a8ab3f 100644
--- a/providers/ofox/models/openai/gpt-5.4.toml
+++ b/providers/ofox/models/openai/gpt-5.4.toml
@@ -1,11 +1,14 @@
 # Source: https://api.ofox.ai/v2/models/catalog?include=provider_price&limit=500
 base_model = "openai/gpt-5.4"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 2.5
-output = 15
-cache_read = 0.25
+input = 2
+output = 12
+cache_read = 0.2
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.5.toml b/providers/ofox/models/openai/gpt-5.5.toml
index 679d0e58dc4..3635ad8bdd9 100644
--- a/providers/ofox/models/openai/gpt-5.5.toml
+++ b/providers/ofox/models/openai/gpt-5.5.toml
@@ -1,10 +1,13 @@
 base_model = "openai/gpt-5.5"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
+input = 4
+output = 24
+cache_read = 0.4
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-5.toml b/providers/ofox/models/openai/gpt-5.toml
index c2393f6cb44..297a9224d26 100644
--- a/providers/ofox/models/openai/gpt-5.toml
+++ b/providers/ofox/models/openai/gpt-5.toml
@@ -3,12 +3,15 @@
 # - https://api.ofox.ai/v2/models/catalog?include=provider_price (input/output/cache_read)
 # Reasoning effort: OpenAI first-party gpt-5 + relay peers (minimal/low/medium/high)
 base_model = "openai/gpt-5"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 1.25
-output = 10
-cache_read = 0.13
+input = 1
+output = 8
+cache_read = 0.104
 
 [provider]
 npm = "@ai-sdk/openai"
diff --git a/providers/ofox/models/openai/gpt-6-astra.toml b/providers/ofox/models/openai/gpt-6-astra.toml
index 4bf30a6b531..56cc44e71c6 100644
--- a/providers/ofox/models/openai/gpt-6-astra.toml
+++ b/providers/ofox/models/openai/gpt-6-astra.toml
@@ -4,13 +4,16 @@
 # - OpenAI model docs (reasoning effort): https://developers.openai.com/api/docs/models/gpt-6-astra
 # Ofox publishes flat $10/$50 (+ cache) with no >=272k context tier in the catalog, so no [[cost.tiers]].
 base_model = "openai/gpt-6-astra"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 10
-output = 50
-cache_read = 1
-cache_write = 12.5
+input = 8
+output = 40
+cache_read = 0.8
+cache_write = 10
 
 [provider]
 npm = "@ai-sdk/openai"

From 1160e57aa98731edaa20fd3b5541608137dbb757 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 04:29:58 +0000
Subject: [PATCH 1572/2012] chore(sync): update Kilo model catalog (#7376)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml            | 6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                  | 1 +
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml              | 6 +++---
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 1ebab0a52b7..3af9cc14392 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 3
-output = 15
-cache_read = 0.3
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bb89edf75cf..f8addeabfc8 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -12,3 +12,4 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
+output = 163_840
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 283ade8c995..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index b6ea0fea56d..7effc6dceef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7735
-output = 2.431
-cache_read = 0.127075
+input = 0.833
+output = 2.618
+cache_read = 0.13685
 
 [limit]
 context = 1_048_576

From 0e576ea593b0c631d3779b41e8fd6c1cc851e52e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 23:57:57 -0500
Subject: [PATCH 1573/2012] feat(cloudflare-ai-gateway): add Pareto (#7378)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/unbiased/pareto.toml                   | 22 +++++++++++++++++++
 .../core/test/cloudflare-ai-gateway.test.ts   | 17 ++++++++++++++
 .../models/unbiased/pareto.toml               |  9 ++++++++
 providers/kilo/models/unbiased/pareto.toml    | 19 +---------------
 .../openrouter/models/unbiased/pareto.toml    | 19 +---------------
 5 files changed, 50 insertions(+), 36 deletions(-)
 create mode 100644 models/unbiased/pareto.toml
 create mode 100644 providers/cloudflare-ai-gateway/models/unbiased/pareto.toml

diff --git a/models/unbiased/pareto.toml b/models/unbiased/pareto.toml
new file mode 100644
index 00000000000..55da1f5151c
--- /dev/null
+++ b/models/unbiased/pareto.toml
@@ -0,0 +1,22 @@
+# Sources:
+# - https://unbiased.ai/model-card/
+# - https://developers.cloudflare.com/ai/models/unbiased/pareto/
+# - https://openrouter.ai/api/v1/models/unbiased/pareto/endpoints
+name = "Pareto"
+description = "Blended multimodal model that runs multiple language models in parallel and returns a synthesized answer"
+release_date = "2026-09-17"
+last_updated = "2026-09-17"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[limit]
+context = 262_144
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/packages/core/test/cloudflare-ai-gateway.test.ts b/packages/core/test/cloudflare-ai-gateway.test.ts
index 316466554dc..3bfdbc566de 100644
--- a/packages/core/test/cloudflare-ai-gateway.test.ts
+++ b/packages/core/test/cloudflare-ai-gateway.test.ts
@@ -293,6 +293,23 @@ test("syncs Jev through shared TypeSafe metadata without inventing tool support"
   }
 });
 
+test("syncs Pareto through shared Unbiased metadata", () => {
+  expect(buildCloudflareAiGatewayModel({
+    model_id: "unbiased/pareto",
+    task: "Text Generation",
+    context_length: 262_144,
+    provider_details: providerDetails({
+      input_tokens: 2.5,
+      output_tokens: 7.5,
+      input_cached_tokens: 0.25,
+    }),
+  }, undefined)).toEqual({
+    base_model: "unbiased/pareto",
+    cost: { input: 2.5, output: 7.5, cache_read: 0.25 },
+    limit: { context: 262_144 },
+  });
+});
+
 test("validates Cloudflare page metadata", async () => {
   const originalFetch = globalThis.fetch;
   const originalToken = process.env.CLOUDFLARE_API_TOKEN;
diff --git a/providers/cloudflare-ai-gateway/models/unbiased/pareto.toml b/providers/cloudflare-ai-gateway/models/unbiased/pareto.toml
new file mode 100644
index 00000000000..20ce41c237d
--- /dev/null
+++ b/providers/cloudflare-ai-gateway/models/unbiased/pareto.toml
@@ -0,0 +1,9 @@
+base_model = "unbiased/pareto"
+
+[cost]
+input = 2.5
+output = 7.5
+cache_read = 0.25
+
+[limit]
+context = 262_144
diff --git a/providers/kilo/models/unbiased/pareto.toml b/providers/kilo/models/unbiased/pareto.toml
index 7e1b27a45c1..19990e346ff 100644
--- a/providers/kilo/models/unbiased/pareto.toml
+++ b/providers/kilo/models/unbiased/pareto.toml
@@ -1,23 +1,6 @@
-name = "Pareto"
-description = "Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks."
-release_date = "2026-09-17"
-last_updated = "2026-09-17"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
+base_model = "unbiased/pareto"
 
 [cost]
 input = 2.5
 output = 7.5
 cache_read = 0.25
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/openrouter/models/unbiased/pareto.toml b/providers/openrouter/models/unbiased/pareto.toml
index 627dcaf474b..19990e346ff 100644
--- a/providers/openrouter/models/unbiased/pareto.toml
+++ b/providers/openrouter/models/unbiased/pareto.toml
@@ -1,23 +1,6 @@
-name = "Pareto"
-description = "Multimodal model for analyzing text, images, documents, and rich media"
-release_date = "2026-09-17"
-last_updated = "2026-09-17"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
+base_model = "unbiased/pareto"
 
 [cost]
 input = 2.5
 output = 7.5
 cache_read = 0.25
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

From 5b42cf9755d9d4c3c077898a56d27c7c4f6de8a8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 05:26:07 +0000
Subject: [PATCH 1574/2012] chore(sync): update OpenRouter model catalog
 (#7379)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/mistralai/mistral-large-2512.toml        | 13 -------------
 providers/openrouter/models/~z-ai/glm-latest.toml   |  8 ++++----
 2 files changed, 4 insertions(+), 17 deletions(-)
 delete mode 100644 providers/openrouter/models/mistralai/mistral-large-2512.toml

diff --git a/providers/openrouter/models/mistralai/mistral-large-2512.toml b/providers/openrouter/models/mistralai/mistral-large-2512.toml
deleted file mode 100644
index 1d8069e28ff..00000000000
--- a/providers/openrouter/models/mistralai/mistral-large-2512.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "mistral/mistral-large-2512"
-structured_output = true
-
-[cost]
-input = 0.55
-output = 1.65
-cache_read = 0.055
-
-[limit]
-output = 209_715
-
-[modalities]
-input = ["text", "image", "pdf"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 9d68ba72961..3e3ce108967 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.833
-output = 2.618
-cache_read = 0.13685
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 235_929
 
 [modalities]
 input = ["text"]

From acd2005918b46e5112069eab8892c69da4227670 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 05:26:09 +0000
Subject: [PATCH 1575/2012] chore(sync): update xAI model catalog (#7380)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../xai/models/grok-imagine-image-2.0.toml     | 13 -------------
 .../xai/models/grok-imagine-image-quality.toml | 18 ------------------
 2 files changed, 31 deletions(-)
 delete mode 100644 providers/xai/models/grok-imagine-image-2.0.toml
 delete mode 100644 providers/xai/models/grok-imagine-image-quality.toml

diff --git a/providers/xai/models/grok-imagine-image-2.0.toml b/providers/xai/models/grok-imagine-image-2.0.toml
deleted file mode 100644
index 0f00ef4f383..00000000000
--- a/providers/xai/models/grok-imagine-image-2.0.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-# Sources:
-# - https://docs.x.ai/docs/models
-# - https://docs.x.ai/developers/models/grok-imagine-image-2.0
-# - https://docs.x.ai/docs/guides/image-generation
-# Pricing: $0.04 per image (not token-based; no [cost] authored, same as sibling imagine image models)
-base_model = "xai/grok-imagine-image-2.0"
-
-[limit]
-context = 64_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["image", "pdf"]
diff --git a/providers/xai/models/grok-imagine-image-quality.toml b/providers/xai/models/grok-imagine-image-quality.toml
deleted file mode 100644
index 12faca1152a..00000000000
--- a/providers/xai/models/grok-imagine-image-quality.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-name = "Grok Imagine Image Quality"
-description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "grok"
-release_date = "2026-04-03"
-last_updated = "2026-04-03"
-attachment = true
-reasoning = false
-temperature = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 16_000
-output = 0
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["image", "pdf"]

From a48d6f2678d5811f7d7bdf5ca2e689ad0b4c8781 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 05:26:13 +0000
Subject: [PATCH 1576/2012] chore(sync): update Cloudflare AI Gateway model
 catalog (#7384)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cloudflare-ai-gateway/models/typesafe/jev.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/cloudflare-ai-gateway/models/typesafe/jev.toml b/providers/cloudflare-ai-gateway/models/typesafe/jev.toml
index b4666c376d8..1e5da088ecd 100644
--- a/providers/cloudflare-ai-gateway/models/typesafe/jev.toml
+++ b/providers/cloudflare-ai-gateway/models/typesafe/jev.toml
@@ -3,6 +3,7 @@ base_model = "typesafe/jev-latest"
 [cost]
 input = 0.042
 output = 0
+cache_read = 0
 
 [limit]
-context = 32000
+context = 32_000

From 5ef285e20ab5e92c06f77a507f69c5ba888afec1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 05:26:18 +0000
Subject: [PATCH 1577/2012] chore(sync): update Kilo model catalog (#7381)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/mistralai/mistral-large-2512.toml  | 14 --------------
 providers/kilo/models/~z-ai/glm-latest.toml        | 10 +++++-----
 2 files changed, 5 insertions(+), 19 deletions(-)
 delete mode 100644 providers/kilo/models/mistralai/mistral-large-2512.toml

diff --git a/providers/kilo/models/mistralai/mistral-large-2512.toml b/providers/kilo/models/mistralai/mistral-large-2512.toml
deleted file mode 100644
index a1f4d549c6b..00000000000
--- a/providers/kilo/models/mistralai/mistral-large-2512.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "mistral/mistral-large-2512"
-description = "Flagship Mistral model for advanced reasoning, coding, and multilingual work"
-structured_output = true
-
-[cost]
-input = 0.5
-output = 1.5
-cache_read = 0.05
-
-[limit]
-output = 209_715
-
-[modalities]
-input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 7effc6dceef..38aa9fed430 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.833
-output = 2.618
-cache_read = 0.13685
+input = 0.8775
+output = 2.97
+cache_read = 0.1755
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From ee6da8c0f6675cf1857d391090d785891ddec186 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 05:26:28 +0000
Subject: [PATCH 1578/2012] chore(sync): update Eden AI model catalog (#7383)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/mistral/mistral-large-2512.toml   | 6 +++---
 providers/edenai/models/mistral/voxtral-small-latest.toml | 1 +
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/edenai/models/mistral/mistral-large-2512.toml b/providers/edenai/models/mistral/mistral-large-2512.toml
index 6944f065750..96902738ba4 100644
--- a/providers/edenai/models/mistral/mistral-large-2512.toml
+++ b/providers/edenai/models/mistral/mistral-large-2512.toml
@@ -2,9 +2,9 @@ base_model = "mistral/mistral-large-2512"
 structured_output = true
 
 [cost]
-input = 0.5
-output = 1.5
-cache_read = 0.05
+input = 0.55
+output = 1.65
+cache_read = 0.055
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/mistral/voxtral-small-latest.toml b/providers/edenai/models/mistral/voxtral-small-latest.toml
index e092715c287..1c35a86c208 100644
--- a/providers/edenai/models/mistral/voxtral-small-latest.toml
+++ b/providers/edenai/models/mistral/voxtral-small-latest.toml
@@ -4,6 +4,7 @@ structured_output = true
 [cost]
 input = 0.1
 output = 0.4
+cache_read = 0.01
 
 [limit]
 context = 32_768

From 97d4c03387252facee6e84f64d1523c33860ffcd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 06:40:31 +0000
Subject: [PATCH 1579/2012] chore(sync): update OpenRouter model catalog
 (#7388)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 6 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 86c8922eef2..95460b6f5f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.94336
-output = 1.88672
-cache_read = 0.079596
+input = 1.6
+output = 3.2
+cache_read = 0.135
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 618c6fb2bbd..3fff8c2342f 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..75641321a8f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b49deb6a06e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 85be7f3e15f755525ed88a55be806ed172e3174a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 06:40:38 +0000
Subject: [PATCH 1580/2012] chore(sync): update NanoGPT model catalog (#7389)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..4e65b119cdc 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..67f17f156de 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.003
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From c655d313f239927ae43e674592f85d93448720b0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 06:40:50 +0000
Subject: [PATCH 1581/2012] chore(sync): update Kilo model catalog (#7390)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml            | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml       | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 3af9cc14392..d09535b980a 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..283ade8c995 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.96
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..f97a233d9f8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2
+output = 11.2
+cache_read = 0.232
 
 [limit]
 context = 1_048_576

From 2dbf7567c471ae0b636557afbf75663649669137 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 07:29:57 +0000
Subject: [PATCH 1582/2012] chore(sync): update OpenRouter model catalog
 (#7393)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 3fff8c2342f..618c6fb2bbd 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index dc34ba0ccf7..bdcd94e72e3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b49deb6a06e..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 74e9f4abd9bf951b43aaad664b18839715c74581 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 07:30:05 +0000
Subject: [PATCH 1583/2012] chore(sync): update Kilo model catalog (#7392)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml              | 6 +++---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml         | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index d09535b980a..3af9cc14392 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 673b730eb39..0b2d5078274 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f97a233d9f8..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2
-output = 11.2
-cache_read = 0.232
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 0c2c6a1bf89bb880aca70e153faac6e9ba47124c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 08:33:34 +0000
Subject: [PATCH 1584/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7394)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gpt-5.6-sol.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/llmgateway/models/gpt-5.6-sol.toml b/providers/llmgateway/models/gpt-5.6-sol.toml
index 45be0a6aee9..8cddfa4d843 100644
--- a/providers/llmgateway/models/gpt-5.6-sol.toml
+++ b/providers/llmgateway/models/gpt-5.6-sol.toml
@@ -5,7 +5,7 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5

From 2cfbacb87ea6cc1546baa0fee24d4cd756123fbf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 08:33:49 +0000
Subject: [PATCH 1585/2012] chore(sync): update LLM Gateway model catalog
 (#7395)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aws-mantle/gpt-5.6-sol.toml                    | 8 ++++----
 .../llmgateway-providers/models/azure/gpt-5.6-sol.toml    | 8 ++++----
 .../llmgateway-providers/models/openai/gpt-5.6-sol.toml   | 8 ++++----
 3 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
index 53e678996d0..671621ddb8f 100644
--- a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
+++ b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
@@ -7,10 +7,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5.5
-output = 33
-cache_read = 0.55
-cache_write = 6.875
+input = 4.4
+output = 22
+cache_read = 0.44
+cache_write = 5.5
 
 [limit]
 context = 278_528
diff --git a/providers/llmgateway-providers/models/azure/gpt-5.6-sol.toml b/providers/llmgateway-providers/models/azure/gpt-5.6-sol.toml
index 31efa601388..1bbdab35ac1 100644
--- a/providers/llmgateway-providers/models/azure/gpt-5.6-sol.toml
+++ b/providers/llmgateway-providers/models/azure/gpt-5.6-sol.toml
@@ -6,7 +6,7 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
diff --git a/providers/llmgateway-providers/models/openai/gpt-5.6-sol.toml b/providers/llmgateway-providers/models/openai/gpt-5.6-sol.toml
index 6ac7e092572..feddda8af87 100644
--- a/providers/llmgateway-providers/models/openai/gpt-5.6-sol.toml
+++ b/providers/llmgateway-providers/models/openai/gpt-5.6-sol.toml
@@ -6,7 +6,7 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 30
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5

From 8f9d1c3eb57b6f164a5f4a18541d7453162cd98a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 09:28:47 +0000
Subject: [PATCH 1586/2012] chore(sync): update Kilo model catalog (#7397)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 0b2d5078274..1a05f7dcedb 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.1485
+output = 0.594
+cache_read = 0.004455
 
 [limit]
 context = 1_048_576

From 1fb89299d8f08e49ae1324ea66ab8ad697709d22 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 09:29:02 +0000
Subject: [PATCH 1587/2012] chore(sync): update OpenRouter model catalog
 (#7399)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.2.toml               | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 95694d19041..426a4d8f486 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.4875
-output = 1.56
-cache_read = 0.091
+input = 0.5625
+output = 1.8
+cache_read = 0.105
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index bdcd94e72e3..07784758cd2 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.1485
+output = 0.594
+cache_read = 0.004455
 
 [limit]
 context = 1_048_576

From 1392fd983795c3d54a4c81594e7afbc3c185bb58 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 09:29:22 +0000
Subject: [PATCH 1588/2012] chore(sync): update LLM Gateway model catalog
 (#7398)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway-providers/models/aws-mantle/gpt-5.6-luna.toml    | 2 +-
 .../llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml     | 2 +-
 .../llmgateway-providers/models/aws-mantle/gpt-5.6-terra.toml   | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-luna.toml b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-luna.toml
index 14b7dc054c3..f1fab1a8efa 100644
--- a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-luna.toml
+++ b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-luna.toml
@@ -13,4 +13,4 @@ cache_read = 0.022
 cache_write = 0.275
 
 [limit]
-context = 278_528
+context = 921_600
diff --git a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
index 671621ddb8f..3c5f40f364f 100644
--- a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
+++ b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-sol.toml
@@ -13,4 +13,4 @@ cache_read = 0.44
 cache_write = 5.5
 
 [limit]
-context = 278_528
+context = 921_600
diff --git a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-terra.toml b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-terra.toml
index febed3b2f4d..66896653c6c 100644
--- a/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-terra.toml
+++ b/providers/llmgateway-providers/models/aws-mantle/gpt-5.6-terra.toml
@@ -13,4 +13,4 @@ cache_read = 0.22
 cache_write = 2.75
 
 [limit]
-context = 278_528
+context = 921_600

From 156818aa9bca4992fc03725e38aeb49ce63eebd8 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Fri, 18 Sep 2026 18:04:35 +0800
Subject: [PATCH 1589/2012] feat(opencode): end gpt-5.6-sol discount

---
 providers/opencode/models/gpt-5.6-sol.toml | 20 +++++++++-----------
 1 file changed, 9 insertions(+), 11 deletions(-)

diff --git a/providers/opencode/models/gpt-5.6-sol.toml b/providers/opencode/models/gpt-5.6-sol.toml
index 1aaeeb843ed..f122b60d0fe 100644
--- a/providers/opencode/models/gpt-5.6-sol.toml
+++ b/providers/opencode/models/gpt-5.6-sol.toml
@@ -1,22 +1,20 @@
-# Source: https://vercel.com/changelog/gpt-5-6-sol-is-50-off-on-ai-gateway-for-the-next-month
 # List pricing: https://developers.openai.com/api/docs/pricing
-# The promotion applies to all token types and long-context tiers through September 18, 2026.
 base_model = "openai/gpt-5.6-sol"
-name = "GPT-5.6 Sol (50% Off)"
+name = "GPT-5.6 Sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
 input = 4
-output = 15
+output = 20
 cache_read = 0.4
 cache_write = 5
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10
+
 [provider]
 npm = "@ai-sdk/openai"

From 9b1b6b09715e08f057104be1cd4c0892ff209901 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:26:35 +0000
Subject: [PATCH 1590/2012] chore(sync): update NanoGPT model catalog (#7402)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml    | 6 +++---
 .../nano-gpt/models/deepseek/deepseek-v4.1-flash.toml    | 2 +-
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml    | 2 +-
 providers/nano-gpt/models/unbiased/pareto.toml           | 9 +++++++++
 4 files changed, 14 insertions(+), 5 deletions(-)
 create mode 100644 providers/nano-gpt/models/unbiased/pareto.toml

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 4e65b119cdc..8e61a5cbf70 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -7,7 +7,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 67f17f156de..90c16dc648c 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -8,7 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.1
 output = 0.4
-cache_read = 0.006
+cache_read = 0.003
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/unbiased/pareto.toml b/providers/nano-gpt/models/unbiased/pareto.toml
new file mode 100644
index 00000000000..5f47e30ab94
--- /dev/null
+++ b/providers/nano-gpt/models/unbiased/pareto.toml
@@ -0,0 +1,9 @@
+base_model = "unbiased/pareto"
+
+[cost]
+input = 2.5
+output = 7.5
+cache_read = 0.25
+
+[limit]
+input = 262_144

From 6a5ec0027c89628cef591dc611d78a41fc11cd1a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:26:42 +0000
Subject: [PATCH 1591/2012] chore(sync): update Kilo model catalog (#7401)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 283ade8c995..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From bae926c752602fd6b5f1ba29081c26d792f433a9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:26:48 +0000
Subject: [PATCH 1592/2012] chore(sync): update OpenRouter model catalog
 (#7403)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 75641321a8f..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.96
-cache_read = 0.033
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576

From 1714693cda0b31846a1c61585066505c871cab3e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 11:24:41 +0000
Subject: [PATCH 1593/2012] chore(sync): update OpenRouter model catalog
 (#7408)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 07784758cd2..dc34ba0ccf7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1485
-output = 0.594
-cache_read = 0.004455
+input = 0.15
+output = 0.6
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From 4c2a945bb69df654bfa788994e6d15d85289e4ab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 11:24:59 +0000
Subject: [PATCH 1594/2012] chore(sync): update Kilo model catalog (#7409)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 1a05f7dcedb..673b730eb39 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1485
-output = 0.594
-cache_read = 0.004455
+input = 0.15
+output = 0.6
+cache_read = 0.015
 
 [limit]
 context = 1_048_576

From ae065079d839cd1e4d824007aa8acef87ff5465f Mon Sep 17 00:00:00 2001
From: 7Sageer <7sageer@djwcb.cn>
Date: Fri, 18 Sep 2026 19:32:13 +0800
Subject: [PATCH 1595/2012] feat(kimi-for-coding): add kimi-code-plan-global
 provider, switch both sites to OpenAI-compatible

- Add kimi-code-plan-global provider ("Kimi For Coding (kimi.ai)") for the
  global deployment at https://api.kimi.ai/coding/v1 with the same 4-model
  catalog (kimi-for-coding, kimi-for-coding-highspeed, k3, k3-256k)
- Rename kimi-for-coding to kimi-code-plan-cn ("Kimi For Coding (kimi.com)")
- Switch kimi-code-plan-cn (api.kimi.com) from @ai-sdk/anthropic to
  @ai-sdk/openai-compatible (/v1/chat/completions)
- Add interleaved reasoning_content to all models on both providers
---
 .../logo.svg                                  |  0
 .../models/k3-256k.toml                       |  3 ++
 .../models/k3.toml                            |  3 ++
 .../models/kimi-for-coding-highspeed.toml     |  3 ++
 .../models/kimi-for-coding.toml               |  5 ++++
 providers/kimi-code-plan-cn/provider.toml     | 10 +++++++
 providers/kimi-code-plan-global/logo.svg      |  3 ++
 .../kimi-code-plan-global/models/k3-256k.toml | 29 +++++++++++++++++++
 .../kimi-code-plan-global/models/k3.toml      | 23 +++++++++++++++
 .../models/kimi-for-coding-highspeed.toml     | 17 +++++++++++
 .../models/kimi-for-coding.toml               | 25 ++++++++++++++++
 providers/kimi-code-plan-global/provider.toml |  9 ++++++
 providers/kimi-for-coding/provider.toml       | 10 -------
 .../models/kimi-k3.toml                       |  2 +-
 14 files changed, 131 insertions(+), 11 deletions(-)
 rename providers/{kimi-for-coding => kimi-code-plan-cn}/logo.svg (100%)
 rename providers/{kimi-for-coding => kimi-code-plan-cn}/models/k3-256k.toml (93%)
 rename providers/{kimi-for-coding => kimi-code-plan-cn}/models/k3.toml (91%)
 rename providers/{kimi-for-coding => kimi-code-plan-cn}/models/kimi-for-coding-highspeed.toml (81%)
 rename providers/{kimi-for-coding => kimi-code-plan-cn}/models/kimi-for-coding.toml (73%)
 create mode 100644 providers/kimi-code-plan-cn/provider.toml
 create mode 100644 providers/kimi-code-plan-global/logo.svg
 create mode 100644 providers/kimi-code-plan-global/models/k3-256k.toml
 create mode 100644 providers/kimi-code-plan-global/models/k3.toml
 create mode 100644 providers/kimi-code-plan-global/models/kimi-for-coding-highspeed.toml
 create mode 100644 providers/kimi-code-plan-global/models/kimi-for-coding.toml
 create mode 100644 providers/kimi-code-plan-global/provider.toml
 delete mode 100644 providers/kimi-for-coding/provider.toml

diff --git a/providers/kimi-for-coding/logo.svg b/providers/kimi-code-plan-cn/logo.svg
similarity index 100%
rename from providers/kimi-for-coding/logo.svg
rename to providers/kimi-code-plan-cn/logo.svg
diff --git a/providers/kimi-for-coding/models/k3-256k.toml b/providers/kimi-code-plan-cn/models/k3-256k.toml
similarity index 93%
rename from providers/kimi-for-coding/models/k3-256k.toml
rename to providers/kimi-code-plan-cn/models/k3-256k.toml
index 2e7cd5d2d1f..3d9c4ba810e 100644
--- a/providers/kimi-for-coding/models/k3-256k.toml
+++ b/providers/kimi-code-plan-cn/models/k3-256k.toml
@@ -22,3 +22,6 @@ context = 262_144
 
 [modalities]
 input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-for-coding/models/k3.toml b/providers/kimi-code-plan-cn/models/k3.toml
similarity index 91%
rename from providers/kimi-for-coding/models/k3.toml
rename to providers/kimi-code-plan-cn/models/k3.toml
index dad0dadecd8..d0dbb42ade9 100644
--- a/providers/kimi-for-coding/models/k3.toml
+++ b/providers/kimi-code-plan-cn/models/k3.toml
@@ -16,3 +16,6 @@ input = 0
 output = 0
 cache_read = 0
 cache_write = 0
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-for-coding/models/kimi-for-coding-highspeed.toml b/providers/kimi-code-plan-cn/models/kimi-for-coding-highspeed.toml
similarity index 81%
rename from providers/kimi-for-coding/models/kimi-for-coding-highspeed.toml
rename to providers/kimi-code-plan-cn/models/kimi-for-coding-highspeed.toml
index 32d5102804a..407bcf710f8 100644
--- a/providers/kimi-for-coding/models/kimi-for-coding-highspeed.toml
+++ b/providers/kimi-code-plan-cn/models/kimi-for-coding-highspeed.toml
@@ -10,3 +10,6 @@ cache_write = 0
 
 [limit]
 output = 32_768
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-for-coding/models/kimi-for-coding.toml b/providers/kimi-code-plan-cn/models/kimi-for-coding.toml
similarity index 73%
rename from providers/kimi-for-coding/models/kimi-for-coding.toml
rename to providers/kimi-code-plan-cn/models/kimi-for-coding.toml
index 2dc81d0d56d..570fea3fd8a 100644
--- a/providers/kimi-for-coding/models/kimi-for-coding.toml
+++ b/providers/kimi-code-plan-cn/models/kimi-for-coding.toml
@@ -1,4 +1,6 @@
 # https://www.kimi.com/code/docs/en/third-party-tools/opencode.html
+# OpenAI-compatible chat completions on api.kimi.com (verified 2026-09-18 with
+# kimi-code OAuth credentials; reasoning returned via reasoning_content).
 # Toggle: thinking.type = "enabled" | "disabled" | "adaptive"
 # Effort: output_config.effort = "low" | "high" | "max" (default: "max")
 # Retain the existing output limit; the K2.8 announcement only specifies context.
@@ -20,3 +22,6 @@ cache_write = 0
 
 [limit]
 output = 32_768
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-code-plan-cn/provider.toml b/providers/kimi-code-plan-cn/provider.toml
new file mode 100644
index 00000000000..ab7b8962d83
--- /dev/null
+++ b/providers/kimi-code-plan-cn/provider.toml
@@ -0,0 +1,10 @@
+# NOTE: api.kimi.com/coding serves BOTH Anthropic (/v1/messages) and
+# OpenAI-compatible (/v1/chat/completions) protocols. All four models
+# verified on /chat/completions 2026-09-18 with kimi-code OAuth credentials
+# (reasoning returned via reasoning_content). The global deployment
+# (api.kimi.ai) is the separate kimi-code-plan-global provider.
+name = "Kimi For Coding (kimi.com)"
+env = ["KIMI_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+doc = "https://www.kimi.com/code/docs/en/kimi-code/models.html"
+api = "https://api.kimi.com/coding/v1"
diff --git a/providers/kimi-code-plan-global/logo.svg b/providers/kimi-code-plan-global/logo.svg
new file mode 100644
index 00000000000..d41029890d2
--- /dev/null
+++ b/providers/kimi-code-plan-global/logo.svg
@@ -0,0 +1,3 @@
+<svg viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path d="M3.51531 15.81L10.908 17.7878C10.8981 18.3134 10.9136 18.8392 10.9545 19.3633L15.5704 20.5979C14.2013 21.163 12.7171 21.3928 11.2413 21.2683L11.1018 21.2559L11.0677 21.2528L11.0026 21.2458L10.9297 21.238C10.8891 21.2332 10.8486 21.228 10.808 21.2225L10.7251 21.2117L10.6399 21.1993C10.557 21.1872 10.4743 21.1741 10.3919 21.1598L10.3593 21.1536L10.3012 21.1435L10.2183 21.128L10.164 21.1164L10.0919 21.1016L10.0338 21.0892L9.96019 21.0737L9.88501 21.0559L9.81216 21.0389L9.75946 21.0257L9.69126 21.0086L9.62151 20.99L9.54789 20.9707L9.48434 20.9528L9.39986 20.9296L9.35181 20.9141L9.28671 20.8947L9.21464 20.873L9.13326 20.8467L9.08831 20.8319L9.02631 20.8118L8.95656 20.7878L8.90541 20.7692C8.89403 20.7653 8.88267 20.7615 8.87131 20.7575L8.81861 20.7382L8.74034 20.7095L8.69617 20.6924L8.63417 20.6692L8.56674 20.6421L8.49854 20.6149L8.43732 20.5901L8.36369 20.5591L8.31487 20.5374L8.26604 20.5165C8.25543 20.5119 8.24484 20.5073 8.23427 20.5026L8.18312 20.4793L8.10329 20.4429L8.06299 20.4243L7.98859 20.3886L7.94054 20.3654L7.87544 20.3344L7.80879 20.3003L7.73672 20.2639L7.69642 20.2429L7.61659 20.2003L7.57242 20.1771L7.52747 20.1523C7.51556 20.1456 7.50367 20.1389 7.49182 20.1321L7.41897 20.091L7.37247 20.0647L7.33294 20.0414L7.27714 20.0097L7.21359 19.9709L7.14152 19.9275L7.10122 19.9027L7.03612 19.8616L6.98885 19.8314L6.92762 19.7927L6.87337 19.7562L6.8323 19.7291C6.81831 19.7199 6.80436 19.7106 6.79045 19.7012L6.75635 19.678L6.72225 19.6547C6.71188 19.6475 6.70155 19.6403 6.69125 19.633L6.64707 19.602L6.58817 19.5602L6.5347 19.5214L6.47735 19.4796L6.43395 19.447L6.37505 19.4028L6.31615 19.3571L6.2495 19.3052L6.21462 19.2781L6.16502 19.2378L6.10767 19.1913L6.0387 19.1347L6.00305 19.1045L5.9674 19.0742C5.95625 19.0647 5.94515 19.0552 5.93407 19.0456L5.8992 19.0146L5.85192 18.9735L5.79767 18.9255L5.74497 18.8789L5.69692 18.834L5.645 18.786L5.60392 18.7472L5.53572 18.6821C5.51002 18.6572 5.48444 18.6321 5.459 18.6069L5.43653 18.5852L5.40475 18.5527L5.35128 18.4984L5.31253 18.4589L5.27378 18.4178C5.2295 18.3724 5.18609 18.3262 5.14358 18.2791L5.08158 18.2109L5.03353 18.1567L4.9785 18.0947L4.94595 18.0567L4.90488 18.0086L4.85993 17.9559L4.82428 17.9125C4.81728 17.904 4.8103 17.8955 4.80335 17.887L4.76848 17.8443L4.71733 17.7808L4.68555 17.7405L4.6468 17.6909L4.6313 17.6715C4.18786 17.0967 3.81331 16.4719 3.51531 15.81ZM2.72481 11.3196L11.5234 13.6733C11.3761 14.1864 11.2547 14.7065 11.1599 15.2318L19.543 17.4747C19.1266 18.0449 18.6469 18.5662 18.1132 19.0285L3.20918 15.0404L3.19678 15.0047L3.16966 14.9241C3.15647 14.8847 3.14356 14.8452 3.13091 14.8056L3.12548 14.7877C3.06583 14.5981 3.01233 14.4066 2.96506 14.2135L2.94181 14.1158L2.92786 14.0538L2.91158 13.9786L2.89763 13.9159L2.88368 13.8461L2.87051 13.781L2.85656 13.7082C2.83641 13.5989 2.81781 13.4888 2.80153 13.378L2.78836 13.2866L2.77983 13.2222L2.76976 13.1432C2.76457 13.1016 2.75966 13.06 2.75503 13.0184L2.75116 12.982C2.69326 12.4297 2.68444 11.8734 2.72481 11.3196ZM3.95938 7.32839L13.2191 9.80528C12.9339 10.2742 12.6727 10.7593 12.4363 11.2584L21.1899 13.6004C21.0799 14.2359 20.9047 14.8497 20.6722 15.4333L11.721 13.0386L2.79611 10.6516L2.80773 10.5741L2.81393 10.5361L2.82168 10.4842L2.83331 10.4168L2.84726 10.3408C2.86741 10.2261 2.89066 10.1122 2.91546 9.99825L2.93716 9.90215L2.95266 9.83628L2.97126 9.7611C2.98831 9.69135 3.00613 9.6216 3.02551 9.5534L3.04721 9.47435L3.06503 9.41003L3.08828 9.33253L3.10766 9.26898L3.13091 9.19458L3.15106 9.13103L3.17508 9.05741C3.37577 8.45541 3.63808 7.87573 3.95783 7.32761L3.95938 7.32839ZM7.40192 3.9184L16.1478 6.25734C15.6862 6.67623 15.2494 7.12157 14.8396 7.59111L20.9024 9.21318C21.1093 9.87348 21.2449 10.5648 21.3 11.2785L4.33215 6.74016L4.36703 6.68979L4.38795 6.65879L4.41895 6.61616L4.4546 6.56579L4.49723 6.50689L4.53908 6.45109L4.58868 6.38444L4.62743 6.33406L4.6716 6.27749L4.71423 6.22324L4.76073 6.16589L4.80335 6.11241L4.85373 6.05274L4.89558 6.00159L4.94673 5.94192L4.9878 5.89542L5.0436 5.83187L5.08468 5.78537L5.1366 5.72802L5.17845 5.68307L5.23503 5.62262L5.27998 5.57612L5.3288 5.52419L5.459 5.39244L5.5365 5.31649L5.58222 5.27309L5.64112 5.21807C6.17525 4.71752 6.76621 4.28132 7.40192 3.9184ZM12.0132 2.7001H12.0883L12.1519 2.70088L12.2054 2.70165L12.2472 2.7032L12.2999 2.70475L12.3356 2.70553L12.3945 2.70785L12.4309 2.7094L12.4774 2.71173L12.5192 2.71328L12.5867 2.71715L12.668 2.72258L12.7796 2.7311L12.8478 2.73653L12.8819 2.73963L12.9416 2.74583L13.0052 2.75203L13.0416 2.7559L13.1206 2.7652L13.1594 2.76985L13.2431 2.7807L13.3059 2.78845L13.3384 2.7931L13.3888 2.80085L13.5492 2.82565L13.6035 2.83495L13.6538 2.84348L13.7623 2.86363L13.8336 2.87758L13.9189 2.89462L13.9545 2.90237L14.0126 2.91477L14.0444 2.92252L14.0925 2.9326L14.125 2.94035L14.1754 2.95197L14.2134 2.96127L14.2684 2.97445L14.3428 2.99305L14.4296 3.0163L14.5172 3.03955L14.6047 3.06435L14.6435 3.07597L14.6977 3.09147L14.7582 3.11007L14.8148 3.1279L14.8535 3.1403L14.8923 3.1527L14.9512 3.17207L15.0279 3.19765L15.1069 3.22555L15.1441 3.23872L15.1937 3.25655L15.2658 3.2829L15.3511 3.31467L15.441 3.34955L15.5185 3.38055L15.5549 3.39605L15.6014 3.41465L15.6332 3.4286L15.682 3.44875L15.713 3.4627L15.7572 3.48207L15.8424 3.51927L15.9199 3.55492L15.9773 3.58205L16.0354 3.60995L16.0819 3.63165L16.1532 3.6673L16.2237 3.70217L16.3028 3.74247L16.3438 3.76417L16.3818 3.78432L16.4175 3.80292L16.464 3.8285L16.4957 3.84555L16.536 3.86802L16.6042 3.90677L16.6864 3.95327L16.7538 3.9928L16.798 4.01915L16.8391 4.04395L16.9135 4.08967L16.9817 4.1323L17.0576 4.18035L17.0855 4.19895L17.1351 4.23072L17.2002 4.27412L17.2312 4.29505L17.2793 4.3276L17.3273 4.36092L17.3451 4.3741C17.387 4.40277 17.4288 4.43222 17.4699 4.46244L17.5342 4.50894L17.5846 4.54614L17.628 4.57947L17.6947 4.62984L17.7582 4.67944L17.7892 4.70269L17.828 4.73447L17.8946 4.78794L17.9558 4.83832L18.0217 4.89334C18.5735 5.35834 19.0703 5.88767 19.5012 6.46736L8.29704 3.47045L8.34509 3.44952L8.39547 3.42782L8.45824 3.40147L8.52489 3.37435C8.61247 3.33947 8.70082 3.3046 8.78916 3.27282L8.86356 3.2457L8.93564 3.22012L9.00074 3.19687L9.07514 3.17285C9.14256 3.1496 9.21154 3.1279 9.27974 3.10697L9.35026 3.08605L9.41691 3.06667L9.49596 3.04342L9.56184 3.0256L9.63934 3.00545L9.70599 2.98685L9.77574 2.96902L9.84626 2.95197L9.91989 2.93492L9.98964 2.91942L10.0656 2.90315L10.1361 2.88765L10.2097 2.8737L10.281 2.85975L10.3585 2.8458L10.4291 2.8334L10.505 2.82023L10.5763 2.80938L10.6515 2.79775L10.7228 2.78768L10.8018 2.7776L10.8724 2.7683L10.9537 2.759L11.0235 2.75125L11.1049 2.7435C11.1769 2.73575 11.249 2.72955 11.3219 2.7249L11.404 2.7187L11.4738 2.71483L11.559 2.71018L11.6311 2.70708L11.7086 2.70398L11.7838 2.70243L11.8605 2.70088L12.0132 2.69933V2.7001Z" fill="currentColor"/>
+</svg>
diff --git a/providers/kimi-code-plan-global/models/k3-256k.toml b/providers/kimi-code-plan-global/models/k3-256k.toml
new file mode 100644
index 00000000000..36af21989bd
--- /dev/null
+++ b/providers/kimi-code-plan-global/models/k3-256k.toml
@@ -0,0 +1,29 @@
+# Verified 2026-09-18 against https://api.kimi.ai/coding/v1/chat/completions
+# with kimi-code OAuth credentials; reasoning returned via reasoning_content.
+# k3-256k is the 256K-context variant of Kimi K3 on the Kimi For Coding
+# endpoint. Unlike full K3 it accepts no video input (image only).
+# reasoning_options per the Kimi Code docs:
+#   reasoning_effort = "low" | "high" | "max" (default "high")
+# Cost is zeroed per this provider's subscription convention.
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3-256K"
+description = "256K-context version of Kimi K3, reducing token consumption for shorter coding sessions"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-code-plan-global/models/k3.toml b/providers/kimi-code-plan-global/models/k3.toml
new file mode 100644
index 00000000000..e2cc98df82b
--- /dev/null
+++ b/providers/kimi-code-plan-global/models/k3.toml
@@ -0,0 +1,23 @@
+# Verified 2026-09-18 against https://api.kimi.ai/coding/v1/chat/completions
+# with kimi-code OAuth credentials; reasoning returned via reasoning_content.
+# reasoning_options mirror the Moonshot AI platform API surface:
+#   thinking.type = "enabled" | "disabled" | "adaptive"
+#   output_config.effort = "low" | "high" | "max"
+# Cost is zeroed per this provider's subscription convention.
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-code-plan-global/models/kimi-for-coding-highspeed.toml b/providers/kimi-code-plan-global/models/kimi-for-coding-highspeed.toml
new file mode 100644
index 00000000000..ff4e80c9104
--- /dev/null
+++ b/providers/kimi-code-plan-global/models/kimi-for-coding-highspeed.toml
@@ -0,0 +1,17 @@
+# Verified 2026-09-18 against https://api.kimi.ai/coding/v1/chat/completions
+# with kimi-code OAuth credentials; reasoning returned via reasoning_content.
+base_model = "moonshotai/kimi-k2.7-code-highspeed"
+name = "Kimi For Coding HighSpeed"
+reasoning_options = []
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+output = 32_768
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-code-plan-global/models/kimi-for-coding.toml b/providers/kimi-code-plan-global/models/kimi-for-coding.toml
new file mode 100644
index 00000000000..766cc5bfb99
--- /dev/null
+++ b/providers/kimi-code-plan-global/models/kimi-for-coding.toml
@@ -0,0 +1,25 @@
+# Verified 2026-09-18 against https://api.kimi.ai/coding/v1/chat/completions
+# with kimi-code OAuth credentials; reasoning returned via reasoning_content.
+# Toggle: thinking.type = "enabled" | "disabled" | "adaptive"
+# Effort: output_config.effort = "low" | "high" | "max" (default: "max")
+base_model = "moonshotai/kimi-k2.8-preview"
+name = "kimi-for-coding"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+output = 32_768
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/kimi-code-plan-global/provider.toml b/providers/kimi-code-plan-global/provider.toml
new file mode 100644
index 00000000000..cf4da1d06d8
--- /dev/null
+++ b/providers/kimi-code-plan-global/provider.toml
@@ -0,0 +1,9 @@
+# Global deployment of Kimi For Coding on api.kimi.ai (separate host and OAuth
+# issuer auth.kimi.ai). OpenAI-compatible chat completions; catalog verified
+# via GET /models on 2026-09-18 (kimi-for-coding, kimi-for-coding-highspeed,
+# k3, k3-256k) with kimi-code OAuth credentials.
+name = "Kimi For Coding (kimi.ai)"
+env = ["KIMI_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+doc = "https://www.kimi.ai/code/docs/en/kimi-code/models.html"
+api = "https://api.kimi.ai/coding/v1"
diff --git a/providers/kimi-for-coding/provider.toml b/providers/kimi-for-coding/provider.toml
deleted file mode 100644
index f8d71f33de3..00000000000
--- a/providers/kimi-for-coding/provider.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-# NOTE: api.kimi.com/coding serves BOTH Anthropic (/v1/messages) and
-# OpenAI-compatible (/v1/chat/completions) protocols (verified 2026-07).
-# Registered as @ai-sdk/anthropic: the Messages surface is the officially
-# documented one (see doc) and round-trips reasoning natively via
-# thinking blocks.
-name = "Kimi For Coding"
-env = ["KIMI_API_KEY"]
-npm = "@ai-sdk/anthropic"
-doc = "https://www.kimi.com/code/docs/en/kimi-code/models.html"
-api = "https://api.kimi.com/coding/v1"
diff --git a/providers/volcengine-coding-plan/models/kimi-k3.toml b/providers/volcengine-coding-plan/models/kimi-k3.toml
index 18f097450bd..7b7fb0c9e7e 100644
--- a/providers/volcengine-coding-plan/models/kimi-k3.toml
+++ b/providers/volcengine-coding-plan/models/kimi-k3.toml
@@ -2,7 +2,7 @@
 # (accessed 2026-09-11), listed alongside kimi-k2.7-code. Subscription tier:
 # no per-token price.
 # reasoning_options mirror the lab (providers/moonshotai/models/kimi-k3.toml)
-# and the kimi-for-coding relay peer. Wire fields on this provider's OpenAI
+# and the kimi-code-plan-cn relay peer. Wire fields on this provider's OpenAI
 # path (provider.toml): POST /api/coding/v3/chat/completions takes
 # thinking.type = enabled|disabled plus reasoning_effort = low|high|max
 # (`adaptive` is the Moonshot lab surface, not accepted here); the separate

From 37482b15ca21f840961b770c0a6c0317de1e8345 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 13:25:43 +0000
Subject: [PATCH 1596/2012] chore(sync): update Kilo model catalog (#7411)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 +
 providers/kilo/models/moonshotai/kimi-k3.toml          |  6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                |  1 -
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     |  8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  2 +-
 providers/kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 7 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 3af9cc14392..55b1dc1c908 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0705
+output = 11.5948
+cache_read = 0.240178
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index f8addeabfc8..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -12,4 +12,3 @@ cache_read = 0.26
 
 [limit]
 context = 1_048_576
-output = 163_840
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..957bc0a82fa 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 396a029912a..26d07e5604f 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -17,7 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.0558
 output = 0.1767
-cache_read = 0.0088
+cache_read = 0.0102
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..2fbb91bbcaa 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0705
+output = 11.5948
+cache_read = 0.240178
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 38aa9fed430..2fc8e4f5ea7 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.9
+output = 3
+cache_read = 0.15
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_000_000
+output = 131_072
 
 [modalities]
 input = ["text"]

From 119bb5a58314f25ba8e1d8bf877cceb4716aaa7d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 13:25:50 +0000
Subject: [PATCH 1597/2012] chore(sync): update OpenRouter model catalog
 (#7412)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 providers/openrouter/models/moonshotai/kimi-k3.toml       | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 7 +++----
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 2 +-
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 8 files changed, 25 insertions(+), 25 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..141d4b41ca6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.049
+output = 0.098
+cache_read = 0.0098
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c307f78addc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 618c6fb2bbd..4bf49607f67 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0705
+output = 11.5948
+cache_read = 0.240178
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 426a4d8f486..4fa1fa642a3 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.5625
-output = 1.8
-cache_read = 0.105
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_048_576
-output = 163_840
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..50d7ce57993 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65868
+output = 1.97604
+cache_read = 0.020958
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 7356cb84c22..ea3e3a2af96 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.0558
 output = 0.1767
-cache_read = 0.0088
+cache_read = 0.0102
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..2ca20af632f 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0705
+output = 11.5948
+cache_read = 0.240178
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 3e3ce108967..0a7b3acfd67 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8775
-output = 2.97
-cache_read = 0.1755
+input = 0.9
+output = 3
+cache_read = 0.15
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 131_072
 
 [modalities]
 input = ["text"]

From c7777f749524662fd06478554eb17922535d7c89 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 14:27:20 +0000
Subject: [PATCH 1598/2012] chore(sync): update OpenRouter model catalog
 (#7413)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/moonshotai/kimi-k3.toml       | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 6 +++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 8 files changed, 23 insertions(+), 24 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 141d4b41ca6..a9469f150d2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.049
-output = 0.098
-cache_read = 0.0098
+input = 0.04984
+output = 0.09968
+cache_read = 0.009968
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c307f78addc..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 4bf49607f67..618c6fb2bbd 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0705
-output = 11.5948
-cache_read = 0.240178
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 4fa1fa642a3..12ed56b2bd6 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.5614
-output = 1.7644
-cache_read = 0.10426
+input = 0.5544
+output = 1.7424
+cache_read = 0.10296
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index dc34ba0ccf7..98fd23a037b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -26,7 +26,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 50d7ce57993..e0b3ef90d2e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index ea3e3a2af96..b207a5a59a3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.0558
-output = 0.1767
-cache_read = 0.0102
+input = 0.055
+output = 0.165
+cache_read = 0.00175
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 2ca20af632f..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0705
-output = 11.5948
-cache_read = 0.240178
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 268d33960435f9214a292630e926fe3f6a542b75 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 14:27:23 +0000
Subject: [PATCH 1599/2012] chore(sync): update Kilo model catalog (#7415)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 -
 providers/kilo/models/moonshotai/kimi-k3.toml          |  6 +++---
 providers/kilo/models/z-ai/glm-5.2.toml                |  3 ++-
 .../kilo/models/~deepseek/deepseek-flash-latest.toml   |  4 ++--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     |  8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     | 10 +++++-----
 providers/kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 7 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 55b1dc1c908..3af9cc14392 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0705
-output = 11.5948
-cache_read = 0.240178
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index bb89edf75cf..dcb3cc80d7f 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,4 +11,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_024_000
+output = 128_000
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 673b730eb39..f491c2853c8 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -20,8 +20,8 @@ output = 0.6
 cache_read = 0.015
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_000_000
+output = 393_216
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 957bc0a82fa..bd4537b7471 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.65868
-output = 1.97604
-cache_read = 0.020958
+input = 0.5808
+output = 1.7424
+cache_read = 0.05808
 
 [limit]
-context = 1_048_576
+context = 1_000_000
 output = 393_216
 
 [modalities]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 26d07e5604f..844f1914d6e 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.0558
-output = 0.1767
-cache_read = 0.0102
+input = 0.055
+output = 0.165
+cache_read = 0.00175
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 2fbb91bbcaa..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.0705
-output = 11.5948
-cache_read = 0.240178
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 19a6cfba6cc3ef772309e09a80e6c20d1a2606ec Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 14:27:27 +0000
Subject: [PATCH 1600/2012] chore(sync): update OVHcloud AI Endpoints model
 catalog (#7414)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/ovhcloud/models/qwen3-32b.toml | 26 ------------------------
 1 file changed, 26 deletions(-)
 delete mode 100644 providers/ovhcloud/models/qwen3-32b.toml

diff --git a/providers/ovhcloud/models/qwen3-32b.toml b/providers/ovhcloud/models/qwen3-32b.toml
deleted file mode 100644
index b117c1bb3b6..00000000000
--- a/providers/ovhcloud/models/qwen3-32b.toml
+++ /dev/null
@@ -1,26 +0,0 @@
-# Put `/no_think` in prompt content to disable the model's default reasoning.
-# https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog/qwen-3-32b/
-
-name = "Qwen3-32B"
-description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
-release_date = "2025-07-16"
-last_updated = "2025-07-16"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.09
-output = 0.25
-
-[limit]
-context = 32_768
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 511b7612e9b764df50381aff89c239800b6421ad Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Fri, 18 Sep 2026 23:09:47 +0800
Subject: [PATCH 1601/2012] feat(aihubmix): add claude-fable-5-1 (#7416)

* feat(aihubmix): add claude-fable-5-1

AIHubMix serves Claude Fable 5.1 but the catalog only had Fable 5. The
numbers come from the provider's public model listing; limit, modalities
and the rest inherit from the Anthropic base entry.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* docs(aihubmix): cite the toggle and effort wire paths

AGENTS.md requires a leading top-of-file comment with the exact request
field for every toggle; the header only cited pricing.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 .../aihubmix/models/claude-fable-5-1.toml      | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/aihubmix/models/claude-fable-5-1.toml

diff --git a/providers/aihubmix/models/claude-fable-5-1.toml b/providers/aihubmix/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..b29ef102523
--- /dev/null
+++ b/providers/aihubmix/models/claude-fable-5-1.toml
@@ -0,0 +1,18 @@
+# Toggle: $.thinking.type = "enabled"|"disabled"|"adaptive" on the Anthropic-compatible
+# /v1/messages path; $.enable_thinking = true|false on /v1/chat/completions.
+# Effort: $.output_config.effort on /v1/messages; $.reasoning_effort on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api-reference/anthropic-compatible/create-a-message
+# Pricing, context window, max output and the effort levels come from
+# https://aihubmix.com/api/v1/models?type=llm (accessed 2026-09-18), which lists this
+# model on both faces. Cache read is a quarter of Claude Fable 5's rate; input, output
+# and cache write are unchanged from it, matching the sibling entry in this directory.
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+interleaved = true
+
+[cost]
+input = 11
+output = 55
+cache_read = 0.275
+cache_write = 13.75

From be7ef0c6b57f85e67f68be9a720b23997afa5622 Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Fri, 18 Sep 2026 23:10:46 +0800
Subject: [PATCH 1602/2012] feat(vivgrid): add claude-opus-5, claude-sonnet-5,
 jev; use anthropic sdk for claude models (#7405)

* feat(vivgrid): add claude-opus-5, claude-sonnet-5; use anthropic sdk for claude models

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* feat(vivgrid): add jev; add doc source links to model files

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/vivgrid/models/claude-fable-5-1.toml  |  5 ++---
 providers/vivgrid/models/claude-fable-5.toml    |  5 ++---
 providers/vivgrid/models/claude-opus-5.toml     | 14 ++++++++++++++
 providers/vivgrid/models/claude-sonnet-5.toml   | 16 ++++++++++++++++
 providers/vivgrid/models/deepseek-v4-flash.toml |  2 ++
 providers/vivgrid/models/glm-5.2.toml           |  2 ++
 providers/vivgrid/models/glm-5.3.toml           |  2 ++
 providers/vivgrid/models/gpt-5.1-codex-max.toml |  2 ++
 providers/vivgrid/models/gpt-5.1-codex.toml     |  2 ++
 providers/vivgrid/models/gpt-5.2-codex.toml     |  2 ++
 providers/vivgrid/models/gpt-5.3-codex.toml     |  2 ++
 providers/vivgrid/models/gpt-5.6-luna.toml      |  2 ++
 providers/vivgrid/models/gpt-5.6-sol.toml       |  2 ++
 providers/vivgrid/models/gpt-5.6-terra.toml     |  2 ++
 providers/vivgrid/models/gpt-6-astra.toml       |  2 ++
 providers/vivgrid/models/jev.toml               |  7 +++++++
 providers/vivgrid/models/kimi-k3.toml           |  2 ++
 17 files changed, 65 insertions(+), 6 deletions(-)
 create mode 100644 providers/vivgrid/models/claude-opus-5.toml
 create mode 100644 providers/vivgrid/models/claude-sonnet-5.toml
 create mode 100644 providers/vivgrid/models/jev.toml

diff --git a/providers/vivgrid/models/claude-fable-5-1.toml b/providers/vivgrid/models/claude-fable-5-1.toml
index 471e3cb571d..756c74d20a3 100644
--- a/providers/vivgrid/models/claude-fable-5-1.toml
+++ b/providers/vivgrid/models/claude-fable-5-1.toml
@@ -1,7 +1,6 @@
 # Pricing, context window and max output:
 # https://docs.vivgrid.com/models/claude-fable-5-1 (accessed 2026-09-07)
-# Cached input is $0.50/MTok. Vivgrid publishes no cache-write rate and states
-# pricing matches the original provider, so Anthropic's $12.50/MTok applies.
+# Effort: output_config.effort = low|medium|high|xhigh|max
 base_model = "anthropic/claude-fable-5-1"
 structured_output = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
@@ -13,4 +12,4 @@ cache_read = 0.5
 cache_write = 12.5
 
 [provider]
-npm = "@ai-sdk/openai-compatible"
\ No newline at end of file
+npm = "@ai-sdk/anthropic"
diff --git a/providers/vivgrid/models/claude-fable-5.toml b/providers/vivgrid/models/claude-fable-5.toml
index 912e2ff029a..31f1362efa1 100644
--- a/providers/vivgrid/models/claude-fable-5.toml
+++ b/providers/vivgrid/models/claude-fable-5.toml
@@ -1,7 +1,6 @@
 # Pricing, context window and max output:
 # https://docs.vivgrid.com/models/claude-fable-5 (accessed 2026-09-07)
-# Cached input is $1.25/MTok. Vivgrid publishes no cache-write rate and states
-# pricing matches the original provider, so Anthropic's $12.50/MTok applies.
+# Effort: output_config.effort = low|medium|high|xhigh|max
 base_model = "anthropic/claude-fable-5"
 structured_output = true
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
@@ -13,4 +12,4 @@ cache_read = 1.25
 cache_write = 12.5
 
 [provider]
-npm = "@ai-sdk/openai-compatible"
\ No newline at end of file
+npm = "@ai-sdk/anthropic"
diff --git a/providers/vivgrid/models/claude-opus-5.toml b/providers/vivgrid/models/claude-opus-5.toml
new file mode 100644
index 00000000000..fbc0e693efc
--- /dev/null
+++ b/providers/vivgrid/models/claude-opus-5.toml
@@ -0,0 +1,14 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/claude-opus-5 (accessed 2026-09-07)
+base_model = "anthropic/claude-opus-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5.0
+output = 25.0
+cache_read = 0.50
+cache_write = 6.25
+
+[provider]
+npm = "@ai-sdk/anthropic"
diff --git a/providers/vivgrid/models/claude-sonnet-5.toml b/providers/vivgrid/models/claude-sonnet-5.toml
new file mode 100644
index 00000000000..44da13b30e5
--- /dev/null
+++ b/providers/vivgrid/models/claude-sonnet-5.toml
@@ -0,0 +1,16 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/claude-sonnet-5 (accessed 2026-09-07)
+# Toggle: thinking.type = adaptive|disabled
+# Effort: output_config.effort = low|medium|high|xhigh|max
+base_model = "anthropic/claude-sonnet-5"
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+structured_output = true
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file
diff --git a/providers/vivgrid/models/deepseek-v4-flash.toml b/providers/vivgrid/models/deepseek-v4-flash.toml
index 38f0598e8fe..464e7b70d2d 100644
--- a/providers/vivgrid/models/deepseek-v4-flash.toml
+++ b/providers/vivgrid/models/deepseek-v4-flash.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/deepseek-v4-flash (accessed 2026-09-18)
 base_model = "deepseek/deepseek-v4-flash-0731"
 name = "DeepSeek V4 Flash"
 
diff --git a/providers/vivgrid/models/glm-5.2.toml b/providers/vivgrid/models/glm-5.2.toml
index ad841133eeb..2ba2cd2bf76 100644
--- a/providers/vivgrid/models/glm-5.2.toml
+++ b/providers/vivgrid/models/glm-5.2.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/glm-5.2 (accessed 2026-09-18)
 base_model = "zhipuai/glm-5.2"
 
 [[reasoning_options]]
diff --git a/providers/vivgrid/models/glm-5.3.toml b/providers/vivgrid/models/glm-5.3.toml
index 704b2199e03..da960ec71fa 100644
--- a/providers/vivgrid/models/glm-5.3.toml
+++ b/providers/vivgrid/models/glm-5.3.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/glm-5.3 (accessed 2026-09-18)
 base_model = "zhipuai/glm-5.3"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
diff --git a/providers/vivgrid/models/gpt-5.1-codex-max.toml b/providers/vivgrid/models/gpt-5.1-codex-max.toml
index b96035a695e..68ec50cbee8 100644
--- a/providers/vivgrid/models/gpt-5.1-codex-max.toml
+++ b/providers/vivgrid/models/gpt-5.1-codex-max.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.1-codex-max (accessed 2026-09-18)
 name = "GPT-5.1 Codex Max"
 description = "Coding-optimized GPT model for repository edits, reviews, and agentic software work"
 family = "gpt-codex"
diff --git a/providers/vivgrid/models/gpt-5.1-codex.toml b/providers/vivgrid/models/gpt-5.1-codex.toml
index c8b1158f20c..f6e6f2fa8ee 100644
--- a/providers/vivgrid/models/gpt-5.1-codex.toml
+++ b/providers/vivgrid/models/gpt-5.1-codex.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.1-codex (accessed 2026-09-18)
 name = "GPT-5.1 Codex"
 description = "Coding-optimized GPT model for repository edits, reviews, and agentic software work"
 family = "gpt-codex"
diff --git a/providers/vivgrid/models/gpt-5.2-codex.toml b/providers/vivgrid/models/gpt-5.2-codex.toml
index e28e266ee10..7319407808d 100644
--- a/providers/vivgrid/models/gpt-5.2-codex.toml
+++ b/providers/vivgrid/models/gpt-5.2-codex.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.2-codex (accessed 2026-09-18)
 name = "GPT-5.2 Codex"
 description = "Coding-optimized GPT model for repository edits, reviews, and agentic software work"
 family = "gpt-codex"
diff --git a/providers/vivgrid/models/gpt-5.3-codex.toml b/providers/vivgrid/models/gpt-5.3-codex.toml
index 715972403fe..454bc79f662 100644
--- a/providers/vivgrid/models/gpt-5.3-codex.toml
+++ b/providers/vivgrid/models/gpt-5.3-codex.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.3-codex (accessed 2026-09-18)
 name = "GPT-5.3 Codex"
 description = "Coding-optimized GPT model for repository edits, reviews, and agentic software work"
 family = "gpt-codex"
diff --git a/providers/vivgrid/models/gpt-5.6-luna.toml b/providers/vivgrid/models/gpt-5.6-luna.toml
index fb81c98d331..9c4852173d2 100644
--- a/providers/vivgrid/models/gpt-5.6-luna.toml
+++ b/providers/vivgrid/models/gpt-5.6-luna.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.6-luna (accessed 2026-09-18)
 base_model = "openai/gpt-5.6-luna"
 name = "GPT 5.6 Luna"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
diff --git a/providers/vivgrid/models/gpt-5.6-sol.toml b/providers/vivgrid/models/gpt-5.6-sol.toml
index df3250cf9b6..65eb349ce72 100644
--- a/providers/vivgrid/models/gpt-5.6-sol.toml
+++ b/providers/vivgrid/models/gpt-5.6-sol.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.6-sol (accessed 2026-09-18)
 base_model = "openai/gpt-5.6-sol"
 name = "GPT 5.6 Sol"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
diff --git a/providers/vivgrid/models/gpt-5.6-terra.toml b/providers/vivgrid/models/gpt-5.6-terra.toml
index 385029c25bd..371995729d7 100644
--- a/providers/vivgrid/models/gpt-5.6-terra.toml
+++ b/providers/vivgrid/models/gpt-5.6-terra.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-5.6-terra (accessed 2026-09-18)
 base_model = "openai/gpt-5.6-terra"
 name = "GPT 5.6 Terra"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
diff --git a/providers/vivgrid/models/gpt-6-astra.toml b/providers/vivgrid/models/gpt-6-astra.toml
index afdc38e7f28..2a246582bfd 100644
--- a/providers/vivgrid/models/gpt-6-astra.toml
+++ b/providers/vivgrid/models/gpt-6-astra.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/gpt-6-astra (accessed 2026-09-18)
 base_model = "openai/gpt-6-astra"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
diff --git a/providers/vivgrid/models/jev.toml b/providers/vivgrid/models/jev.toml
new file mode 100644
index 00000000000..f499745e785
--- /dev/null
+++ b/providers/vivgrid/models/jev.toml
@@ -0,0 +1,7 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/jev
+base_model = "typesafe/jev-latest"
+
+[cost]
+input = 0.042
+output = 0
diff --git a/providers/vivgrid/models/kimi-k3.toml b/providers/vivgrid/models/kimi-k3.toml
index 7e11b4c39c2..1ca4fad3636 100644
--- a/providers/vivgrid/models/kimi-k3.toml
+++ b/providers/vivgrid/models/kimi-k3.toml
@@ -1,3 +1,5 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/kimi-k3 (accessed 2026-09-18)
 base_model = "moonshotai/kimi-k3"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 reasoning_options = []

From a72e9666472c91973fcdc97d808029c2a60f4274 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:10:56 -0500
Subject: [PATCH 1603/2012] fix: [DeepSeek Lab] The weights of DeepSeek V4
 Flash Vision Exp and DeepSeek V4.1 Flash are available on Hugging Face
 (#7407)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 models/deepseek/deepseek-v4-flash-vision-exp.toml | 10 ++++++++--
 models/deepseek/deepseek-v4.1-flash.toml          |  7 ++++++-
 2 files changed, 14 insertions(+), 3 deletions(-)

diff --git a/models/deepseek/deepseek-v4-flash-vision-exp.toml b/models/deepseek/deepseek-v4-flash-vision-exp.toml
index ef5426b119c..18d7744f7c8 100644
--- a/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -1,14 +1,16 @@
+# Open weights (MIT): https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp
 name = "DeepSeek V4 Flash Vision Exp"
 description = "Experimental multimodal DeepSeek V4 Flash model for image understanding, coding, and agentic work"
 family = "deepseek-flash"
 release_date = "2026-08-21"
-last_updated = "2026-08-21"
+last_updated = "2026-09-01"
 attachment = true
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = false
+open_weights = true
+license = "MIT"
 
 [limit]
 context = 1_000_000
@@ -17,3 +19,7 @@ output = 384_000
 [modalities]
 input = ["text", "image"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"
diff --git a/models/deepseek/deepseek-v4.1-flash.toml b/models/deepseek/deepseek-v4.1-flash.toml
index aac72588623..68b2de09bb4 100644
--- a/models/deepseek/deepseek-v4.1-flash.toml
+++ b/models/deepseek/deepseek-v4.1-flash.toml
@@ -1,3 +1,4 @@
+# Open weights (MIT): https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash
 name = "DeepSeek V4.1 Flash"
 description = "DeepSeek V4.1 Flash model for reasoning and agentic coding"
 family = "deepseek-flash"
@@ -18,4 +19,8 @@ output = 384_000
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash"

From 4056f866d2777df0e5d278740f8faf3f1ae9d8b7 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=BC=A0=E5=9C=A3=E5=A5=87?=
 <zhangshengqi20@mails.ucas.ac.cn>
Date: Fri, 18 Sep 2026 23:11:11 +0800
Subject: [PATCH 1604/2012] feat(alibaba-cn): add deepseek-v4.1-flash (#7396)

---
 .../models/deepseek-v4.1-flash.toml           | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100644 providers/alibaba-cn/models/deepseek-v4.1-flash.toml

diff --git a/providers/alibaba-cn/models/deepseek-v4.1-flash.toml b/providers/alibaba-cn/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..933dcb373c8
--- /dev/null
+++ b/providers/alibaba-cn/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,26 @@
+# Sources (accessed 2026-09-18):
+# https://bailian.console.aliyun.com/cn-beijing?tab=model#/model-market/detail/deepseek-v4.1-flash?serviceSite=asia-pacific-china
+# https://help.aliyun.com/zh/model-studio/deepseek-api — capability table,
+# reasoning_effort ladder, max_tokens default (393_216, shared with thinking)
+# https://help.aliyun.com/zh/model-studio/billing-for-model-studio — Beijing
+# peak/off-peak pricing: CNY 2/1 input, 8/4 output per 1M tokens with context
+# cache discount; peak price converted to USD following the existing
+# deepseek-v4-flash.toml convention on this host (off-peak is half).
+# Hybrid reasoning: enable_thinking = true|false toggles thinking;
+# reasoning_effort = low|high|max (default high) — low is only supported on
+# v4.1-flash / v4-flash-0731 / v4-pro-0813, and v4.1-flash is the current
+# low-tier bearer. reasoning_content streams in deltas. Responses API
+# supported (Beijing and Singapore only). Limits and modalities are identical
+# to the deepseek lab entry (1M context, 384_000 output, text+image input),
+# so they are inherited and not restated.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.278
+output = 1.111
+cache_read = 0.014

From b17f6fcd1bde3fb35f28213e2847f00e72b71049 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:14:46 -0500
Subject: [PATCH 1605/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7382)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): correct GLM 5.3 FlashX metadata

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/zai/glm-5.3-flashx.toml      | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/vercel/models/zai/glm-5.3-flashx.toml

diff --git a/providers/vercel/models/zai/glm-5.3-flashx.toml b/providers/vercel/models/zai/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..9cf4fc74365
--- /dev/null
+++ b/providers/vercel/models/zai/glm-5.3-flashx.toml
@@ -0,0 +1,18 @@
+# GLM-5.3-FlashX uses the same model and forced-thinking effort levels as Flash.
+# https://docs.z.ai/guides/vlm/glm-5.3-flash
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 FlashX"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.37
+output = 1.25
+cache_read = 0.075
+
+[modalities]
+input = ["text", "image"]

From 79bdd0da1da7bab187c850ff45dff0d834b365fe Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 15:25:57 +0000
Subject: [PATCH 1606/2012] chore(sync): update Kilo model catalog (#7420)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 1 +
 providers/kilo/models/z-ai/glm-5.2.toml                  | 3 +--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/z-ai/glm-5.2.toml b/providers/kilo/models/z-ai/glm-5.2.toml
index dcb3cc80d7f..bb89edf75cf 100644
--- a/providers/kilo/models/z-ai/glm-5.2.toml
+++ b/providers/kilo/models/z-ai/glm-5.2.toml
@@ -11,5 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_024_000
-output = 128_000
+context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index bd4537b7471..522fff2ce46 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57816
+output = 1.73448
+cache_read = 0.018396
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 393_216
 
 [modalities]

From 127aff4232309717d4198d5cc4087928a6da3af0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 15:26:00 +0000
Subject: [PATCH 1607/2012] chore(sync): update Eden AI model catalog (#7421)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml  | 2 +-
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml     | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml      | 4 ++--
 providers/edenai/models/mistral/mistral-large-2512.toml     | 6 +++---
 .../edenai/models/scaleway/deepseek-v4-flash-0731.toml      | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml          | 4 ++--
 .../edenai/models/scaleway/llama-3.3-70b-instruct.toml      | 4 ++--
 providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml   | 1 -
 9 files changed, 16 insertions(+), 17 deletions(-)

diff --git a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
index 8a3564408ab..1e1e8029798 100644
--- a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
@@ -12,4 +12,4 @@ input = 0.065
 output = 0.18
 
 [limit]
-context = 786_432
+context = 1_048_576
diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index d5d18a5d5f0..9b2ca243154 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34443
-output = 0.45924
+input = 0.3438
+output = 0.4584
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index fc91f75adc5..b5b5ca537eb 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.746265
-output = 0.746265
+input = 0.7449
+output = 0.7449
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 488fd48dab9..fb25f513d77 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.172215
-output = 0.746265
+input = 0.1719
+output = 0.7449
diff --git a/providers/edenai/models/mistral/mistral-large-2512.toml b/providers/edenai/models/mistral/mistral-large-2512.toml
index 96902738ba4..6944f065750 100644
--- a/providers/edenai/models/mistral/mistral-large-2512.toml
+++ b/providers/edenai/models/mistral/mistral-large-2512.toml
@@ -2,9 +2,9 @@ base_model = "mistral/mistral-large-2512"
 structured_output = true
 
 [cost]
-input = 0.55
-output = 1.65
-cache_read = 0.055
+input = 0.5
+output = 1.5
+cache_read = 0.05
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 2236e082a15..ed1db7ea77d 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.45924
-output = 0.91848
+input = 0.4584
+output = 0.9168
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index c78acf705d7..df8da020b1b 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.172215
-output = 0.68886
+input = 0.1719
+output = 0.6876
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index b6133a15c15..da84cd257a7 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.03329
-output = 1.03329
+input = 1.0314
+output = 1.0314
diff --git a/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml b/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
index 87d3cc945d7..fc327494824 100644
--- a/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
+++ b/providers/edenai/models/tensorx/moonshotai/kimi-k2.5.toml
@@ -1,6 +1,5 @@
 base_model = "moonshotai/kimi-k2.5"
 name = "Kimi K2.5 (TensorX)"
-structured_output = false
 reasoning_options = []
 
 [cost]

From 68147ad3c36ba3dd8abe24fa0cd8b9185492bf9d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 15:26:04 +0000
Subject: [PATCH 1608/2012] chore(sync): update OpenRouter model catalog
 (#7419)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 ++++---
 providers/openrouter/models/meta/muse-glimmer-30b.toml     | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml   | 6 +++---
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..7c435d69ac6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.57816
+output = 1.73448
+cache_read = 0.018396
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index 12578d9072c..3b46ee2987a 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -10,8 +10,8 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.3
-output = 1.1
+input = 0.35
+output = 1.5
 cache_read = 0.04
 
 [limit]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e0b3ef90d2e..42e4245ad89 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5808
-output = 1.7424
-cache_read = 0.05808
+input = 0.57816
+output = 1.73448
+cache_read = 0.018396
 
 [limit]
 context = 1_048_576

From b1be347e4a1464ee069ae9a383b4236e1227d286 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 15:26:14 +0000
Subject: [PATCH 1609/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7418)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/zai/glm-5.3-flashx.toml | 2 --
 1 file changed, 2 deletions(-)

diff --git a/providers/vercel/models/zai/glm-5.3-flashx.toml b/providers/vercel/models/zai/glm-5.3-flashx.toml
index 9cf4fc74365..c7fc6f5e326 100644
--- a/providers/vercel/models/zai/glm-5.3-flashx.toml
+++ b/providers/vercel/models/zai/glm-5.3-flashx.toml
@@ -2,8 +2,6 @@
 # https://docs.z.ai/guides/vlm/glm-5.3-flash
 base_model = "zhipuai/glm-5.3-flash"
 name = "GLM 5.3 FlashX"
-release_date = "2026-09-18"
-last_updated = "2026-09-18"
 
 [[reasoning_options]]
 type = "effort"

From 58f89895180de2cd5b56110a8da1dc6dfa1a7732 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 10:31:18 -0500
Subject: [PATCH 1610/2012] fix(alibaba-cn): document DeepSeek exchange rate
 (#7417)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 providers/alibaba-cn/models/deepseek-v4.1-flash.toml | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/providers/alibaba-cn/models/deepseek-v4.1-flash.toml b/providers/alibaba-cn/models/deepseek-v4.1-flash.toml
index 933dcb373c8..6fe0ab81592 100644
--- a/providers/alibaba-cn/models/deepseek-v4.1-flash.toml
+++ b/providers/alibaba-cn/models/deepseek-v4.1-flash.toml
@@ -1,11 +1,12 @@
+# CNY→USD rate: 6.721845, 2026-09-18, source: https://open.er-api.com/v6/latest/USD
 # Sources (accessed 2026-09-18):
 # https://bailian.console.aliyun.com/cn-beijing?tab=model#/model-market/detail/deepseek-v4.1-flash?serviceSite=asia-pacific-china
 # https://help.aliyun.com/zh/model-studio/deepseek-api — capability table,
 # reasoning_effort ladder, max_tokens default (393_216, shared with thinking)
 # https://help.aliyun.com/zh/model-studio/billing-for-model-studio — Beijing
 # peak/off-peak pricing: CNY 2/1 input, 8/4 output per 1M tokens with context
-# cache discount; peak price converted to USD following the existing
-# deepseek-v4-flash.toml convention on this host (off-peak is half).
+# cache discount; peak price converted to USD using the rate above (off-peak
+# is half).
 # Hybrid reasoning: enable_thinking = true|false toggles thinking;
 # reasoning_effort = low|high|max (default high) — low is only supported on
 # v4.1-flash / v4-flash-0731 / v4-pro-0813, and v4.1-flash is the current
@@ -21,6 +22,6 @@ reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "h
 field = "reasoning_content"
 
 [cost]
-input = 0.278
-output = 1.111
-cache_read = 0.014
+input = 0.29754
+output = 1.19015
+cache_read = 0.01488

From 0204777328cb663967b9b52093d88c62b4730854 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 16:26:49 +0000
Subject: [PATCH 1611/2012] chore(sync): update OpenRouter model catalog
 (#7431)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/tencent/hy3.toml         | 6 +++---
 providers/openrouter/models/upstage/solar-pro-3.toml | 3 ++-
 providers/openrouter/models/upstage/solar-pro4.toml  | 3 ++-
 3 files changed, 7 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/upstage/solar-pro-3.toml b/providers/openrouter/models/upstage/solar-pro-3.toml
index 146505f4863..c9eda434a24 100644
--- a/providers/openrouter/models/upstage/solar-pro-3.toml
+++ b/providers/openrouter/models/upstage/solar-pro-3.toml
@@ -13,7 +13,8 @@ structured_output = true
 open_weights = false
 
 [[reasoning_options]]
-type = "toggle"
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0.15
diff --git a/providers/openrouter/models/upstage/solar-pro4.toml b/providers/openrouter/models/upstage/solar-pro4.toml
index adc9761fb07..7e3d19352a7 100644
--- a/providers/openrouter/models/upstage/solar-pro4.toml
+++ b/providers/openrouter/models/upstage/solar-pro4.toml
@@ -13,7 +13,8 @@ structured_output = true
 open_weights = false
 
 [[reasoning_options]]
-type = "toggle"
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.09

From 7a433d1ea8dd9c508d7f0ca44cb6197bf7d41bab Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 16:26:59 +0000
Subject: [PATCH 1612/2012] chore(sync): update Kilo model catalog (#7432)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml         | 6 +++---
 providers/kilo/models/upstage/solar-pro-3.toml | 2 +-
 providers/kilo/models/upstage/solar-pro4.toml  | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 8ea8b76bcff..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/upstage/solar-pro-3.toml b/providers/kilo/models/upstage/solar-pro-3.toml
index 54205b2bb45..ad7adb31fe8 100644
--- a/providers/kilo/models/upstage/solar-pro-3.toml
+++ b/providers/kilo/models/upstage/solar-pro-3.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "high"]
+values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
 input = 0.15
diff --git a/providers/kilo/models/upstage/solar-pro4.toml b/providers/kilo/models/upstage/solar-pro4.toml
index 4b483744118..d3486cf5460 100644
--- a/providers/kilo/models/upstage/solar-pro4.toml
+++ b/providers/kilo/models/upstage/solar-pro4.toml
@@ -12,7 +12,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.3

From 0efba743ed6ff791cc5aa52643e4f20dd46583b4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 17:22:35 +0000
Subject: [PATCH 1613/2012] chore(sync): update Kilo model catalog (#7436)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.3.toml     | 3 +--
 providers/kilo/models/~z-ai/glm-latest.toml | 8 ++++----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-5.3.toml b/providers/kilo/models/z-ai/glm-5.3.toml
index 70fec865ada..e9d5bdad891 100644
--- a/providers/kilo/models/z-ai/glm-5.3.toml
+++ b/providers/kilo/models/z-ai/glm-5.3.toml
@@ -11,5 +11,4 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_575
-output = 943_717
+context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2fc8e4f5ea7..c6f7c87e901 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.8988
+output = 2.8248
+cache_read = 0.16692
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 131_072
 
 [modalities]

From 1dc425d366c3c0ab4511b9504f6fbf01f78f5e9c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 17:22:50 +0000
Subject: [PATCH 1614/2012] chore(sync): update Merge Gateway model catalog
 (#7435)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index 5ef98e7a3a0..b86a0e589cd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,7 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
-attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +11,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 131_072
-output = 32_768
+context = 256_000
+output = 64_000
 
 [modalities]
-input = ["text"]
+input = ["text", "image"]

From c78add9b32185e8282abaa042b9fdd8b88f73574 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 17:23:34 +0000
Subject: [PATCH 1615/2012] chore(sync): update OpenRouter model catalog
 (#7437)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml      | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml              | 7 +++----
 providers/openrouter/models/~z-ai/glm-latest.toml          | 6 +++---
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a9469f150d2..887f8fb33c5 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04984
-output = 0.09968
-cache_read = 0.009968
+input = 0.04956
+output = 0.09912
+cache_read = 0.009912
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index d077461d393..6ca3ad0c4f6 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,10 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 0.91
+output = 2.86
+cache_read = 0.169
 
 [limit]
 context = 1_310_720
-output = 943_717
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0a7b3acfd67..1aaa44a3087 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.9
-output = 3
-cache_read = 0.15
+input = 0.8988
+output = 2.8248
+cache_read = 0.16692
 
 [limit]
 context = 1_310_720

From 76bb6adbd1f734d4473f3fb789c89141a79139f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 18:29:52 +0000
Subject: [PATCH 1616/2012] chore(sync): update Kilo model catalog (#7439)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml   | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index f491c2853c8..69e45ef88a2 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.14
+output = 0.42
+cache_read = 0.0042
 
 [limit]
-context = 1_000_000
-output = 393_216
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text", "image"]

From bcf2e8b580e7b0721d29767ecb0551080b22168e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 18:29:54 +0000
Subject: [PATCH 1617/2012] chore(sync): update OpenRouter model catalog
 (#7438)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 887f8fb33c5..58e87cdc226 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04956
-output = 0.09912
-cache_read = 0.009912
+input = 0.04928
+output = 0.09856
+cache_read = 0.009856
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 98fd23a037b..5e8eab978e2 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.015
+input = 0.14
+output = 0.42
+cache_read = 0.0042
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 131_072
 
 [modalities]
 input = ["text", "image"]

From b9e66aafa07959402c64d998930d8909de9762bd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 19:23:07 +0000
Subject: [PATCH 1618/2012] chore(sync): update Kilo model catalog (#7441)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/prism-ml/ternary-bonsai-2-27b.toml | 26 +++++++++++++++++++
 providers/kilo/models/~z-ai/glm-latest.toml   |  8 +++---
 2 files changed, 30 insertions(+), 4 deletions(-)
 create mode 100644 providers/kilo/models/prism-ml/ternary-bonsai-2-27b.toml

diff --git a/providers/kilo/models/prism-ml/ternary-bonsai-2-27b.toml b/providers/kilo/models/prism-ml/ternary-bonsai-2-27b.toml
new file mode 100644
index 00000000000..a1eb487d1bb
--- /dev/null
+++ b/providers/kilo/models/prism-ml/ternary-bonsai-2-27b.toml
@@ -0,0 +1,26 @@
+name = "PrismML: Ternary Bonsai 2 27B"
+description = "Bonsai 2 27B is a 27B-parameter reasoning model from PrismML derived from Qwen3.8-27B. It supports coding, mathematics, tool calling, and image understanding with a 262K-token context window. Ternary compression shrinks..."
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "xhigh"]
+
+[cost]
+input = 0.075
+output = 0.5
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index c6f7c87e901..da0527bbfae 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8988
-output = 2.8248
-cache_read = 0.16692
+input = 0.8925
+output = 2.805
+cache_read = 0.146625
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 743e022ceb024827c6aa737a4ff4883161c98489 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 19:23:15 +0000
Subject: [PATCH 1619/2012] chore(sync): update OpenRouter model catalog
 (#7440)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++--
 .../models/prism-ml/ternary-bonsai-2-27b.toml | 31 +++++++++++++++++++
 .../openrouter/models/~z-ai/glm-latest.toml   |  8 ++---
 3 files changed, 38 insertions(+), 7 deletions(-)
 create mode 100644 providers/openrouter/models/prism-ml/ternary-bonsai-2-27b.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 58e87cdc226..141d4b41ca6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04928
-output = 0.09856
-cache_read = 0.009856
+input = 0.049
+output = 0.098
+cache_read = 0.0098
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/prism-ml/ternary-bonsai-2-27b.toml b/providers/openrouter/models/prism-ml/ternary-bonsai-2-27b.toml
new file mode 100644
index 00000000000..316e8fde320
--- /dev/null
+++ b/providers/openrouter/models/prism-ml/ternary-bonsai-2-27b.toml
@@ -0,0 +1,31 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Ternary Bonsai 2 27B"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["medium", "xhigh"]
+
+[cost]
+input = 0.075
+output = 0.5
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 1aaa44a3087..0d21da4aeaf 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8988
-output = 2.8248
-cache_read = 0.16692
+input = 0.8925
+output = 2.805
+cache_read = 0.146625
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 5ecf83932e114a43cd4c31a2f5d7e137ffd23fd4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 20:25:52 +0000
Subject: [PATCH 1620/2012] chore(sync): update OpenRouter model catalog
 (#7443)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml       | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 141d4b41ca6..2309a32472b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.049
-output = 0.098
-cache_read = 0.0098
+input = 0.04872
+output = 0.09744
+cache_read = 0.009744
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 618c6fb2bbd..4e8299c6b8d 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.95
+output = 10.92
+cache_read = 0.2262
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..56f8b023ae1 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.95
+output = 10.92
+cache_read = 0.2262
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 0d21da4aeaf..d3b500c9777 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8925
-output = 2.805
-cache_read = 0.146625
+input = 0.8918
+output = 2.8028
+cache_read = 0.16562
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From fe33af6dcb907ac1ccf1b5fedb5d3d6dbacafccd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 20:26:10 +0000
Subject: [PATCH 1621/2012] chore(sync): update Kilo model catalog (#7442)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml      | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml        | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 3af9cc14392..9fe8a4fd3f8 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.95
+output = 10.92
+cache_read = 0.2262
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..fcb5b91d30e 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 1.95
+output = 10.92
+cache_read = 0.2262
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index da0527bbfae..73bfc195431 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8925
-output = 2.805
-cache_read = 0.146625
+input = 0.8918
+output = 2.8028
+cache_read = 0.16562
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 115bd0c7a86ad1a2635fbd7c3aa268ce10490de4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 21:23:41 +0000
Subject: [PATCH 1622/2012] chore(sync): update Kilo model catalog (#7445)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 844f1914d6e..b96573b6b7a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.055
-output = 0.165
-cache_read = 0.00175
+input = 0.05412
+output = 0.16236
+cache_read = 0.001722
 
 [limit]
 context = 1_024_000

From 709448a4719255680e5727052faafd3661e4206d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 21:23:45 +0000
Subject: [PATCH 1623/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7446)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/fish-audio/s1-free.toml  | 16 ----------------
 .../vercel/models/fish-audio/s2-pro-free.toml    | 16 ----------------
 .../vercel/models/fish-audio/s2.1-pro-free.toml  | 16 ----------------
 .../models/fish-audio/transcribe-1-free.toml     | 16 ----------------
 4 files changed, 64 deletions(-)
 delete mode 100644 providers/vercel/models/fish-audio/s1-free.toml
 delete mode 100644 providers/vercel/models/fish-audio/s2-pro-free.toml
 delete mode 100644 providers/vercel/models/fish-audio/s2.1-pro-free.toml
 delete mode 100644 providers/vercel/models/fish-audio/transcribe-1-free.toml

diff --git a/providers/vercel/models/fish-audio/s1-free.toml b/providers/vercel/models/fish-audio/s1-free.toml
deleted file mode 100644
index 2eb0f456def..00000000000
--- a/providers/vercel/models/fish-audio/s1-free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-name = "S1 (Free)"
-description = "Speech generation model for controllable voice, narration, and audio delivery"
-release_date = "2025-10-20"
-last_updated = "2025-10-20"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["text"]
-output = ["audio"]
diff --git a/providers/vercel/models/fish-audio/s2-pro-free.toml b/providers/vercel/models/fish-audio/s2-pro-free.toml
deleted file mode 100644
index c2bbbb724c8..00000000000
--- a/providers/vercel/models/fish-audio/s2-pro-free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-name = "S2 Pro (Free)"
-description = "Speech generation model for controllable voice, narration, and audio delivery"
-release_date = "2026-03-09"
-last_updated = "2026-03-09"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["text"]
-output = ["audio"]
diff --git a/providers/vercel/models/fish-audio/s2.1-pro-free.toml b/providers/vercel/models/fish-audio/s2.1-pro-free.toml
deleted file mode 100644
index c580e92cd34..00000000000
--- a/providers/vercel/models/fish-audio/s2.1-pro-free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-name = "S2.1 Pro (Free)"
-description = "Speech generation model for controllable voice, narration, and audio delivery"
-release_date = "2026-07-28"
-last_updated = "2026-07-28"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["text"]
-output = ["audio"]
diff --git a/providers/vercel/models/fish-audio/transcribe-1-free.toml b/providers/vercel/models/fish-audio/transcribe-1-free.toml
deleted file mode 100644
index e21df2298c0..00000000000
--- a/providers/vercel/models/fish-audio/transcribe-1-free.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-name = "Transcribe-1 (Free)"
-description = "Speech transcription model for accurate audio-to-text and captioning workflows"
-release_date = "2026-03-01"
-last_updated = "2026-03-01"
-attachment = false
-reasoning = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 0
-output = 0
-
-[modalities]
-input = ["audio"]
-output = ["text"]

From 6a9e725a50369370c91d12c06cb8d5dc67e3f059 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 21:23:52 +0000
Subject: [PATCH 1624/2012] chore(sync): update OpenRouter model catalog
 (#7444)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/google/gemma-4-26b-a4b-it.toml  | 1 +
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 2309a32472b..814013f36dd 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04872
-output = 0.09744
-cache_read = 0.009744
+input = 0.04844
+output = 0.09688
+cache_read = 0.009688
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index 7d230a2be6f..872203b2862 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -11,6 +11,7 @@ output = 0.3
 cache_read = 0.05
 
 [limit]
+context = 1_000_000
 output = 235_929
 
 [modalities]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index b207a5a59a3..9f76b8f4c29 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.055
-output = 0.165
-cache_read = 0.00175
+input = 0.05412
+output = 0.16236
+cache_read = 0.001722
 
 [limit]
 context = 1_310_720

From 9f4dbb29e1960215c03d76714c6675910e0fb9bb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 22:24:16 +0000
Subject: [PATCH 1625/2012] chore(sync): update OpenRouter model catalog
 (#7449)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/google/gemma-4-26b-a4b-it.toml      | 1 -
 .../models/~deepseek/deepseek-flash-latest.toml           | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 6 +++---
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
index 872203b2862..7d230a2be6f 100644
--- a/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
+++ b/providers/openrouter/models/google/gemma-4-26b-a4b-it.toml
@@ -11,7 +11,6 @@ output = 0.3
 cache_read = 0.05
 
 [limit]
-context = 1_000_000
 output = 235_929
 
 [modalities]
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 5e8eab978e2..bdcd94e72e3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.42
-cache_read = 0.0042
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 9f76b8f4c29..f00eacc90b8 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05412
-output = 0.16236
-cache_read = 0.001722
+input = 0.05324
+output = 0.15972
+cache_read = 0.001694
 
 [limit]
 context = 1_310_720

From ff883d057d726877568b1d89cec733ea95df9ef0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 22:24:20 +0000
Subject: [PATCH 1626/2012] chore(sync): update Kilo model catalog (#7450)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml      | 8 ++++----
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml   | 6 +++---
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 69e45ef88a2..0b2d5078274 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.14
-output = 0.42
-cache_read = 0.0042
+input = 0.135
+output = 0.54
+cache_read = 0.00405
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index b96573b6b7a..cdd99b9392d 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05412
-output = 0.16236
-cache_read = 0.001722
+input = 0.05324
+output = 0.15972
+cache_read = 0.001694
 
 [limit]
 context = 1_024_000

From 91086d2c8d667ca5058a2bc77f8f8684293c871e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 23:24:12 +0000
Subject: [PATCH 1627/2012] chore(sync): update OpenRouter model catalog
 (#7455)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index f00eacc90b8..83799fb5d7c 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05324
-output = 0.15972
-cache_read = 0.001694
+input = 0.05236
+output = 0.15708
+cache_read = 0.001666
 
 [limit]
 context = 1_310_720

From 6b1fd113c90ca967bd9e4d8d3a529cceb9a226a3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 23:24:16 +0000
Subject: [PATCH 1628/2012] chore(sync): update Kilo model catalog (#7456)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index cdd99b9392d..64711cafb8f 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05324
-output = 0.15972
-cache_read = 0.001694
+input = 0.05236
+output = 0.15708
+cache_read = 0.001666
 
 [limit]
 context = 1_024_000

From 6b5d1eaa97c36e23adaa4584640d60651f5fce94 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Fri, 18 Sep 2026 23:24:46 +0000
Subject: [PATCH 1629/2012] chore(sync): update NanoGPT model catalog (#7457)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../mistral-large-3-675b-instruct-2512.toml   | 24 -----------------
 .../models/prism-ml/ternary-bonsai-2-27b.toml | 27 +++++++++++++++++++
 2 files changed, 27 insertions(+), 24 deletions(-)
 delete mode 100644 providers/nano-gpt/models/mistralai/mistral-large-3-675b-instruct-2512.toml
 create mode 100644 providers/nano-gpt/models/prism-ml/ternary-bonsai-2-27b.toml

diff --git a/providers/nano-gpt/models/mistralai/mistral-large-3-675b-instruct-2512.toml b/providers/nano-gpt/models/mistralai/mistral-large-3-675b-instruct-2512.toml
deleted file mode 100644
index 0a0650a3987..00000000000
--- a/providers/nano-gpt/models/mistralai/mistral-large-3-675b-instruct-2512.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Mistral Large 3 675B"
-description = "Flagship Mistral model for advanced reasoning, coding, and multilingual work"
-family = "mistral-large"
-release_date = "2025-12-25"
-last_updated = "2025-12-02"
-attachment = true
-reasoning = false
-tool_call = false
-structured_output = false
-open_weights = true
-
-[cost]
-input = 1
-output = 3
-cache_read = 0.5
-
-[limit]
-context = 262_144
-input = 262_144
-output = 256_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/prism-ml/ternary-bonsai-2-27b.toml b/providers/nano-gpt/models/prism-ml/ternary-bonsai-2-27b.toml
new file mode 100644
index 00000000000..32eb4c91ae5
--- /dev/null
+++ b/providers/nano-gpt/models/prism-ml/ternary-bonsai-2-27b.toml
@@ -0,0 +1,27 @@
+name = "Ternary Bonsai 2 27B"
+description = "Ternary Bonsai 2 27B is a 27B-parameter reasoning model from PrismML derived from Qwen3.8-27B. It supports coding, mathematics, tool calling, and image understanding with a 262K-token context window."
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["medium", "xhigh"]
+
+[cost]
+input = 0.075
+output = 0.5
+cache_read = 0.0375
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From dcd5fc29801fdfa285c8cd0553bfbca3bb7652a4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 00:57:46 +0000
Subject: [PATCH 1630/2012] chore(sync): update OpenRouter model catalog
 (#7460)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 7 +++----
 .../models/nvidia/nemotron-3-ultra-550b-a55b.toml         | 8 ++++----
 .../openrouter/models/nvidia/nemotron-3.5-lightning.toml  | 4 ++--
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 4 ++--
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 6 +++---
 6 files changed, 17 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 95460b6f5f3..4ce5d404923 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,10 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 1.6
-output = 3.2
-cache_read = 0.135
+input = 0.675816
+output = 1.351632
+cache_read = 0.056318
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index f471e71439f..5dcd117448e 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -14,10 +14,10 @@ values = ["medium", "high"]
 type = "budget_tokens"
 
 [cost]
-input = 0.625
-output = 3.125
-cache_read = 0.1875
+input = 0.6
+output = 2.4
+cache_read = 0.12
 
 [limit]
 context = 262_144
-output = 32_768
+output = 182_520
diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
index 3b251679339..842e5779095 100644
--- a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
@@ -7,9 +7,9 @@ description = "Nemotron model for efficient reasoning, coding, and specialized A
 type = "toggle"
 
 [cost]
-input = 0.08
+input = 0.07
 output = 0.2
 cache_read = 0.04
 
 [limit]
-output = 131_072
+output = 235_929
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index d857fa53dab..91402212177 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.13
-output = 0.52
+input = 0.2
+output = 0.7
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 83799fb5d7c..f51872c2fb8 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05236
-output = 0.15708
-cache_read = 0.001666
+input = 0.05148
+output = 0.15444
+cache_read = 0.001638
 
 [limit]
 context = 1_310_720

From 255919f646ed4f9382b0b5fbe7eb23fe18be1126 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 00:57:52 +0000
Subject: [PATCH 1631/2012] chore(sync): update Kilo model catalog (#7462)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro.toml    |  3 +--
 .../kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml |  4 ++--
 .../kilo/models/nvidia/nemotron-3.5-lightning.toml     |  2 +-
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml     | 10 +++++-----
 5 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro.toml b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
index e58e24281d4..88500bd7dd8 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro.toml
@@ -11,5 +11,4 @@ output = 3.2
 cache_read = 0.135
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
index 212e280b8a9..25c72a0b0fb 100644
--- a/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
+++ b/providers/kilo/models/nvidia/nemotron-3-ultra-550b-a55b.toml
@@ -12,5 +12,5 @@ output = 2.2
 cache_read = 0.1
 
 [limit]
-context = 256_000
-output = 32_768
+context = 202_800
+output = 182_520
diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index 78104d7319f..5244dc1ac02 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -10,4 +10,4 @@ input = 0.065
 output = 0.18
 
 [limit]
-output = 131_072
+output = 235_929
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 64711cafb8f..dcbc3433279 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05236
-output = 0.15708
-cache_read = 0.001666
+input = 0.05104
+output = 0.15312
+cache_read = 0.001624
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text"]

From fead8587cea02673f3a10c6be712ae83851f8c17 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 00:57:54 +0000
Subject: [PATCH 1632/2012] chore(sync): update NanoGPT model catalog (#7461)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml  | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
index abb7e798efd..fb7f9db2975 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-latest.toml
@@ -19,9 +19,9 @@ output = 0.16
 cache_read = 0.013
 
 [limit]
-context = 1_048_576
-input = 1_048_576
-output = 384_000
+context = 1_000_000
+input = 1_000_000
+output = 131_072
 
 [modalities]
 input = ["text"]

From dff014f67f04d7f17b6a7024510c77dd4a5b47e5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 00:57:57 +0000
Subject: [PATCH 1633/2012] chore(sync): update Venice model catalog (#7463)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-56-sol-pro.toml     | 16 ++++++++--------
 providers/venice/models/openai-gpt-56-sol.toml   | 16 ++++++++--------
 2 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/providers/venice/models/openai-gpt-56-sol-pro.toml b/providers/venice/models/openai-gpt-56-sol-pro.toml
index 36093d01914..1b837b0d610 100644
--- a/providers/venice/models/openai-gpt-56-sol-pro.toml
+++ b/providers/venice/models/openai-gpt-56-sol-pro.toml
@@ -8,18 +8,18 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.5
-output = 12.5
-cache_read = 0.25
-cache_write = 3.125
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
 input = 5
-output = 18.75
+output = 25
 cache_read = 0.5
 cache_write = 6.25
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 37.5
+cache_read = 1
+cache_write = 12.5
+
 [limit]
 context = 1_000_000
 
diff --git a/providers/venice/models/openai-gpt-56-sol.toml b/providers/venice/models/openai-gpt-56-sol.toml
index dedfc528097..a0a89340131 100644
--- a/providers/venice/models/openai-gpt-56-sol.toml
+++ b/providers/venice/models/openai-gpt-56-sol.toml
@@ -7,18 +7,18 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2.5
-output = 12.5
-cache_read = 0.25
-cache_write = 3.125
-
-[[cost.tiers]]
-tier = { type = "context", size = 272_000 }
 input = 5
-output = 18.75
+output = 25
 cache_read = 0.5
 cache_write = 6.25
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 37.5
+cache_read = 1
+cache_write = 12.5
+
 [limit]
 context = 1_000_000
 

From 248a44ccbcf6c2bc38fe8212720b5050cff01d76 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 01:34:31 +0000
Subject: [PATCH 1634/2012] chore(sync): update Kilo model catalog (#7465)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml          |  6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml     | 10 +++++-----
 providers/kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 3 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 9fe8a4fd3f8..3af9cc14392 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.95
-output = 10.92
-cache_read = 0.2262
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index dcbc3433279..84e8197e6b4 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.05104
-output = 0.15312
-cache_read = 0.001624
+input = 0.0506
+output = 0.1518
+cache_read = 0.00161
 
 [limit]
-context = 1_048_576
-output = 131_072
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index fcb5b91d30e..9bbf3c7be77 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.95
-output = 10.92
-cache_read = 0.2262
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 6303a062a3782391f3147e9db1100ee93139abbb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 01:34:47 +0000
Subject: [PATCH 1635/2012] chore(sync): update OpenRouter model catalog
 (#7466)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4ce5d404923..62b0bbc3ecf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.675816
-output = 1.351632
-cache_read = 0.056318
+input = 0.66555
+output = 1.3311
+cache_read = 0.055463
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 4e8299c6b8d..618c6fb2bbd 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.95
-output = 10.92
-cache_read = 0.2262
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index f51872c2fb8..7c73e02e9d6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05148
-output = 0.15444
-cache_read = 0.001638
+input = 0.0506
+output = 0.1518
+cache_read = 0.00161
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 56f8b023ae1..d1d254d74de 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.95
-output = 10.92
-cache_read = 0.2262
+input = 2.1
+output = 10.95
+cache_read = 0.23
 
 [limit]
 context = 1_048_576

From 63ad21b91c8a4dd1be7331a44313ca8dcdeefce5 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Sat, 19 Sep 2026 10:21:04 +0800
Subject: [PATCH 1636/2012] feat(aihubmix): add ox-alpha (#7428)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/zhipuai/glm-5.3-flash.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/ox-alpha.toml | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
 create mode 100644 providers/aihubmix/models/ox-alpha.toml

diff --git a/providers/aihubmix/models/ox-alpha.toml b/providers/aihubmix/models/ox-alpha.toml
new file mode 100644
index 00000000000..800e20046f6
--- /dev/null
+++ b/providers/aihubmix/models/ox-alpha.toml
@@ -0,0 +1,23 @@
+# Toggle:
+# $.enable_thinking = true|false on the OpenAI-compatible /v1/chat/completions path (verified live 2026-09-11);
+# $.thinking.type = "enabled"|"disabled"|"adaptive" on /v1/messages; $.generationConfig.thinkingConfig on the Gemini path.
+# Effort: low|high|max
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "zhipuai/glm-5.3-flash"
+name = "Ox Alpha"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0

From af57606c96af3f5960fe7a12f5e6cca77a8518d2 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Sat, 19 Sep 2026 10:21:13 +0800
Subject: [PATCH 1637/2012] feat(aihubmix): add deepseek-v4-flash-vision-exp
 (#7427)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/deepseek/deepseek-v4-flash-vision-exp.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 .../models/deepseek-v4-flash-vision-exp.toml  | 23 +++++++++++++++++++
 1 file changed, 23 insertions(+)
 create mode 100644 providers/aihubmix/models/deepseek-v4-flash-vision-exp.toml

diff --git a/providers/aihubmix/models/deepseek-v4-flash-vision-exp.toml b/providers/aihubmix/models/deepseek-v4-flash-vision-exp.toml
new file mode 100644
index 00000000000..3cd40d48314
--- /dev/null
+++ b/providers/aihubmix/models/deepseek-v4-flash-vision-exp.toml
@@ -0,0 +1,23 @@
+# Toggle:
+# $.enable_thinking = true|false on the OpenAI-compatible /v1/chat/completions path (verified live 2026-09-11);
+# $.thinking.type = "enabled"|"disabled"|"adaptive" on /v1/messages; $.generationConfig.thinkingConfig on the Gemini path.
+# Effort: low|high|max
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.155
+output = 0.62
+cache_read = 0.0031

From d809a3f9909fb8320cd17acf4c6ab6aeb0eb8077 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Sat, 19 Sep 2026 10:21:21 +0800
Subject: [PATCH 1638/2012] feat(aihubmix): add qwen3.8-flash (#7426)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/alibaba/qwen3.8-flash.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/qwen3.8-flash.toml | 31 ++++++++++++++++++++
 1 file changed, 31 insertions(+)
 create mode 100644 providers/aihubmix/models/qwen3.8-flash.toml

diff --git a/providers/aihubmix/models/qwen3.8-flash.toml b/providers/aihubmix/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..69e8766006b
--- /dev/null
+++ b/providers/aihubmix/models/qwen3.8-flash.toml
@@ -0,0 +1,31 @@
+# Toggle:
+# $.enable_thinking = true|false on the OpenAI-compatible /v1/chat/completions path (verified live 2026-09-11);
+# $.thinking.type = "enabled"|"disabled"|"adaptive" on /v1/messages; $.generationConfig.thinkingConfig on the Gemini path.
+# Effort: low|medium|xhigh
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# Budget:
+# integer $.reasoning.max_tokens on /v1/chat/completions; $.thinking.budget_tokens >= 1024 on /v1/messages;
+# integer $.generationConfig.thinkingConfig.thinkingBudget on the Gemini path (-1 dynamic, 0 off where supported);
+# the Responses path carries effort but has no reasoning-token budget field.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "alibaba/qwen3.8-flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.1126
+output = 0.380025
+cache_read = 0.014075
+cache_write = 0.175937

From f14883a075e95ed2dad7565a276c90dfa30d1481 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 02:33:13 +0000
Subject: [PATCH 1639/2012] chore(sync): update OpenRouter model catalog
 (#7469)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index df31c0bbb32..1273df92943 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.06
-output = 0.12
-cache_read = 0.012
+input = 0.05
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 7c73e02e9d6..5a559b9f1f5 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.0506
-output = 0.1518
-cache_read = 0.00161
+input = 0.04972
+output = 0.14916
+cache_read = 0.001582
 
 [limit]
 context = 1_310_720
-output = 384_000
+output = 131_072
 
 [modalities]
 input = ["text"]

From bc21426e3a59ca8b9b1afd8d24973bdc16e02217 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 02:33:15 +0000
Subject: [PATCH 1640/2012] chore(sync): update Kilo model catalog (#7468)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~deepseek/deepseek-v4-flash-latest.toml     | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 84e8197e6b4..857a18e8b48 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.0506
-output = 0.1518
-cache_read = 0.00161
+input = 0.04972
+output = 0.14916
+cache_read = 0.001582
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text"]

From e775635942b83c102ebde5b772e1ab8c7663339e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 03:29:08 +0000
Subject: [PATCH 1641/2012] chore(sync): update Kilo model catalog (#7471)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/nvidia/nemotron-3.5-lightning.toml |  2 +-
 providers/kilo/models/openai/gpt-oss-20b.toml |  4 +--
 .../kilo/models/z-ai/glm-5.3-flashx.toml      | 28 +++++++++++++++++++
 .../~deepseek/deepseek-v4-flash-latest.toml   | 10 +++----
 4 files changed, 36 insertions(+), 8 deletions(-)
 create mode 100644 providers/kilo/models/z-ai/glm-5.3-flashx.toml

diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index 5244dc1ac02..6759948a30e 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -6,7 +6,7 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.065
+input = 0.04
 output = 0.18
 
 [limit]
diff --git a/providers/kilo/models/openai/gpt-oss-20b.toml b/providers/kilo/models/openai/gpt-oss-20b.toml
index 895f7fb4602..fe3d2508db6 100644
--- a/providers/kilo/models/openai/gpt-oss-20b.toml
+++ b/providers/kilo/models/openai/gpt-oss-20b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.02
-output = 0.1
+input = 0.018
+output = 0.09
 
 [limit]
 output = 117_964
diff --git a/providers/kilo/models/z-ai/glm-5.3-flashx.toml b/providers/kilo/models/z-ai/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..a730a63619c
--- /dev/null
+++ b/providers/kilo/models/z-ai/glm-5.3-flashx.toml
@@ -0,0 +1,28 @@
+name = "Z.ai: GLM 5.3 FlashX"
+description = "GLM-5.3-FlashX is the high-speed variant of Z.ai's GLM-5.3-Flash, a native multimodal model delivering inference speeds of up to 200 tokens/s. Built on the same hybrid sparse and linear attention architecture..."
+family = "glm"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.37
+output = 1.25
+cache_read = 0.075
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 857a18e8b48..50af69be57a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04972
-output = 0.14916
-cache_read = 0.001582
+input = 0.04928
+output = 0.14784
+cache_read = 0.001568
 
 [limit]
-context = 1_048_576
-output = 131_072
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From 4676efad5e24df24ef63f4624696df368c22f06f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 03:29:25 +0000
Subject: [PATCH 1642/2012] chore(sync): update OpenRouter model catalog
 (#7470)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731.toml      |  6 ++--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/z-ai/glm-5.3-flashx.toml           | 28 +++++++++++++++++++
 .../~deepseek/deepseek-v4-flash-latest.toml   |  8 +++---
 4 files changed, 38 insertions(+), 10 deletions(-)
 create mode 100644 providers/openrouter/models/z-ai/glm-5.3-flashx.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 1273df92943..d06dd2cd8d1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.05
-output = 0.1
-cache_read = 0.01
+input = 0.065
+output = 0.18
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 62b0bbc3ecf..a3a0173e68a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.66555
-output = 1.3311
-cache_read = 0.055463
+input = 0.645366
+output = 1.290732
+cache_read = 0.053781
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flashx.toml b/providers/openrouter/models/z-ai/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..ea5b2ee8747
--- /dev/null
+++ b/providers/openrouter/models/z-ai/glm-5.3-flashx.toml
@@ -0,0 +1,28 @@
+name = "GLM 5.3 FlashX"
+description = "GLM vision model for visual reasoning, documents, and multimodal agents"
+family = "glm"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.37
+output = 1.25
+cache_read = 0.075
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 5a559b9f1f5..189296e3dde 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04972
-output = 0.14916
-cache_read = 0.001582
+input = 0.04928
+output = 0.14784
+cache_read = 0.001568
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 384_000
 
 [modalities]
 input = ["text"]

From 227516b6e6f96960c5b952e7c0dfaef00a91b54a Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Sat, 19 Sep 2026 11:45:24 +0800
Subject: [PATCH 1643/2012] feat(opencode): add qwen3.8 flash to zen

---
 providers/opencode/models/qwen3.8-flash.toml | 25 ++++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 providers/opencode/models/qwen3.8-flash.toml

diff --git a/providers/opencode/models/qwen3.8-flash.toml b/providers/opencode/models/qwen3.8-flash.toml
new file mode 100644
index 00000000000..1cd079b5cb3
--- /dev/null
+++ b/providers/opencode/models/qwen3.8-flash.toml
@@ -0,0 +1,25 @@
+# https://opencode.ai/docs/go/#endpoints
+# https://help.aliyun.com/en/model-studio/anthropic-api-messages
+# Toggle (Messages): thinking.type = enabled|disabled
+# Effort (Messages): output_config.effort = low|medium|xhigh
+base_model = "alibaba/qwen3.8-flash"
+temperature = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
+
+[provider]
+npm = "@ai-sdk/anthropic"
\ No newline at end of file

From 27c0683399bdf23e3bfecc716cba524c8ad5afb6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 04:28:48 +0000
Subject: [PATCH 1644/2012] chore(sync): update OpenRouter model catalog
 (#7474)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml   | 4 ++--
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 6 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 814013f36dd..b5c7d255938 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04844
-output = 0.09688
-cache_read = 0.009688
+input = 0.04816
+output = 0.09632
+cache_read = 0.009632
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a3a0173e68a..ff6ad910cd9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.645366
-output = 1.290732
-cache_read = 0.053781
+input = 0.614916
+output = 1.229832
+cache_read = 0.051243
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 618c6fb2bbd..d374007e872 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0695
+output = 11.5892
+cache_read = 0.240062
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 91402212177..d857fa53dab 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.7
+input = 0.13
+output = 0.52
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 189296e3dde..5aff2dc4d72 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04928
-output = 0.14784
-cache_read = 0.001568
+input = 0.0484
+output = 0.1452
+cache_read = 0.00154
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index d1d254d74de..b15245b9227 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0695
+output = 11.5892
+cache_read = 0.240062
 
 [limit]
 context = 1_048_576

From 7e6dd18cd010b678157324056be2fe1be53ac0b3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 04:28:53 +0000
Subject: [PATCH 1645/2012] chore(sync): update Kilo model catalog (#7473)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml               | 6 +++---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml          | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 3af9cc14392..7965ee661dd 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0695
+output = 11.5892
+cache_read = 0.240062
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 50af69be57a..4cea316de5e 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04928
-output = 0.14784
-cache_read = 0.001568
+input = 0.0484
+output = 0.1452
+cache_read = 0.00154
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 9bbf3c7be77..86ff6cbcca0 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.1
-output = 10.95
-cache_read = 0.23
+input = 2.0695
+output = 11.5892
+cache_read = 0.240062
 
 [limit]
 context = 1_048_576

From 043d5793090f6a78503f36abf97214782f10281a Mon Sep 17 00:00:00 2001
From: Daniel Chen <daniel85.chen@gmail.com>
Date: Fri, 18 Sep 2026 21:30:58 -0700
Subject: [PATCH 1646/2012] feat(opencode):added deepseek-v4.1-flash to zen

---
 providers/opencode/models/deepseek-v4.1-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/opencode/models/deepseek-v4.1-flash.toml

diff --git a/providers/opencode/models/deepseek-v4.1-flash.toml b/providers/opencode/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..20a4928d010
--- /dev/null
+++ b/providers/opencode/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006

From ebfeed5f5eb53e48f6fdf659cf9907ee0172ec03 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 00:17:06 -0500
Subject: [PATCH 1647/2012] feat: add GLM-5.3-FlashX to Z.AI providers (#7475)

Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>
---
 providers/zai/models/glm-5.3-flash.toml      |  7 ++++---
 providers/zai/models/glm-5.3-flashx.toml     | 21 ++++++++++++++++++++
 providers/zhipuai/models/glm-5.3-flash.toml  |  9 ++++-----
 providers/zhipuai/models/glm-5.3-flashx.toml | 21 ++++++++++++++++++++
 4 files changed, 50 insertions(+), 8 deletions(-)
 create mode 100644 providers/zai/models/glm-5.3-flashx.toml
 create mode 100644 providers/zhipuai/models/glm-5.3-flashx.toml

diff --git a/providers/zai/models/glm-5.3-flash.toml b/providers/zai/models/glm-5.3-flash.toml
index 59029d6336a..5850a4727dc 100644
--- a/providers/zai/models/glm-5.3-flash.toml
+++ b/providers/zai/models/glm-5.3-flash.toml
@@ -2,6 +2,7 @@ base_model = "zhipuai/glm-5.3-flash"
 # GLM-5.3-Flash reasons by default; effort levels low|high|max, and thinking can
 # be suppressed with {"thinking": {"type": "disabled"}}.
 # https://z.ai/blog/glm-5.3-flash
+# Pricing: https://docs.z.ai/guides/overview/pricing (accessed 2026-09-19)
 # Served by https://api.z.ai/api/paas/v4 (verified against GET /models 2026-08-26).
 
 [[reasoning_options]]
@@ -12,7 +13,7 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.50
+cache_read = 0.03
 cache_write = 0
diff --git a/providers/zai/models/glm-5.3-flashx.toml b/providers/zai/models/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..059cb836933
--- /dev/null
+++ b/providers/zai/models/glm-5.3-flashx.toml
@@ -0,0 +1,21 @@
+# GLM-5.3-FlashX is the high-speed serving option for GLM-5.3-Flash.
+# https://docs.z.ai/guides/vlm/glm-5.3-flash
+# Pricing: https://docs.z.ai/guides/overview/pricing (accessed 2026-09-19)
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3-FlashX"
+description = "High-speed GLM-5.3-Flash serving option for coding and agent workflows"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.37
+output = 1.25
+cache_read = 0.075
+cache_write = 0
diff --git a/providers/zhipuai/models/glm-5.3-flash.toml b/providers/zhipuai/models/glm-5.3-flash.toml
index 84efd9a9cef..d2ff44820e7 100644
--- a/providers/zhipuai/models/glm-5.3-flash.toml
+++ b/providers/zhipuai/models/glm-5.3-flash.toml
@@ -1,8 +1,7 @@
 # GLM-5.3-Flash always reasons (thinking cannot be disabled); text params
 # match GLM-5.3 — effort low|high|max with default max.
 # https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash
-# Cost: https://docs.z.ai/guides/overview/pricing (accessed 2026-08-26)
-# GLM-5.3-Flash 50% promo ends 2026-09-09 24:00 UTC+8; list $0.15/$0.03/$0.50.
+# Pricing: https://docs.z.ai/guides/overview/pricing (accessed 2026-09-19)
 base_model = "zhipuai/glm-5.3-flash"
 
 [[reasoning_options]]
@@ -13,7 +12,7 @@ values = ["low", "high", "max"]
 field = "reasoning_content"
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.015
+input = 0.15
+output = 0.50
+cache_read = 0.03
 cache_write = 0
diff --git a/providers/zhipuai/models/glm-5.3-flashx.toml b/providers/zhipuai/models/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..851a6969e41
--- /dev/null
+++ b/providers/zhipuai/models/glm-5.3-flashx.toml
@@ -0,0 +1,21 @@
+# GLM-5.3-FlashX is the high-speed serving option for GLM-5.3-Flash.
+# https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash
+# Pricing: https://docs.z.ai/guides/overview/pricing (accessed 2026-09-19)
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3-FlashX"
+description = "High-speed GLM-5.3-Flash serving option for coding and agent workflows"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.37
+output = 1.25
+cache_read = 0.075
+cache_write = 0

From a25d952334e2331e6649b1e054b9a7b4579dd978 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 05:25:15 +0000
Subject: [PATCH 1648/2012] chore(sync): update OpenRouter model catalog
 (#7476)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml   | 4 ++--
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b5c7d255938..cc0ba12eb64 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04816
-output = 0.09632
-cache_read = 0.009632
+input = 0.0476
+output = 0.0952
+cache_read = 0.00952
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ff6ad910cd9..b670ae33837 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.614916
-output = 1.229832
-cache_read = 0.051243
+input = 0.594558
+output = 1.189116
+cache_read = 0.049547
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index d857fa53dab..91402212177 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.13
-output = 0.52
+input = 0.2
+output = 0.7
 
 [limit]
 context = 262_144

From b979ab5354dd2b222740c3cce5eda0db8d281394 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 06:37:07 +0000
Subject: [PATCH 1649/2012] chore(sync): update OpenRouter model catalog
 (#7478)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 5 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index cc0ba12eb64..c5c46f3604a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0476
-output = 0.0952
-cache_read = 0.00952
+input = 0.04732
+output = 0.09464
+cache_read = 0.009464
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b670ae33837..f453f0bbdbd 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.594558
-output = 1.189116
-cache_read = 0.049547
+input = 0.564282
+output = 1.128564
+cache_read = 0.047024
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index d374007e872..c1033f6ad0b 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0695
-output = 11.5892
-cache_read = 0.240062
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 5aff2dc4d72..8324901e3db 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.0484
-output = 0.1452
-cache_read = 0.00154
+input = 0.04752
+output = 0.14256
+cache_read = 0.001512
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index b15245b9227..ec32a69e6bd 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0695
-output = 11.5892
-cache_read = 0.240062
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From d25c8085b265e08910be5c834098c7afa65b58ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 06:37:12 +0000
Subject: [PATCH 1650/2012] chore(sync): update Kilo model catalog (#7480)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml |  6 ++--
 .../models/openai/gpt-5.6-sol-discounted.toml | 30 -------------------
 .../~deepseek/deepseek-v4-flash-latest.toml   |  6 ++--
 .../kilo/models/~moonshotai/kimi-latest.toml  |  6 ++--
 4 files changed, 9 insertions(+), 39 deletions(-)
 delete mode 100644 providers/kilo/models/openai/gpt-5.6-sol-discounted.toml

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 7965ee661dd..90c5d22728f 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.0695
-output = 11.5892
-cache_read = 0.240062
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml b/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml
deleted file mode 100644
index 6008af017c3..00000000000
--- a/providers/kilo/models/openai/gpt-5.6-sol-discounted.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "OpenAI: GPT-5.6 Sol (50% off)"
-description = "GPT-5.6 Sol served by OpenAI through Vercel AI Gateway at 50% lower cost than other available inference providers. This promotion runs through September 18, 2026."
-family = "gpt"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 2
-output = 10
-reasoning = 0
-cache_read = 0.2
-cache_write = 2.5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 4cea316de5e..87db0089581 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.0484
-output = 0.1452
-cache_read = 0.00154
+input = 0.04752
+output = 0.14256
+cache_read = 0.001512
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 86ff6cbcca0..f09092fb568 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 2.0695
-output = 11.5892
-cache_read = 0.240062
+input = 1.875
+output = 10.5
+cache_read = 0.2175
 
 [limit]
 context = 1_048_576

From 97816c1e80054356e29db7c4542a257248d0594e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 06:37:15 +0000
Subject: [PATCH 1651/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7479)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/openai/gpt-5.6-sol-fast.toml | 8 ++++----
 providers/vercel/models/openai/gpt-5.6-sol.toml      | 8 ++++----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
index adf0da2fcb0..007699dc75e 100644
--- a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
@@ -6,7 +6,7 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 4
-output = 20
-cache_read = 0.4
-cache_write = 5
+input = 8
+output = 40
+cache_read = 0.8
+cache_write = 10
diff --git a/providers/vercel/models/openai/gpt-5.6-sol.toml b/providers/vercel/models/openai/gpt-5.6-sol.toml
index 8b88a743a65..b07d7577a9a 100644
--- a/providers/vercel/models/openai/gpt-5.6-sol.toml
+++ b/providers/vercel/models/openai/gpt-5.6-sol.toml
@@ -8,7 +8,7 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5

From 1337abc7d3d91826557f6cf4e87c45ccc5424f60 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 07:25:51 +0000
Subject: [PATCH 1652/2012] chore(sync): update OpenRouter model catalog
 (#7488)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml       | 6 +++---
 .../models/~deepseek/deepseek-flash-latest.toml           | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 8 ++++----
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 6 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index d06dd2cd8d1..8dce0b17a1e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.016
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c5c46f3604a..664a084d4ad 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04732
-output = 0.09464
-cache_read = 0.009464
+input = 0.04704
+output = 0.09408
+cache_read = 0.009408
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index c1033f6ad0b..22b1407542a 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index bdcd94e72e3..643ffd542be 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.13
+output = 0.52
+cache_read = 0.0026
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 8324901e3db..a8bf0f0a408 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04752
-output = 0.14256
-cache_read = 0.001512
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
 context = 1_310_720
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index ec32a69e6bd..7031b748f98 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 context = 1_048_576

From 8fcdf73a0977a6e0b7b222c24556ac19a347f0da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 07:26:02 +0000
Subject: [PATCH 1653/2012] chore(sync): update Kilo model catalog (#7487)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml          |  6 +++---
 .../kilo/models/~deepseek/deepseek-flash-latest.toml   |  6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml     | 10 +++++-----
 providers/kilo/models/~moonshotai/kimi-latest.toml     |  6 +++---
 4 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 90c5d22728f..fd81e087ca6 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 0b2d5078274..9e0f74cdd43 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.135
-output = 0.54
-cache_read = 0.00405
+input = 0.13
+output = 0.52
+cache_read = 0.0026
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 87db0089581..c34bd2f2e08 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04752
-output = 0.14256
-cache_read = 0.001512
+input = 0.04
+output = 0.08
+cache_read = 0.008
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index f09092fb568..631c1758bcd 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.875
-output = 10.5
-cache_read = 0.2175
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 context = 1_048_576

From bfc571555f1f22a908b248c30e2be9a248eddaba Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Sat, 19 Sep 2026 04:07:21 -0400
Subject: [PATCH 1654/2012] update zen models

---
 providers/opencode/models/jev-1.13-free.toml | 7 +++++++
 providers/opencode/models/jev-1.13.toml      | 8 ++++++++
 2 files changed, 15 insertions(+)
 create mode 100644 providers/opencode/models/jev-1.13-free.toml
 create mode 100644 providers/opencode/models/jev-1.13.toml

diff --git a/providers/opencode/models/jev-1.13-free.toml b/providers/opencode/models/jev-1.13-free.toml
new file mode 100644
index 00000000000..3c2a23a4f39
--- /dev/null
+++ b/providers/opencode/models/jev-1.13-free.toml
@@ -0,0 +1,7 @@
+# Availability: https://opencode.ai/zen/v1/models
+base_model = "typesafe/jev-latest"
+name = "Jev 1.13 Free"
+
+[cost]
+input = 0
+output = 0
diff --git a/providers/opencode/models/jev-1.13.toml b/providers/opencode/models/jev-1.13.toml
new file mode 100644
index 00000000000..c31caeddd67
--- /dev/null
+++ b/providers/opencode/models/jev-1.13.toml
@@ -0,0 +1,8 @@
+# Availability: https://opencode.ai/zen/v1/models
+# Pricing: https://docs.typesafe.ai/models.md
+base_model = "typesafe/jev-latest"
+name = "Jev 1.13"
+
+[cost]
+input = 0.042
+output = 0

From 0cc3d9e4b02208767a532fd4c43a2a72892c77d9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 08:30:20 +0000
Subject: [PATCH 1655/2012] chore(sync): update OpenRouter model catalog
 (#7490)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 664a084d4ad..f3c7c5a82dc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04704
-output = 0.09408
-cache_read = 0.009408
+input = 0.04592
+output = 0.09184
+cache_read = 0.009184
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index f453f0bbdbd..b7faf4d187f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.564282
-output = 1.128564
-cache_read = 0.047024
+input = 0.533832
+output = 1.067664
+cache_read = 0.044486
 
 [limit]
 context = 1_048_576

From 66563f8f71f4077633ceaa18ec4196f48df43188 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 09:25:43 +0000
Subject: [PATCH 1656/2012] chore(sync): update OpenRouter model catalog
 (#7492)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml  | 2 +-
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 4 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 8dce0b17a1e..5e9065ff206 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -12,7 +12,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.04
 output = 0.08
-cache_read = 0.008
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f3c7c5a82dc..a93a1a5d132 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04592
-output = 0.09184
-cache_read = 0.009184
+input = 0.04536
+output = 0.09072
+cache_read = 0.009072
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b7faf4d187f..d8e373b8232 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.533832
-output = 1.067664
-cache_read = 0.044486
+input = 0.503382
+output = 1.006764
+cache_read = 0.041949
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index a8bf0f0a408..cd84b32afe3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.04
 output = 0.08
-cache_read = 0.008
+cache_read = 0.016
 
 [limit]
 context = 1_310_720

From a8604c4080b59f43b16bb85f31f70591a90f57d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 09:25:48 +0000
Subject: [PATCH 1657/2012] chore(sync): update Kilo model catalog (#7491)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index c34bd2f2e08..2c01541878c 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -17,7 +17,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.04
 output = 0.08
-cache_read = 0.008
+cache_read = 0.016
 
 [limit]
 context = 1_048_576

From b507a904ffd5408f3f207e9c073d9bc0e1f6c8a5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 09:26:21 +0000
Subject: [PATCH 1658/2012] chore(sync): update NanoGPT model catalog (#7493)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/{slowburn => google}/gemma4-31b-splituntied.toml     | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
 rename providers/nano-gpt/models/{slowburn => google}/gemma4-31b-splituntied.toml (65%)

diff --git a/providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml b/providers/nano-gpt/models/google/gemma4-31b-splituntied.toml
similarity index 65%
rename from providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml
rename to providers/nano-gpt/models/google/gemma4-31b-splituntied.toml
index 60615a27959..2a2440d5a87 100644
--- a/providers/nano-gpt/models/slowburn/gemma4-31b-splituntied.toml
+++ b/providers/nano-gpt/models/google/gemma4-31b-splituntied.toml
@@ -1,5 +1,5 @@
 name = "Gemma 4 31B Split-Untied"
-description = "Slowburn's Split-Untied is a text-only Gemma 4 31B community finetune with an untied BF16 output head, built for creative writing, roleplay, expressive dialogue, and tool use."
+description = "Blazed-Forge's Split-Untied is a text-only Gemma 4 31B community finetune with an untied BF16 output head, built for creative writing, roleplay, expressive dialogue, and tool use."
 family = "gemma"
 release_date = "2026-09-17"
 last_updated = "2026-09-17"

From 66b55e0e71c11c6cc3426c0d344f7293fbc8d14c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 10:24:15 +0000
Subject: [PATCH 1659/2012] chore(sync): update OpenRouter model catalog
 (#7496)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a93a1a5d132..be38a00704b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04536
-output = 0.09072
-cache_read = 0.009072
+input = 0.04508
+output = 0.09016
+cache_read = 0.009016
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index d8e373b8232..cf0b7555b6f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.503382
-output = 1.006764
-cache_read = 0.041949
+input = 0.49329
+output = 0.98658
+cache_read = 0.041108
 
 [limit]
 context = 1_048_576

From 19203ba786113081f7f6eb76720915f89bbc017a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 11:22:24 +0000
Subject: [PATCH 1660/2012] chore(sync): update OpenRouter model catalog
 (#7497)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index be38a00704b..99fb509a97e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04508
-output = 0.09016
-cache_read = 0.009016
+input = 0.04452
+output = 0.08904
+cache_read = 0.008904
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index cf0b7555b6f..8f63f2cd80d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.49329
-output = 0.98658
-cache_read = 0.041108
+input = 0.46284
+output = 0.92568
+cache_read = 0.03857
 
 [limit]
 context = 1_048_576

From 34885ce1a71bdd23359484a4cf1a40d6e2c143f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 12:32:48 +0000
Subject: [PATCH 1661/2012] chore(sync): update OpenRouter model catalog
 (#7499)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 99fb509a97e..638aaa8a0cd 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04452
-output = 0.08904
-cache_read = 0.008904
+input = 0.04368
+output = 0.08736
+cache_read = 0.008736
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 8f63f2cd80d..5bc792c3ae0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.46284
-output = 0.92568
-cache_read = 0.03857
+input = 0.422298
+output = 0.844596
+cache_read = 0.035192
 
 [limit]
 context = 1_048_576

From 2111a4c4ecb43abb4d7bb823e6b4fec2147731b1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 12:32:57 +0000
Subject: [PATCH 1662/2012] chore(sync): update NanoGPT model catalog (#7498)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../gemma-4-26b-a4b-it-cybersecurity.toml     | 25 +++++++++++++++++
 .../nvidia/nemotron-3.5-content-safety.toml   | 17 +++++++++++
 .../qwen/qwen3.8-27b-cybersecurity.toml       | 25 +++++++++++++++++
 .../models/qwen/qwen3.8-27b-uncensored.toml   |  6 ++--
 .../qwen/qwen3.8-27b-uncensored:thinking.toml |  6 ++--
 .../z-ai/glm-5.3-flash-cybersecurity.toml     | 28 +++++++++++++++++++
 6 files changed, 101 insertions(+), 6 deletions(-)
 create mode 100644 providers/nano-gpt/models/google/gemma-4-26b-a4b-it-cybersecurity.toml
 create mode 100644 providers/nano-gpt/models/nvidia/nemotron-3.5-content-safety.toml
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.8-27b-cybersecurity.toml
 create mode 100644 providers/nano-gpt/models/z-ai/glm-5.3-flash-cybersecurity.toml

diff --git a/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-cybersecurity.toml b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-cybersecurity.toml
new file mode 100644
index 00000000000..502d73081ea
--- /dev/null
+++ b/providers/nano-gpt/models/google/gemma-4-26b-a4b-it-cybersecurity.toml
@@ -0,0 +1,25 @@
+name = "Gemma 4 26B A4B Cybersecurity"
+description = "Gemma 4 26B A4B Cybersecurity is a cybersecurity-focused variant based on the uncensored model, with provider moderation for illegal activities. It supports optional reasoning, image understanding, tool calling, and a 262,144-token context window."
+family = "gemma"
+release_date = "2026-09-19"
+last_updated = "2026-09-19"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.1056
+output = 0.3344
+cache_read = 0.0528
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/nvidia/nemotron-3.5-content-safety.toml b/providers/nano-gpt/models/nvidia/nemotron-3.5-content-safety.toml
new file mode 100644
index 00000000000..8564f16dea6
--- /dev/null
+++ b/providers/nano-gpt/models/nvidia/nemotron-3.5-content-safety.toml
@@ -0,0 +1,17 @@
+base_model = "nvidia/nemotron-3.5-content-safety"
+attachment = false
+structured_output = false
+reasoning_options = []
+
+[cost]
+input = 0.05
+output = 0.15
+cache_read = 0.025
+
+[limit]
+context = 131_072
+input = 131_072
+output = 32_768
+
+[modalities]
+input = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-cybersecurity.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-cybersecurity.toml
new file mode 100644
index 00000000000..fcae44481c3
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-cybersecurity.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.8 27B Cybersecurity"
+description = "Qwen 3.8 27B Cybersecurity is a cybersecurity-focused variant based on the uncensored model, with provider moderation for illegal activities. It supports optional reasoning, image understanding, tool calling, and a 262,144-token context window."
+family = "qwen"
+release_date = "2026-09-19"
+last_updated = "2026-09-19"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.1
+output = 0.6
+cache_read = 0.05
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index 516d35fc982..e34ca7cb8aa 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.25
-output = 1.5
-cache_read = 0.125
+input = 0.2
+output = 1.7
+cache_read = 0.2
 
 [limit]
 context = 524_288
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 169511f2187..17f4a01df20 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.25
-output = 1.5
-cache_read = 0.125
+input = 0.2
+output = 1.7
+cache_read = 0.2
 
 [limit]
 context = 524_288
diff --git a/providers/nano-gpt/models/z-ai/glm-5.3-flash-cybersecurity.toml b/providers/nano-gpt/models/z-ai/glm-5.3-flash-cybersecurity.toml
new file mode 100644
index 00000000000..b55aaf3a6a5
--- /dev/null
+++ b/providers/nano-gpt/models/z-ai/glm-5.3-flash-cybersecurity.toml
@@ -0,0 +1,28 @@
+name = "GLM 5.3 Flash Cybersecurity"
+description = "GLM 5.3 Flash Cybersecurity is a cybersecurity-focused variant based on the uncensored model, with provider moderation for illegal activities. It supports always-on reasoning, image understanding, tool calling, and a 1,048,576-token context window."
+family = "glm"
+release_date = "2026-09-19"
+last_updated = "2026-09-19"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.075
+
+[limit]
+context = 1_048_576
+input = 1_048_576
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 797f773e891df6e8c800464d20a01bc694150e25 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 13:23:47 +0000
Subject: [PATCH 1663/2012] chore(sync): update OpenRouter model catalog
 (#7502)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 +++------
 2 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 638aaa8a0cd..ac1de47f47f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04368
-output = 0.08736
-cache_read = 0.008736
+input = 0.04312
+output = 0.08624
+cache_read = 0.008624
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index 9a813d7b44e..e562f006f6c 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,9 +7,6 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
-
-[limit]
-output = 16_384
+input = 0.5
+output = 2
+cache_read = 0.1

From 29e117b6ea7fa6dc95ec3cdc4a270ddaab8f973a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 13:23:59 +0000
Subject: [PATCH 1664/2012] chore(sync): update Kilo model catalog (#7503)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index 2a611ba49e7..fd68fe0c68b 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,10 +7,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.43
-output = 1.75
-cache_read = 0.08
+input = 0.5
+output = 2
+cache_read = 0.1
 
 [limit]
-context = 198_000
-output = 16_384
+context = 202_752

From add971a83271a43e7487626eec10131f14763ed7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 14:23:46 +0000
Subject: [PATCH 1665/2012] chore(sync): update NanoGPT model catalog (#7505)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml      | 2 +-
 .../nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index e34ca7cb8aa..612900717c2 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -13,7 +13,7 @@ reasoning_options = []
 [cost]
 input = 0.2
 output = 1.7
-cache_read = 0.2
+cache_read = 0.18
 
 [limit]
 context = 524_288
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 17f4a01df20..46cc8830338 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -13,7 +13,7 @@ reasoning_options = []
 [cost]
 input = 0.2
 output = 1.7
-cache_read = 0.2
+cache_read = 0.18
 
 [limit]
 context = 524_288

From d3acf770c2294aeda45e8892205dc624b378dd3f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 14:23:48 +0000
Subject: [PATCH 1666/2012] chore(sync): update OpenRouter model catalog
 (#7504)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ac1de47f47f..b4277c2cb62 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04312
-output = 0.08624
-cache_read = 0.008624
+input = 0.04284
+output = 0.08568
+cache_read = 0.008568
 
 [limit]
 context = 1_048_576

From 7d315b77805625381ea9f5c663fdcc019c99dcd9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 15:23:21 +0000
Subject: [PATCH 1667/2012] chore(sync): update OpenRouter model catalog
 (#7506)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b4277c2cb62..37587a85fb8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04284
-output = 0.08568
-cache_read = 0.008568
+input = 0.04228
+output = 0.08456
+cache_read = 0.008456
 
 [limit]
 context = 1_048_576

From b638683cd779bde5e76708f96b895075dc1a024d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 16:25:26 +0000
Subject: [PATCH 1668/2012] chore(sync): update LLM Gateway model catalog
 (#7508)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/vichar-ai/glm-5.3-flash.toml       | 20 -------------------
 .../models/vichar-ai/glm-5.3.toml             | 16 ---------------
 2 files changed, 36 deletions(-)
 delete mode 100644 providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml
 delete mode 100644 providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml

diff --git a/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml b/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml
deleted file mode 100644
index a8d0eb4cba1..00000000000
--- a/providers/llmgateway-providers/models/vichar-ai/glm-5.3-flash.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-base_model = "zhipuai/glm-5.3-flash"
-name = "GLM-5.3 Flash (vichar-ai)"
-attachment = false
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 0.15
-output = 0.5
-cache_read = 0.03
-
-[limit]
-context = 1_048_000
-output = 128_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml b/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml
deleted file mode 100644
index 561369dc9e2..00000000000
--- a/providers/llmgateway-providers/models/vichar-ai/glm-5.3.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "zhipuai/glm-5.3"
-name = "GLM-5.3 (vichar-ai)"
-structured_output = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
-
-[limit]
-context = 1_048_000
-output = 128_000

From b55fb9f10b8499c4d7bdfd63b100bce976d634b9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 16:25:32 +0000
Subject: [PATCH 1669/2012] chore(sync): update OpenRouter model catalog
 (#7507)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml   | 4 ++--
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 37587a85fb8..ff66884b828 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04228
-output = 0.08456
-cache_read = 0.008456
+input = 0.04172
+output = 0.08344
+cache_read = 0.008344
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 91402212177..d857fa53dab 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.7
+input = 0.13
+output = 0.52
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 6b9d0a0d2de2753979c3c503baae4f9d35026493 Mon Sep 17 00:00:00 2001
From: bpnrc <burak.pinarci@outlook.com>
Date: Sat, 19 Sep 2026 19:11:25 +0200
Subject: [PATCH 1670/2012] chore: update TensorX model catalog (add 6, remove
 7 retired) (#7486)

* chore: update TensorX model catalog (add 6, remove 7 retired)

Sync TensorX provider catalog against the live GET /v1/models endpoint.

Add (6):
- z-ai/glm-5.3 (effort low|high|max)
- z-ai/glm-5.3-flash (effort low|high|max)
- qwen/qwen3.8-2.4t-a95b (effort low|medium|xhigh)
- qwen/qwen3.8-27b (effort low|medium|xhigh)
- qwen/qwen3.8-flash-next (effort low|medium|xhigh)
- deepseek/deepseek-v4-pro-0813 (toggle thinking)

Remove (7, no longer served): deepseek-chat-v3.1, deepseek-v4-flash,
nemotron-3-super-120b-a12b, gpt-oss-120b, qwen3-coder-30b-a3b-instruct,
qwen3-vl-235b-a22b-instruct, glm-4.7.

Pricing, context/output limits and reasoning_options verified against the
live TensorX API (2026-09-19).

* fix: doc-accurate reasoning_options and override-only limits for new TensorX models

Align reasoning controls with docs.tensorx.ai/api-reference/reasoning:
- GLM 5.3 / 5.3-Flash: toggle enable_thinking + effort low|high
- DeepSeek V4 Pro 0813: toggle thinking + effort high|max
- Qwen 3.8: effort low|medium|xhigh (always-on)

Drop redundant [limit] overrides that merely restate the lab (Qwen 3.8
context/output) and keep only host deltas (output = 64000).

* fix: GLM-5.3 effort includes max; exact toggle wire paths

- GLM 5.3 / 5.3-Flash: effort low|high|max (docs + live probe confirm max
  is accepted; max is the default depth). Note that TensorX allows disabling
  thinking (enable_thinking) unlike first-party Z.AI.
- DeepSeek V4 Pro 0813: document the exact toggle wire path
  chat_template_kwargs.thinking in a leading header.

---------

Co-authored-by: bpnrc <bpnrc@users.noreply.github.com>
---
 .../models/deepseek/deepseek-chat-v3.1.toml   | 30 -------------------
 .../models/deepseek/deepseek-v4-flash.toml    | 13 --------
 .../models/deepseek/deepseek-v4-pro-0813.toml | 19 ++++++++++++
 .../nvidia/nemotron-3-super-120b-a12b.toml    | 11 -------
 .../tensorx/models/openai/gpt-oss-120b.toml   | 12 --------
 .../qwen/qwen3-coder-30b-a3b-instruct.toml    | 10 -------
 .../qwen/qwen3-vl-235b-a22b-instruct.toml     | 25 ----------------
 .../models/qwen/qwen3.8-2.4t-a95b.toml        | 15 ++++++++++
 .../tensorx/models/qwen/qwen3.8-27b.toml      | 11 +++++++
 .../models/qwen/qwen3.8-flash-next.toml       | 15 ++++++++++
 providers/tensorx/models/z-ai/glm-4.7.toml    | 19 ------------
 .../tensorx/models/z-ai/glm-5.3-flash.toml    | 20 +++++++++++++
 providers/tensorx/models/z-ai/glm-5.3.toml    | 20 +++++++++++++
 13 files changed, 100 insertions(+), 120 deletions(-)
 delete mode 100644 providers/tensorx/models/deepseek/deepseek-chat-v3.1.toml
 delete mode 100644 providers/tensorx/models/deepseek/deepseek-v4-flash.toml
 create mode 100644 providers/tensorx/models/deepseek/deepseek-v4-pro-0813.toml
 delete mode 100644 providers/tensorx/models/nvidia/nemotron-3-super-120b-a12b.toml
 delete mode 100644 providers/tensorx/models/openai/gpt-oss-120b.toml
 delete mode 100644 providers/tensorx/models/qwen/qwen3-coder-30b-a3b-instruct.toml
 delete mode 100644 providers/tensorx/models/qwen/qwen3-vl-235b-a22b-instruct.toml
 create mode 100644 providers/tensorx/models/qwen/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/tensorx/models/qwen/qwen3.8-27b.toml
 create mode 100644 providers/tensorx/models/qwen/qwen3.8-flash-next.toml
 delete mode 100644 providers/tensorx/models/z-ai/glm-4.7.toml
 create mode 100644 providers/tensorx/models/z-ai/glm-5.3-flash.toml
 create mode 100644 providers/tensorx/models/z-ai/glm-5.3.toml

diff --git a/providers/tensorx/models/deepseek/deepseek-chat-v3.1.toml b/providers/tensorx/models/deepseek/deepseek-chat-v3.1.toml
deleted file mode 100644
index 93143b88c68..00000000000
--- a/providers/tensorx/models/deepseek/deepseek-chat-v3.1.toml
+++ /dev/null
@@ -1,30 +0,0 @@
-name = "DeepSeek Chat V3.1"
-description = "DeepSeek chat model for instruction following, coding, and analysis"
-family = "deepseek"
-release_date = "2025-08-21"
-last_updated = "2025-08-21"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-knowledge = "2024-11"
-open_weights = true
-
-
-[[reasoning_options]]
-type = "effort" # API: {"reasoning_effort": <value>}; "none" disables reasoning
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.2
-output = 0.8
-cache_read = 0.05
-cache_write = 0.25
-
-[limit]
-context = 164000
-output = 163840
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/tensorx/models/deepseek/deepseek-v4-flash.toml b/providers/tensorx/models/deepseek/deepseek-v4-flash.toml
deleted file mode 100644
index fdea8b9a03d..00000000000
--- a/providers/tensorx/models/deepseek/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash"
-
-[[reasoning_options]]
-type = "toggle" # API: {"chat_template_kwargs": {"thinking": true}} (default off)
-
-[cost]
-input = 0.15
-output = 0.3
-cache_read = 0.0375
-cache_write = 0.1875
-
-[limit]
-context = 1048576
diff --git a/providers/tensorx/models/deepseek/deepseek-v4-pro-0813.toml b/providers/tensorx/models/deepseek/deepseek-v4-pro-0813.toml
new file mode 100644
index 00000000000..82492863b2c
--- /dev/null
+++ b/providers/tensorx/models/deepseek/deepseek-v4-pro-0813.toml
@@ -0,0 +1,19 @@
+base_model = "deepseek/deepseek-v4-pro-0813"
+# DeepSeek V4 reasons off by default; toggle chat_template_kwargs.thinking, effort high|max when on.
+# https://docs.tensorx.ai/api-reference/reasoning
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 2.0
+output = 4.0
+cache_read = 0.5
+
+[limit]
+context = 1048576
+output = 64000
\ No newline at end of file
diff --git a/providers/tensorx/models/nvidia/nemotron-3-super-120b-a12b.toml b/providers/tensorx/models/nvidia/nemotron-3-super-120b-a12b.toml
deleted file mode 100644
index 279eb86ea6a..00000000000
--- a/providers/tensorx/models/nvidia/nemotron-3-super-120b-a12b.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "nvidia/nemotron-3-super-120b-a12b"
-
-[[reasoning_options]]
-type = "effort" # API: {"reasoning_effort": <value>}; "none" disables reasoning
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.3
-output = 0.9
-cache_read = 0.075
-cache_write = 0.375
diff --git a/providers/tensorx/models/openai/gpt-oss-120b.toml b/providers/tensorx/models/openai/gpt-oss-120b.toml
deleted file mode 100644
index b34a24534e3..00000000000
--- a/providers/tensorx/models/openai/gpt-oss-120b.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "openai/gpt-oss-120b"
-knowledge = "2024-10"
-
-[[reasoning_options]]
-type = "effort" # API: {"reasoning_effort": <value>}; reasoning is mandatory, "none" is rejected
-values = ["minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.04
-output = 0.2
-cache_read = 0.01
-cache_write = 0.05
diff --git a/providers/tensorx/models/qwen/qwen3-coder-30b-a3b-instruct.toml b/providers/tensorx/models/qwen/qwen3-coder-30b-a3b-instruct.toml
deleted file mode 100644
index bf2ca2550b6..00000000000
--- a/providers/tensorx/models/qwen/qwen3-coder-30b-a3b-instruct.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "alibaba/qwen3-coder-30b-a3b-instruct"
-
-[cost]
-input = 0.06
-output = 0.25
-cache_read = 0.015
-cache_write = 0.075
-
-[limit]
-context = 262000
diff --git a/providers/tensorx/models/qwen/qwen3-vl-235b-a22b-instruct.toml b/providers/tensorx/models/qwen/qwen3-vl-235b-a22b-instruct.toml
deleted file mode 100644
index 63f03d54a44..00000000000
--- a/providers/tensorx/models/qwen/qwen3-vl-235b-a22b-instruct.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Qwen3 VL 235B-A22B Instruct"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-09-23"
-last_updated = "2025-09-23"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2025-03-31"
-open_weights = true
-
-[cost]
-input = 0.21
-output = 1.9
-cache_read = 0.0525
-cache_write = 0.2625
-
-[limit]
-context = 131000
-output = 131072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/tensorx/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/tensorx/models/qwen/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..70a027487ce
--- /dev/null
+++ b/providers/tensorx/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+# Qwen3.8 always reasons; effort xhigh (default) | medium | low.
+# https://docs.tensorx.ai/api-reference/reasoning
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 2.5
+output = 6.0
+cache_read = 0.63
+
+[limit]
+output = 64000
\ No newline at end of file
diff --git a/providers/tensorx/models/qwen/qwen3.8-27b.toml b/providers/tensorx/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..552bc1978f7
--- /dev/null
+++ b/providers/tensorx/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,11 @@
+base_model = "alibaba/qwen3.8-27b"
+# Thinking is always on; reasoning_effort low|medium|xhigh (default xhigh).
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.4
+output = 2.4
+cache_read = 0.1
\ No newline at end of file
diff --git a/providers/tensorx/models/qwen/qwen3.8-flash-next.toml b/providers/tensorx/models/qwen/qwen3.8-flash-next.toml
new file mode 100644
index 00000000000..7df5eb110ec
--- /dev/null
+++ b/providers/tensorx/models/qwen/qwen3.8-flash-next.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-flash-next"
+# Qwen3.8 always reasons; effort xhigh (default) | medium | low.
+# https://docs.tensorx.ai/api-reference/reasoning
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+output = 64000
\ No newline at end of file
diff --git a/providers/tensorx/models/z-ai/glm-4.7.toml b/providers/tensorx/models/z-ai/glm-4.7.toml
deleted file mode 100644
index 931c963e6a7..00000000000
--- a/providers/tensorx/models/z-ai/glm-4.7.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "zhipuai/glm-4.7"
-
-[[reasoning_options]]
-type = "effort" # API: {"reasoning_effort": <value>}; "none" disables reasoning
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 0.6
-output = 2.2
-cache_read = 0.15
-cache_write = 0.75
-
-[limit]
-context = 200000
-output = 200000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/tensorx/models/z-ai/glm-5.3-flash.toml b/providers/tensorx/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..ad6a7edac8e
--- /dev/null
+++ b/providers/tensorx/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,20 @@
+base_model = "zhipuai/glm-5.3-flash"
+# GLM-5.3-Flash reasons by default; toggle enable_thinking, effort low|high|max (default max).
+# Unlike first-party Z.AI, TensorX allows thinking to be disabled (enable_thinking).
+# https://docs.tensorx.ai/api-reference/reasoning
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.05
+
+[limit]
+context = 1048576
+output = 64000
\ No newline at end of file
diff --git a/providers/tensorx/models/z-ai/glm-5.3.toml b/providers/tensorx/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..85587ef550d
--- /dev/null
+++ b/providers/tensorx/models/z-ai/glm-5.3.toml
@@ -0,0 +1,20 @@
+base_model = "zhipuai/glm-5.3"
+# GLM-5.3 reasons by default; toggle enable_thinking, effort low|high|max (default max).
+# Unlike first-party Z.AI, TensorX allows thinking to be disabled (enable_thinking).
+# https://docs.tensorx.ai/api-reference/reasoning
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.75
+output = 4.5
+cache_read = 0.44
+
+[limit]
+context = 1048576
+output = 64000
\ No newline at end of file

From bc1e1a83ac95e718b48e4122d553ef0e706ffd82 Mon Sep 17 00:00:00 2001
From: Alcatraz-Zhang <zy.zhang760@gmail.com>
Date: Sun, 20 Sep 2026 01:11:44 +0800
Subject: [PATCH 1671/2012] chore(zenmux): refresh Z.AI and xAI families
 (#7484)

* chore(zenmux): refresh Z.AI and xAI families

* fix(zenmux): document Z.AI host deltas
---
 models/xai/grok-voice-stt-1.0.toml            | 17 ++++++++
 models/xai/grok-voice-tts-1.0.toml            | 17 ++++++++
 models/zhipuai/glm-image.toml                 | 22 ++++++++++
 providers/zenmux/models/x-ai/grok-4-fast.toml | 25 -----------
 .../x-ai/grok-4.1-fast-non-reasoning.toml     | 24 -----------
 .../zenmux/models/x-ai/grok-4.1-fast.toml     | 25 -----------
 .../x-ai/grok-4.2-fast-non-reasoning.toml     | 26 ++++++------
 .../zenmux/models/x-ai/grok-4.2-fast.toml     | 32 +++++++-------
 providers/zenmux/models/x-ai/grok-4.3.toml    | 23 +++++-----
 providers/zenmux/models/x-ai/grok-4.6.toml    | 19 +++++++++
 providers/zenmux/models/x-ai/grok-4.toml      | 25 -----------
 .../zenmux/models/x-ai/grok-build-0.1.toml    | 14 +++++--
 .../zenmux/models/x-ai/grok-code-fast-1.toml  | 25 -----------
 .../models/x-ai/grok-imagine-image-2.0.toml   |  7 ++++
 .../models/x-ai/grok-voice-stt-1.0.toml       |  3 ++
 .../models/x-ai/grok-voice-tts-1.0.toml       |  3 ++
 providers/zenmux/models/z-ai/glm-4.5-air.toml | 35 ++++++++--------
 providers/zenmux/models/z-ai/glm-4.5.toml     | 36 ++++++++--------
 providers/zenmux/models/z-ai/glm-4.6.toml     | 36 ++++++++--------
 .../models/z-ai/glm-4.6v-flash-free.toml      | 34 ++++++++-------
 .../zenmux/models/z-ai/glm-4.6v-flash.toml    | 36 ++++++++--------
 providers/zenmux/models/z-ai/glm-4.6v.toml    | 35 ++++++++--------
 .../models/z-ai/glm-4.7-flash-free.toml       | 35 +++++++---------
 .../zenmux/models/z-ai/glm-4.7-flashx.toml    | 36 +++++++---------
 providers/zenmux/models/z-ai/glm-4.7.toml     | 42 +++++++++----------
 providers/zenmux/models/z-ai/glm-5-turbo.toml | 36 ++++++++--------
 providers/zenmux/models/z-ai/glm-5.1.toml     | 32 +++++++-------
 .../zenmux/models/z-ai/glm-5.2-free.toml      | 11 -----
 providers/zenmux/models/z-ai/glm-5.2.toml     | 17 ++++++--
 .../zenmux/models/z-ai/glm-5.3-flash.toml     | 20 +++++++++
 .../zenmux/models/z-ai/glm-5.3-flashx.toml    | 20 +++++++++
 providers/zenmux/models/z-ai/glm-5.3.toml     | 20 +++++++++
 providers/zenmux/models/z-ai/glm-5.toml       | 28 ++++++-------
 .../zenmux/models/z-ai/glm-5v-turbo.toml      | 27 ++++++------
 providers/zenmux/models/z-ai/glm-image.toml   |  9 ++++
 35 files changed, 439 insertions(+), 413 deletions(-)
 create mode 100644 models/xai/grok-voice-stt-1.0.toml
 create mode 100644 models/xai/grok-voice-tts-1.0.toml
 create mode 100644 models/zhipuai/glm-image.toml
 delete mode 100644 providers/zenmux/models/x-ai/grok-4-fast.toml
 delete mode 100644 providers/zenmux/models/x-ai/grok-4.1-fast-non-reasoning.toml
 delete mode 100644 providers/zenmux/models/x-ai/grok-4.1-fast.toml
 create mode 100644 providers/zenmux/models/x-ai/grok-4.6.toml
 delete mode 100644 providers/zenmux/models/x-ai/grok-4.toml
 delete mode 100644 providers/zenmux/models/x-ai/grok-code-fast-1.toml
 create mode 100644 providers/zenmux/models/x-ai/grok-imagine-image-2.0.toml
 create mode 100644 providers/zenmux/models/x-ai/grok-voice-stt-1.0.toml
 create mode 100644 providers/zenmux/models/x-ai/grok-voice-tts-1.0.toml
 delete mode 100644 providers/zenmux/models/z-ai/glm-5.2-free.toml
 create mode 100644 providers/zenmux/models/z-ai/glm-5.3-flash.toml
 create mode 100644 providers/zenmux/models/z-ai/glm-5.3-flashx.toml
 create mode 100644 providers/zenmux/models/z-ai/glm-5.3.toml
 create mode 100644 providers/zenmux/models/z-ai/glm-image.toml

diff --git a/models/xai/grok-voice-stt-1.0.toml b/models/xai/grok-voice-stt-1.0.toml
new file mode 100644
index 00000000000..725c96c69d7
--- /dev/null
+++ b/models/xai/grok-voice-stt-1.0.toml
@@ -0,0 +1,17 @@
+name = "Grok Voice STT 1.0"
+description = "Grok Voice STT 1.0 is xAI's speech-to-text model. It supports transcription with word-level timestamps, optional speaker diarization, and multichannel audio."
+release_date = "2026-08-04"
+last_updated = "2026-08-04"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 15_000
+output = 15_000
+
+[modalities]
+input = ["audio"]
+output = ["text"]
diff --git a/models/xai/grok-voice-tts-1.0.toml b/models/xai/grok-voice-tts-1.0.toml
new file mode 100644
index 00000000000..b519e016a19
--- /dev/null
+++ b/models/xai/grok-voice-tts-1.0.toml
@@ -0,0 +1,17 @@
+name = "Grok Voice TTS 1.0"
+description = "Convert text into spoken audio with a single API call. The API supports a rich set of expressive voices, inline speech tags for fine-grained delivery control, and output formats from high-fidelity MP3 to telephony-optimized μ-law."
+release_date = "2026-07-31"
+last_updated = "2026-07-31"
+attachment = false
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 15_000
+output = 15_000
+
+[modalities]
+input = ["text"]
+output = ["audio"]
diff --git a/models/zhipuai/glm-image.toml b/models/zhipuai/glm-image.toml
new file mode 100644
index 00000000000..b3f3746cb2e
--- /dev/null
+++ b/models/zhipuai/glm-image.toml
@@ -0,0 +1,22 @@
+name = "GLM-Image"
+description = "GLM-Image is an image generation model adopts a hybrid autoregressive + diffusion decoder architecture. In general image generation quality, GLM‑Image aligns with mainstream latent diffusion approaches, but it shows significant advantages in text-rendering and knowledge‑intensive generation scenarios. It performs especially well in tasks requiring precise semantic understanding and complex information expression, while maintaining strong capabilities in high‑fidelity and fine‑grained detail generation. In addition to text‑to‑image generation, GLM‑Image also supports a rich set of image‑to‑image tasks including image editing, style transfer, identity‑preserving generation, and multi‑subject consistency."
+release_date = "2026-01-19"
+last_updated = "2026-01-19"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = true
+license = "MIT"
+
+[limit]
+context = 10_240
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/zai-org/GLM-Image"
diff --git a/providers/zenmux/models/x-ai/grok-4-fast.toml b/providers/zenmux/models/x-ai/grok-4-fast.toml
deleted file mode 100644
index f722ac7ac71..00000000000
--- a/providers/zenmux/models/x-ai/grok-4-fast.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Grok 4 Fast"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2025-09-19"
-last_updated = "2025-09-19"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-status = "deprecated"
-
-[cost]
-input = 0.20
-output = 0.50
-cache_read = 0.05
-
-[limit]
-context = 2000_000
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/zenmux/models/x-ai/grok-4.1-fast-non-reasoning.toml b/providers/zenmux/models/x-ai/grok-4.1-fast-non-reasoning.toml
deleted file mode 100644
index 997935a9745..00000000000
--- a/providers/zenmux/models/x-ai/grok-4.1-fast-non-reasoning.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Grok 4.1 Fast Non Reasoning"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2025-11-20"
-last_updated = "2025-11-20"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-status = "deprecated"
-
-[cost]
-input = 0.20
-output = 0.50
-cache_read = 0.05
-
-[limit]
-context = 2000_000
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/zenmux/models/x-ai/grok-4.1-fast.toml b/providers/zenmux/models/x-ai/grok-4.1-fast.toml
deleted file mode 100644
index 6ddced52aef..00000000000
--- a/providers/zenmux/models/x-ai/grok-4.1-fast.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Grok 4.1 Fast"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2025-11-20"
-last_updated = "2025-11-20"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-status = "deprecated"
-
-[cost]
-input = 0.20
-output = 0.50
-cache_read = 0.05
-
-[limit]
-context = 2000_000
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/zenmux/models/x-ai/grok-4.2-fast-non-reasoning.toml b/providers/zenmux/models/x-ai/grok-4.2-fast-non-reasoning.toml
index 5fd0f203453..b521181860f 100644
--- a/providers/zenmux/models/x-ai/grok-4.2-fast-non-reasoning.toml
+++ b/providers/zenmux/models/x-ai/grok-4.2-fast-non-reasoning.toml
@@ -1,22 +1,20 @@
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "xai/grok-4.20-0309-non-reasoning"
 name = "Grok 4.2 Fast Non Reasoning"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2026-03-20"
-last_updated = "2026-03-20"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2025-08-31"
-open_weights = false
 
 [cost]
-input = 3.00
-output = 9.00
+input = 2
+output = 6
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 4
+output = 12
+cache_read = 0.2
 
 [limit]
 context = 2_000_000
-output = 30_000
 
 [modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+input = ["text", "image"]
diff --git a/providers/zenmux/models/x-ai/grok-4.2-fast.toml b/providers/zenmux/models/x-ai/grok-4.2-fast.toml
index c768fe47135..fc524864b65 100644
--- a/providers/zenmux/models/x-ai/grok-4.2-fast.toml
+++ b/providers/zenmux/models/x-ai/grok-4.2-fast.toml
@@ -1,23 +1,25 @@
+# ZenMux's 2026-09-19 page has supports_reasoning=0 and both live OpenAI- and
+# Anthropic-compatible entries have capabilities.reasoning=false for this id.
+# This route therefore uses the reasoning checkpoint identity but disables the
+# reasoning surface as an explicit host delta.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "xai/grok-4.20-0309-reasoning"
 name = "Grok 4.2 Fast"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2026-03-20"
-last_updated = "2026-03-20"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-08-31"
-open_weights = false
+reasoning = false
 
 [cost]
-input = 3.00
-output = 9.00
+input = 2
+output = 6
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 4
+output = 12
+cache_read = 0.2
 
 [limit]
 context = 2_000_000
-output = 30_000
 
 [modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+input = ["text", "image"]
diff --git a/providers/zenmux/models/x-ai/grok-4.3.toml b/providers/zenmux/models/x-ai/grok-4.3.toml
index 455779665fb..ab8ed82e9a7 100644
--- a/providers/zenmux/models/x-ai/grok-4.3.toml
+++ b/providers/zenmux/models/x-ai/grok-4.3.toml
@@ -1,19 +1,22 @@
+# Effort: reasoning_effort = none|low|medium|high
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
 base_model = "xai/grok-4.3"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
 
 [cost]
 input = 1.25
-output = 2.50
-cache_read = 0.20
-cache_write = 0
+output = 2.5
+cache_read = 0.2
 
 [[cost.tiers]]
-tier = { size = 200_000 }
-input = 2.50
-output = 5.00
-cache_read = 0.40
-cache_write = 0
+tier = { type = "context", size = 200_000 }
+input = 2.5
+output = 5
+cache_read = 0.4
 
 [limit]
-context = 1_000_000
 output = 1_000_000
diff --git a/providers/zenmux/models/x-ai/grok-4.6.toml b/providers/zenmux/models/x-ai/grok-4.6.toml
new file mode 100644
index 00000000000..7b40b61706e
--- /dev/null
+++ b/providers/zenmux/models/x-ai/grok-4.6.toml
@@ -0,0 +1,19 @@
+# Effort: reasoning_effort = low|medium|high|xhigh
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "xai/grok-4.6"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
diff --git a/providers/zenmux/models/x-ai/grok-4.toml b/providers/zenmux/models/x-ai/grok-4.toml
deleted file mode 100644
index 9a3bd9acea0..00000000000
--- a/providers/zenmux/models/x-ai/grok-4.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Grok 4"
-description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
-release_date = "2025-07-09"
-last_updated = "2025-07-09"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-status = "deprecated"
-
-[cost]
-input = 3.00
-output = 15.00
-cache_read = 0.75
-
-[limit]
-context = 256_000
-output = 64_000
-
-[modalities]
-input = ["image", "text"]
-output = ["text"]
diff --git a/providers/zenmux/models/x-ai/grok-build-0.1.toml b/providers/zenmux/models/x-ai/grok-build-0.1.toml
index 59480b618c9..f00530ac65f 100644
--- a/providers/zenmux/models/x-ai/grok-build-0.1.toml
+++ b/providers/zenmux/models/x-ai/grok-build-0.1.toml
@@ -1,7 +1,13 @@
+# ZenMux's 2026-09-19 page has supports_reasoning=0 and both live OpenAI- and
+# Anthropic-compatible entries have capabilities.reasoning=false for this id.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
 base_model = "xai/grok-build-0.1"
-reasoning_options = []
+reasoning = false
 
 [cost]
-input = 1.00
-output = 2.00
-cache_read = 0.20
+input = 1
+output = 2
+cache_read = 0.2
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/zenmux/models/x-ai/grok-code-fast-1.toml b/providers/zenmux/models/x-ai/grok-code-fast-1.toml
deleted file mode 100644
index 5a692a5e5d9..00000000000
--- a/providers/zenmux/models/x-ai/grok-code-fast-1.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "Grok Code Fast 1"
-description = "Fast Grok model for responsive chat, reasoning, and tool-assisted work"
-release_date = "2025-08-26"
-last_updated = "2025-08-26"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-status = "deprecated"
-
-[cost]
-input = 0.20
-output = 1.50
-cache_read = 0.02
-
-[limit]
-context = 256_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/zenmux/models/x-ai/grok-imagine-image-2.0.toml b/providers/zenmux/models/x-ai/grok-imagine-image-2.0.toml
new file mode 100644
index 00000000000..05abc6d0374
--- /dev/null
+++ b/providers/zenmux/models/x-ai/grok-imagine-image-2.0.toml
@@ -0,0 +1,7 @@
+# Host context: page and OpenAI list both report 66000 tokens.
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "xai/grok-imagine-image-2.0"
+
+[limit]
+context = 66_000
diff --git a/providers/zenmux/models/x-ai/grok-voice-stt-1.0.toml b/providers/zenmux/models/x-ai/grok-voice-stt-1.0.toml
new file mode 100644
index 00000000000..c4bf10390e9
--- /dev/null
+++ b/providers/zenmux/models/x-ai/grok-voice-stt-1.0.toml
@@ -0,0 +1,3 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "xai/grok-voice-stt-1.0"
diff --git a/providers/zenmux/models/x-ai/grok-voice-tts-1.0.toml b/providers/zenmux/models/x-ai/grok-voice-tts-1.0.toml
new file mode 100644
index 00000000000..27f8f70b699
--- /dev/null
+++ b/providers/zenmux/models/x-ai/grok-voice-tts-1.0.toml
@@ -0,0 +1,3 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "xai/grok-voice-tts-1.0"
diff --git a/providers/zenmux/models/z-ai/glm-4.5-air.toml b/providers/zenmux/models/z-ai/glm-4.5-air.toml
index 756ca62b321..3797e830f0e 100644
--- a/providers/zenmux/models/z-ai/glm-4.5-air.toml
+++ b/providers/zenmux/models/z-ai/glm-4.5-air.toml
@@ -1,24 +1,23 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.5-air"
 name = "GLM 4.5 Air"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-release_date = "2025-07-25"
-last_updated = "2025-07-25"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.11
-output = 0.56
-cache_read = 0.02
+input = 0.1165
+output = 0.2911
+cache_read = 0.0233
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.1747
+output = 1.1645
+cache_read = 0.0349
 
 [limit]
 context = 128_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 96_000
diff --git a/providers/zenmux/models/z-ai/glm-4.5.toml b/providers/zenmux/models/z-ai/glm-4.5.toml
index b2f7197b924..fffd19cacc1 100644
--- a/providers/zenmux/models/z-ai/glm-4.5.toml
+++ b/providers/zenmux/models/z-ai/glm-4.5.toml
@@ -1,24 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.5"
 name = "GLM 4.5"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-release_date = "2025-07-25"
-last_updated = "2025-07-25"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.35
-output = 1.54
-cache_read = 0.07
+input = 0.2911
+output = 1.1645
+cache_read = 0.0582
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.5823
+output = 2.3291
+cache_read = 0.1165
 
 [limit]
 context = 128_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 96_000
diff --git a/providers/zenmux/models/z-ai/glm-4.6.toml b/providers/zenmux/models/z-ai/glm-4.6.toml
index e5f9d2ca1a4..d4d4456a446 100644
--- a/providers/zenmux/models/z-ai/glm-4.6.toml
+++ b/providers/zenmux/models/z-ai/glm-4.6.toml
@@ -1,24 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.6"
 name = "GLM 4.6"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-release_date = "2025-09-30"
-last_updated = "2025-09-30"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.35
-output = 1.54
-cache_read = 0.07
+input = 0.2911
+output = 1.1645
+cache_read = 0.0582
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.5823
+output = 2.3291
+cache_read = 0.1165
 
 [limit]
 context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-4.6v-flash-free.toml b/providers/zenmux/models/z-ai/glm-4.6v-flash-free.toml
index 33b4db29d44..18b34502274 100644
--- a/providers/zenmux/models/z-ai/glm-4.6v-flash-free.toml
+++ b/providers/zenmux/models/z-ai/glm-4.6v-flash-free.toml
@@ -1,23 +1,27 @@
+# Toggle: reasoning.enabled = true|false
+# Host modalities: the live free route includes file/PDF input; the paid route omits it.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.6v-flash"
 name = "GLM 4.6V Flash (Free)"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.00
-output = 0.00
+input = 0
+output = 0
+cache_read = 0
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0
+output = 0
+cache_read = 0
 
 [limit]
 context = 200_000
-output = 64_000
+output = 128_000
 
 [modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+input = ["text", "image", "video", "pdf"]
diff --git a/providers/zenmux/models/z-ai/glm-4.6v-flash.toml b/providers/zenmux/models/z-ai/glm-4.6v-flash.toml
index 00a8fa34b67..6c53afd51b0 100644
--- a/providers/zenmux/models/z-ai/glm-4.6v-flash.toml
+++ b/providers/zenmux/models/z-ai/glm-4.6v-flash.toml
@@ -1,24 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# Host naming: the 2026-09-19 live route z-ai/glm-4.6v-flash is labeled "GLM 4.6V FlashX".
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.6v-flash"
 name = "GLM 4.6V FlashX"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.02
-output = 0.21
-cache_read = 0.0043
+input = 0.0218
+output = 0.2184
+cache_read = 0.0044
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.0437
+output = 0.4367
+cache_read = 0.0044
 
 [limit]
 context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-4.6v.toml b/providers/zenmux/models/z-ai/glm-4.6v.toml
index 73690ef9e4f..e8c345954c8 100644
--- a/providers/zenmux/models/z-ai/glm-4.6v.toml
+++ b/providers/zenmux/models/z-ai/glm-4.6v.toml
@@ -1,24 +1,23 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.6v"
 name = "GLM 4.6V"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.14
-output = 0.42
-cache_read = 0.03
+input = 0.1456
+output = 0.4367
+cache_read = 0.0291
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.2911
+output = 0.8734
+cache_read = 0.0582
 
 [limit]
 context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-4.7-flash-free.toml b/providers/zenmux/models/z-ai/glm-4.7-flash-free.toml
index b2ddbce96ec..df0e92ac02a 100644
--- a/providers/zenmux/models/z-ai/glm-4.7-flash-free.toml
+++ b/providers/zenmux/models/z-ai/glm-4.7-flash-free.toml
@@ -1,26 +1,19 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.7-flash"
 name = "GLM 4.7 Flash (Free)"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-
-[cost]
-input = 0.00
-output = 0.00
-
-[limit]
-context = 200_000
-output = 64_000
 
 [interleaved]
 field = "reasoning_content"
 
-[modalities]
-input = ["text"]
-output = ["text"]
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-4.7-flashx.toml b/providers/zenmux/models/z-ai/glm-4.7-flashx.toml
index eba199ff720..b131a75aa2a 100644
--- a/providers/zenmux/models/z-ai/glm-4.7-flashx.toml
+++ b/providers/zenmux/models/z-ai/glm-4.7-flashx.toml
@@ -1,27 +1,19 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.7-flashx"
 name = "GLM 4.7 FlashX"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
-
-[cost]
-input = 0.07
-output = 0.42
-cache_read = 0.01
-
-[limit]
-context = 200_000
-output = 64_000
 
 [interleaved]
 field = "reasoning_content"
 
-[modalities]
-input = ["text"]
-output = ["text"]
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.0728
+output = 0.4367
+cache_read = 0.0146
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-4.7.toml b/providers/zenmux/models/z-ai/glm-4.7.toml
index 266005dda7e..c32d76030a6 100644
--- a/providers/zenmux/models/z-ai/glm-4.7.toml
+++ b/providers/zenmux/models/z-ai/glm-4.7.toml
@@ -1,27 +1,27 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-4.7"
 name = "GLM 4.7"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-release_date = "2025-12-23"
-last_updated = "2025-12-23"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.28
-output = 1.14
-cache_read = 0.06
+input = 0.2911
+output = 1.1645
+cache_read = 0.0582
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.5823
+output = 2.3291
+cache_read = 0.1165
 
 [limit]
 context = 200_000
-output = 64_000
-
-[interleaved]
-field = "reasoning_content"
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5-turbo.toml b/providers/zenmux/models/z-ai/glm-5-turbo.toml
index ef6438e58ac..ecc0fbae95d 100644
--- a/providers/zenmux/models/z-ai/glm-5-turbo.toml
+++ b/providers/zenmux/models/z-ai/glm-5-turbo.toml
@@ -1,23 +1,25 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5-turbo"
 name = "GLM 5 Turbo"
-description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-release_date = "2026-03-20"
-last_updated = "2026-03-20"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0.88
-output = 3.48
+input = 0.73
+output = 3.19
+cache_read = 0.174
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.02
+output = 3.77
+cache_read = 0.261
 
 [limit]
-context = 200_000
 output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/zenmux/models/z-ai/glm-5.1.toml b/providers/zenmux/models/z-ai/glm-5.1.toml
index 3b4c0bec049..3a64eedea4a 100644
--- a/providers/zenmux/models/z-ai/glm-5.1.toml
+++ b/providers/zenmux/models/z-ai/glm-5.1.toml
@@ -1,27 +1,25 @@
-name = "GLM-5.1"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-release_date = "2026-04-03"
-last_updated = "2026-04-03"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5.1"
+name = "GLM 5.1"
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.8781
 output = 3.5126
 cache_read = 0.1903
 
-[limit]
-context = 200000
-output = 131072
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.1709
+output = 4.098
+cache_read = 0.2927
 
-[modalities]
-input = ["text"]
-output = ["text"]
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5.2-free.toml b/providers/zenmux/models/z-ai/glm-5.2-free.toml
deleted file mode 100644
index 561b079445e..00000000000
--- a/providers/zenmux/models/z-ai/glm-5.2-free.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "zhipuai/glm-5.2"
-name = "GLM 5.2 (Free)"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
diff --git a/providers/zenmux/models/z-ai/glm-5.2.toml b/providers/zenmux/models/z-ai/glm-5.2.toml
index 7d3e1a5ebf4..09a38c80a35 100644
--- a/providers/zenmux/models/z-ai/glm-5.2.toml
+++ b/providers/zenmux/models/z-ai/glm-5.2.toml
@@ -1,11 +1,20 @@
+# Effort: reasoning_effort = high|max
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2"
 
-[cost]
-input = 1.40
-output = 4.50
-cache_read = 0.26
+[interleaved]
+field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
 values = ["high", "max"]
+
+[cost]
+input = 0.98
+output = 3.08
+cache_read = 0.182
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5.3-flash.toml b/providers/zenmux/models/z-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..3bd77e65bd9
--- /dev/null
+++ b/providers/zenmux/models/z-ai/glm-5.3-flash.toml
@@ -0,0 +1,20 @@
+# Effort: reasoning_effort = low|high|max
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 Flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5.3-flashx.toml b/providers/zenmux/models/z-ai/glm-5.3-flashx.toml
new file mode 100644
index 00000000000..50ad6eb5ea5
--- /dev/null
+++ b/providers/zenmux/models/z-ai/glm-5.3-flashx.toml
@@ -0,0 +1,20 @@
+# Effort: reasoning_effort = low|high|max
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM 5.3 FlashX"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.375
+output = 1.25
+cache_read = 0.075
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5.3.toml b/providers/zenmux/models/z-ai/glm-5.3.toml
new file mode 100644
index 00000000000..18b45ff89ad
--- /dev/null
+++ b/providers/zenmux/models/z-ai/glm-5.3.toml
@@ -0,0 +1,20 @@
+# Effort: reasoning_effort = low|high|max
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+output = 128_000
diff --git a/providers/zenmux/models/z-ai/glm-5.toml b/providers/zenmux/models/z-ai/glm-5.toml
index 3be16bf629a..f2ee297f1d0 100644
--- a/providers/zenmux/models/z-ai/glm-5.toml
+++ b/providers/zenmux/models/z-ai/glm-5.toml
@@ -1,27 +1,27 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5"
 name = "GLM 5"
-description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-release_date = "2026-02-12"
-last_updated = "2026-02-12"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = true
+structured_output = true
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.58
 output = 2.6
 cache_read = 0.14
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.87
+output = 3.18
+cache_read = 0.22
+
 [limit]
 context = 200_000
 output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/zenmux/models/z-ai/glm-5v-turbo.toml b/providers/zenmux/models/z-ai/glm-5v-turbo.toml
index 00fc2f71901..0e75e57e194 100644
--- a/providers/zenmux/models/z-ai/glm-5v-turbo.toml
+++ b/providers/zenmux/models/z-ai/glm-5v-turbo.toml
@@ -1,26 +1,25 @@
+# Toggle: reasoning.enabled = true|false
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "zhipuai/glm-5v-turbo"
 name = "GLM 5V Turbo"
-description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-release_date = "2026-04-01"
-last_updated = "2026-04-01"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-open_weights = false
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.726
 output = 3.1946
 cache_read = 0.1743
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.0165
+output = 3.7754
+cache_read = 0.2614
+
 [limit]
-context = 200_000
 output = 128_000
-
-[modalities]
-input = ["text", "image", "video", "pdf"]
-output = ["text"]
diff --git a/providers/zenmux/models/z-ai/glm-image.toml b/providers/zenmux/models/z-ai/glm-image.toml
new file mode 100644
index 00000000000..79c6282b394
--- /dev/null
+++ b/providers/zenmux/models/z-ai/glm-image.toml
@@ -0,0 +1,9 @@
+# ZenMux's 2026-09-19 page and live Google-compatible list expose text input only;
+# image editing exists in the lab model but is unavailable on this hosted route.
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "zhipuai/glm-image"
+attachment = false
+
+[modalities]
+input = ["text"]

From 55a53aa2df138ce8a26af0e198025fa41185c219 Mon Sep 17 00:00:00 2001
From: Sujee Maniyam <sujee@nebius.com>
Date: Sat, 19 Sep 2026 10:12:03 -0700
Subject: [PATCH 1672/2012] feat(nebius): add DeepSeek V4.1 Flash, V4 Pro 0813,
 GLM-5.3 (#7477)

---
 .../deepseek-ai/DeepSeek-V4-Pro-0813.toml     | 28 +++++++++++++++++
 .../deepseek-ai/DeepSeek-V4.1-Flash.toml      | 30 +++++++++++++++++++
 providers/nebius/models/zai-org/GLM-5.3.toml  | 28 +++++++++++++++++
 3 files changed, 86 insertions(+)
 create mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
 create mode 100644 providers/nebius/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
 create mode 100644 providers/nebius/models/zai-org/GLM-5.3.toml

diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
new file mode 100644
index 00000000000..b439ca31a66
--- /dev/null
+++ b/providers/nebius/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -0,0 +1,28 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/api/public/models_info (pricing, context, max length)
+# Accessed 2026-09-18.
+# Effort: reasoning_effort = none|low|high|max (none = thinking off). Same gateway
+# behavior as the other DeepSeek V4 entries on this host (DeepSeek-V4-Flash-0731,
+# DeepSeek-V4-Pro): the generic schema accepts none/minimal/low/medium/high/xhigh/max,
+# but the mid tiers collapse (minimal=low=medium, high=xhigh) and none disables
+# thinking, leaving none/low/high/max as the distinct levels.
+# Served window: models_info max_model_len = 979_000 for this snapshot with no separate
+# output cap, so context and output both take the full served window.
+# No discounted prompt-cache tier is published in the public catalog: cached and fresh
+# input are billed identically, so cache_read = input (same rationale as
+# DeepSeek-V4-Flash-0731 / Kimi-K3 / GLM-5.3-Flash).
+base_model = "deepseek/deepseek-v4-pro-0813"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.32
+output = 3.96
+cache_read = 1.32
+
+[limit]
+context = 979_000
+output = 979_000
diff --git a/providers/nebius/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/nebius/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..cc631b9cfee
--- /dev/null
+++ b/providers/nebius/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,30 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/api/public/models_info (pricing, context, max length)
+# Accessed 2026-09-18.
+# Effort: reasoning_effort = none|low|high|max (none = thinking off). Same gateway
+# behavior as the other DeepSeek V4 entries on this host (DeepSeek-V4-Flash-0731,
+# DeepSeek-V4-Pro): the generic schema accepts none/minimal/low/medium/high/xhigh/max,
+# but the mid tiers collapse (minimal=low=medium, high=xhigh) and none disables
+# thinking, leaving none/low/high/max as the distinct levels.
+# Server reports model type image2text, so attachment/modalities are inherited from the
+# lab entry.
+# Served window: models_info max_model_len = 1_048_000 for this model with no separate
+# output cap, so context and output both take the full served window.
+# No discounted prompt-cache tier is published in the public catalog: cached and fresh
+# input are billed identically, so cache_read = input (same rationale as
+# DeepSeek-V4-Flash-0731 / Kimi-K3 / GLM-5.3-Flash).
+base_model = "deepseek/deepseek-v4.1-flash"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.3
+
+[limit]
+context = 1_048_000
+output = 1_048_000
diff --git a/providers/nebius/models/zai-org/GLM-5.3.toml b/providers/nebius/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..be8b6208f43
--- /dev/null
+++ b/providers/nebius/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,28 @@
+# Sources:
+# - https://tokenfactory.nebius.com/ (model catalog)
+# - https://tokenfactory.nebius.com/api/public/models_info (pricing, context, max length)
+# - https://docs.bigmodel.cn/cn/guide/models/text/glm-5.3 (effort set; thinking cannot
+#   be disabled, so there is no toggle — effort per lab: low|high|max, default max)
+# Accessed 2026-09-18.
+# Served window: models_info max_model_len = 1_024_000 with no separate output cap, so
+# context and output both take the full served window.
+# No discounted prompt-cache tier is published in the public catalog: cached and fresh
+# input are billed identically, so cache_read = input (same rationale as
+# GLM-5.3-Flash / DeepSeek-V4-Flash-0731 / Kimi-K3).
+base_model = "zhipuai/glm-5.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 1.4
+
+[limit]
+context = 1_024_000
+output = 1_024_000

From a55d9f67599f0dd5d974913d68622777122b787b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 12:13:22 -0500
Subject: [PATCH 1673/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7509)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): normalize synced model metadata

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/mixedbread/toast-1.toml                | 19 +++++++++++++++++++
 models/quiverai/arrow-2-telos.toml            | 19 +++++++++++++++++++
 models/quiverai/arrow-2.toml                  | 19 +++++++++++++++++++
 .../vercel/models/mixedbread/toast-1.toml     |  6 ++++++
 .../vercel/models/quiverai/arrow-2-telos.toml | 13 +++++++++++++
 providers/vercel/models/quiverai/arrow-2.toml | 12 ++++++++++++
 providers/vercel/models/typesafe-ai/jev.toml  |  3 +++
 7 files changed, 91 insertions(+)
 create mode 100644 models/mixedbread/toast-1.toml
 create mode 100644 models/quiverai/arrow-2-telos.toml
 create mode 100644 models/quiverai/arrow-2.toml
 create mode 100644 providers/vercel/models/mixedbread/toast-1.toml
 create mode 100644 providers/vercel/models/quiverai/arrow-2-telos.toml
 create mode 100644 providers/vercel/models/quiverai/arrow-2.toml

diff --git a/models/mixedbread/toast-1.toml b/models/mixedbread/toast-1.toml
new file mode 100644
index 00000000000..775acb667f6
--- /dev/null
+++ b/models/mixedbread/toast-1.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-19):
+# https://www.mixedbread.com/docs/agent/models
+# https://www.mixedbread.com/docs/agent/chat-completions
+name = "Toast 1"
+description = "Specialized search model for knowledge-intensive questions, multi-step retrieval, and evidence synthesis"
+release_date = "2026-08-13"
+last_updated = "2026-08-13"
+attachment = false
+reasoning = false
+tool_call = true
+open_weights = false
+
+[limit]
+context = 131_000
+output = 4_000
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/models/quiverai/arrow-2-telos.toml b/models/quiverai/arrow-2-telos.toml
new file mode 100644
index 00000000000..ae0e17b52d6
--- /dev/null
+++ b/models/quiverai/arrow-2-telos.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-19):
+# https://docs.quiver.ai/developers/models
+# https://docs.quiver.ai/developers/models/arrow-2-telos
+name = "Arrow 2 Telos"
+description = "High-fidelity SVG generation model for complex vector work and long-context refinement"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_050_000
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text", "image"]
diff --git a/models/quiverai/arrow-2.toml b/models/quiverai/arrow-2.toml
new file mode 100644
index 00000000000..5d73a411794
--- /dev/null
+++ b/models/quiverai/arrow-2.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-19):
+# https://docs.quiver.ai/developers/models
+# https://docs.quiver.ai/developers/models/arrow-2
+name = "Arrow 2"
+description = "Fast SVG generation model for creation, vectorization, editing, and animation"
+release_date = "2026-09-16"
+last_updated = "2026-09-16"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 131_072
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text", "image"]
diff --git a/providers/vercel/models/mixedbread/toast-1.toml b/providers/vercel/models/mixedbread/toast-1.toml
new file mode 100644
index 00000000000..b50f8e55367
--- /dev/null
+++ b/providers/vercel/models/mixedbread/toast-1.toml
@@ -0,0 +1,6 @@
+base_model = "mixedbread/toast-1"
+
+[cost]
+input = 0.3
+output = 0.72
+cache_read = 0.036
diff --git a/providers/vercel/models/quiverai/arrow-2-telos.toml b/providers/vercel/models/quiverai/arrow-2-telos.toml
new file mode 100644
index 00000000000..59dae7b0880
--- /dev/null
+++ b/providers/vercel/models/quiverai/arrow-2-telos.toml
@@ -0,0 +1,13 @@
+# https://docs.quiver.ai/api-reference/openresponses/createopenresponse
+base_model = "quiverai/arrow-2-telos"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 6
+output = 30
+cache_read = 0.6
+cache_write = 7.5
+
+[limit]
+context = 131_072
+output = 131_072
diff --git a/providers/vercel/models/quiverai/arrow-2.toml b/providers/vercel/models/quiverai/arrow-2.toml
new file mode 100644
index 00000000000..2b7145c9081
--- /dev/null
+++ b/providers/vercel/models/quiverai/arrow-2.toml
@@ -0,0 +1,12 @@
+# https://docs.quiver.ai/api-reference/openresponses/createopenresponse
+base_model = "quiverai/arrow-2"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+output = 131_072
diff --git a/providers/vercel/models/typesafe-ai/jev.toml b/providers/vercel/models/typesafe-ai/jev.toml
index e3c1d0652ad..f18f60fb34b 100644
--- a/providers/vercel/models/typesafe-ai/jev.toml
+++ b/providers/vercel/models/typesafe-ai/jev.toml
@@ -3,3 +3,6 @@ base_model = "typesafe/jev-latest"
 [cost]
 input = 0.042
 output = 0
+
+[limit]
+context = 32_000

From a685c6900d143d3dd381be35f04bf5feb7f045ab Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 12:15:59 -0500
Subject: [PATCH 1674/2012] =?UTF-8?q?fix:=20MiniMax=20China=20API=20endpoi?=
 =?UTF-8?q?nt=20migrated=20from=20api.minimaxi.com=20to=20api.minimax.cn?=
 =?UTF-8?q?=20=E2=80=94=20update=20minimax-cn=20providers=20(#7501)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/minimax-cn-coding-plan/provider.toml | 9 +++++++--
 providers/minimax-cn/provider.toml             | 9 +++++++--
 2 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/providers/minimax-cn-coding-plan/provider.toml b/providers/minimax-cn-coding-plan/provider.toml
index b72447a7d59..28bd4e0b23d 100644
--- a/providers/minimax-cn-coding-plan/provider.toml
+++ b/providers/minimax-cn-coding-plan/provider.toml
@@ -1,5 +1,10 @@
-name = "MiniMax Token Plan (minimaxi.com)"
+# China API host migrated from api.minimaxi.com → api.minimax.cn
+# Sources:
+# - https://platform.minimaxi.com/docs/guides/quickstart-preparation
+# - https://platform.minimaxi.com/docs/token-plan/other-tools
+# Anthropic-compatible base: https://api.minimax.cn/anthropic
+name = "MiniMax Token Plan (minimax.cn)"
 env = ["MINIMAX_API_KEY"]
 npm = "@ai-sdk/anthropic"
 doc = "https://platform.minimaxi.com/docs/token-plan/intro"
-api = "https://api.minimaxi.com/anthropic/v1"
+api = "https://api.minimax.cn/anthropic/v1"
diff --git a/providers/minimax-cn/provider.toml b/providers/minimax-cn/provider.toml
index a72226e7345..2a75ce8de3a 100644
--- a/providers/minimax-cn/provider.toml
+++ b/providers/minimax-cn/provider.toml
@@ -1,5 +1,10 @@
-name = "MiniMax (minimaxi.com)"
+# China API host migrated from api.minimaxi.com → api.minimax.cn
+# Sources:
+# - https://platform.minimaxi.com/docs/guides/quickstart-preparation
+# - https://platform.minimaxi.com/docs/api-reference/text-anthropic-api
+# Anthropic-compatible base: https://api.minimax.cn/anthropic
+name = "MiniMax (minimax.cn)"
 env = ["MINIMAX_API_KEY"]
 npm = "@ai-sdk/anthropic"
 doc = "https://platform.minimaxi.com/docs/guides/quickstart"
-api = "https://api.minimaxi.com/anthropic/v1"
+api = "https://api.minimax.cn/anthropic/v1"

From ceebaa8234220ca37e83517dc58a10c95c15d3b1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 17:21:30 +0000
Subject: [PATCH 1675/2012] chore(sync): update NanoGPT model catalog (#7510)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/diffusiongemma.toml         | 27 +++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 providers/nano-gpt/models/google/diffusiongemma.toml

diff --git a/providers/nano-gpt/models/google/diffusiongemma.toml b/providers/nano-gpt/models/google/diffusiongemma.toml
new file mode 100644
index 00000000000..244dd04defb
--- /dev/null
+++ b/providers/nano-gpt/models/google/diffusiongemma.toml
@@ -0,0 +1,27 @@
+name = "DiffusionGemma"
+description = "DiffusionGemma is a high-speed diffusion-based version of Gemma 4 26B A4B. It supports optional reasoning and a 262,144-token context window."
+release_date = "2026-09-19"
+last_updated = "2026-09-19"
+attachment = false
+reasoning = true
+tool_call = false
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "xhigh"]
+
+[cost]
+input = 0.05
+output = 0.15
+cache_read = 0.025
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 9f4c5f6929250e13482037e71d5ff0546c67d0c2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 17:21:35 +0000
Subject: [PATCH 1676/2012] chore(sync): update OpenRouter model catalog
 (#7511)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml    | 6 +++---
 providers/openrouter/models/z-ai/glm-4.6.toml            | 9 ++++++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index ff66884b828..6d4eb7a48b8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04172
-output = 0.08344
-cache_read = 0.008344
+input = 0.04116
+output = 0.08232
+cache_read = 0.008232
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-4.6.toml b/providers/openrouter/models/z-ai/glm-4.6.toml
index e562f006f6c..9a813d7b44e 100644
--- a/providers/openrouter/models/z-ai/glm-4.6.toml
+++ b/providers/openrouter/models/z-ai/glm-4.6.toml
@@ -7,6 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
+
+[limit]
+output = 16_384

From e68a3aad18efe08a2bf44401eefedd2a9df0f9f5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 17:22:09 +0000
Subject: [PATCH 1677/2012] chore(sync): update Kilo model catalog (#7512)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml  | 6 +++---
 providers/kilo/models/z-ai/glm-4.6.toml | 9 +++++----
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/z-ai/glm-4.6.toml b/providers/kilo/models/z-ai/glm-4.6.toml
index fd68fe0c68b..2a611ba49e7 100644
--- a/providers/kilo/models/z-ai/glm-4.6.toml
+++ b/providers/kilo/models/z-ai/glm-4.6.toml
@@ -7,9 +7,10 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.5
-output = 2
-cache_read = 0.1
+input = 0.43
+output = 1.75
+cache_read = 0.08
 
 [limit]
-context = 202_752
+context = 198_000
+output = 16_384

From b6e376eda74813bbfe3bf907b0ad99c8198017d2 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Sat, 19 Sep 2026 12:36:09 -0500
Subject: [PATCH 1678/2012] Update reasoning_options in muse-spark-1.3.toml

Removed 'max' from reasoning_options values.
---
 providers/opencode/models/muse-spark-1.3.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/opencode/models/muse-spark-1.3.toml b/providers/opencode/models/muse-spark-1.3.toml
index cfaa3f534d3..80f0b886543 100644
--- a/providers/opencode/models/muse-spark-1.3.toml
+++ b/providers/opencode/models/muse-spark-1.3.toml
@@ -1,5 +1,5 @@
 base_model = "meta/muse-spark-1.3"
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh", "max"] }]
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
 
 [cost]
 input = 1.25

From d68788d330a64dc6a3edf0c165db9d65747a2314 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 18:27:34 +0000
Subject: [PATCH 1679/2012] chore(sync): update Kilo model catalog (#7513)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 73bfc195431..c5af7ff24a4 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8918
-output = 2.8028
-cache_read = 0.16562
+input = 0.8442
+output = 2.6532
+cache_read = 0.15678
 
 [limit]
 context = 1_048_576

From c820ca27ddc04b04bacc60df09fa33b2e9fd986c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 18:27:45 +0000
Subject: [PATCH 1680/2012] chore(sync): update OpenRouter model catalog
 (#7514)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6d4eb7a48b8..22943b94574 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04116
-output = 0.08232
-cache_read = 0.008232
+input = 0.0406
+output = 0.0812
+cache_read = 0.00812
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index d3b500c9777..f42b1197f93 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8918
-output = 2.8028
-cache_read = 0.16562
+input = 0.8442
+output = 2.6532
+cache_read = 0.15678
 
 [limit]
 context = 1_310_720

From 3baa7cd3ba4a2269ee94399e6eacb32a1eeecd38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 19:20:27 +0000
Subject: [PATCH 1681/2012] chore(sync): update OpenRouter model catalog
 (#7516)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 22943b94574..463a135fc41 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0406
-output = 0.0812
-cache_read = 0.00812
+input = 0.04004
+output = 0.08008
+cache_read = 0.008008
 
 [limit]
 context = 1_048_576

From 8e6b6b80eb3a3cdbead0e94d46c61a23e747cbc2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 20:24:31 +0000
Subject: [PATCH 1682/2012] chore(sync): update OpenRouter model catalog
 (#7517)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 463a135fc41..f8d9e7ed131 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.04004
-output = 0.08008
-cache_read = 0.008008
+input = 0.03976
+output = 0.07952
+cache_read = 0.007952
 
 [limit]
 context = 1_048_576

From d748926e1ab74e8da86e3b9f2fa942118f609112 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 21:22:23 +0000
Subject: [PATCH 1683/2012] chore(sync): update OpenRouter model catalog
 (#7519)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f8d9e7ed131..21112c8f9b1 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03976
-output = 0.07952
-cache_read = 0.007952
+input = 0.03948
+output = 0.07896
+cache_read = 0.007896
 
 [limit]
 context = 1_048_576

From 36bf49018bd02e4987f6cfc890a1fb1f479ebb5d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 22:23:26 +0000
Subject: [PATCH 1684/2012] chore(sync): update OpenRouter model catalog
 (#7520)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 21112c8f9b1..87ebe0279bb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03948
-output = 0.07896
-cache_read = 0.007896
+input = 0.03864
+output = 0.07728
+cache_read = 0.007728
 
 [limit]
 context = 1_048_576

From f85553799b09968cd6e0d4b94a65e771d8eaf397 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sat, 19 Sep 2026 23:21:56 +0000
Subject: [PATCH 1685/2012] chore(sync): update OpenRouter model catalog
 (#7521)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml               | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 87ebe0279bb..c1ff87abe3f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03864
-output = 0.07728
-cache_read = 0.007728
+input = 0.03808
+output = 0.07616
+cache_read = 0.007616
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 12ed56b2bd6..e1246ed1b0a 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.5544
-output = 1.7424
-cache_read = 0.10296
+input = 0.6496
+output = 2.0416
+cache_read = 0.12064
 
 [limit]
 context = 1_048_576

From 132454ed014ae43c9199b232a2ace3bd37031dae Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 01:05:27 +0000
Subject: [PATCH 1686/2012] chore(sync): update OpenRouter model catalog
 (#7522)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml     | 8 ++++++--
 providers/openrouter/models/qwen/qwen3.5-9b.toml          | 4 +++-
 providers/openrouter/models/qwen/qwen3.8-27b.toml         | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml             | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 6 +++---
 7 files changed, 24 insertions(+), 18 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index c1ff87abe3f..d635c80b599 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03808
-output = 0.07616
-cache_read = 0.007616
+input = 0.03724
+output = 0.07448
+cache_read = 0.007448
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
index 34aa7a00097..8257248c271 100644
--- a/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-35b-a3b.toml
@@ -6,8 +6,12 @@ base_model = "alibaba/qwen3.5-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.1625
-output = 1.3
+input = 0.3125
+output = 1.25
+cache_read = 0.15625
+
+[limit]
+output = 16_384
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/qwen/qwen3.5-9b.toml b/providers/openrouter/models/qwen/qwen3.5-9b.toml
index 34660f4a8ed..8f34db0bc11 100644
--- a/providers/openrouter/models/qwen/qwen3.5-9b.toml
+++ b/providers/openrouter/models/qwen/qwen3.5-9b.toml
@@ -1,3 +1,5 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
 base_model = "alibaba/qwen3.5-9b"
 
 [[reasoning_options]]
@@ -8,4 +10,4 @@ input = 0.1
 output = 0.15
 
 [limit]
-output = 235_929
+output = 32_768
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index 488923c2abf..d1ddcffa2ed 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.214
-output = 2.55
-cache_read = 0.15
+input = 0.42
+output = 3
+cache_read = 0.085
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index e1246ed1b0a..12ed56b2bd6 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.6496
-output = 2.0416
-cache_read = 0.12064
+input = 0.5544
+output = 1.7424
+cache_read = 0.10296
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 6ca3ad0c4f6..7e314ece82e 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.91
-output = 2.86
-cache_read = 0.169
+input = 0.896
+output = 2.816
+cache_read = 0.1664
 
 [limit]
 context = 1_310_720

From 198dec37e53fa369af5943183ebcd3ae75cea55a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 01:05:37 +0000
Subject: [PATCH 1687/2012] chore(sync): update Kilo model catalog (#7523)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.5-35b-a3b.toml | 4 ++++
 providers/kilo/models/qwen/qwen3.5-9b.toml      | 3 ++-
 providers/kilo/models/qwen/qwen3.8-27b.toml     | 1 +
 providers/kilo/models/tencent/hy3.toml          | 6 +++---
 4 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
index 96076003f63..bdebd95519c 100644
--- a/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-35b-a3b.toml
@@ -9,5 +9,9 @@ values = ["none", "high"]
 input = 0.1625
 output = 1.3
 
+[limit]
+context = 256_000
+output = 16_384
+
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/qwen/qwen3.5-9b.toml b/providers/kilo/models/qwen/qwen3.5-9b.toml
index bc4a13e3e11..7c1d5856e31 100644
--- a/providers/kilo/models/qwen/qwen3.5-9b.toml
+++ b/providers/kilo/models/qwen/qwen3.5-9b.toml
@@ -10,4 +10,5 @@ input = 0.1
 output = 0.15
 
 [limit]
-output = 235_929
+context = 256_000
+output = 32_768
diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 9b18fb98b33..820f69fb194 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,4 +12,5 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
+context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 1dfbdde36eec9ea3c13ebeebea7fcddbb2e0a20b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 02:34:41 +0000
Subject: [PATCH 1688/2012] chore(sync): update Kilo model catalog (#7524)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 -
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 522fff2ce46..384684585d6 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.018396
+input = 0.57684
+output = 1.73052
+cache_read = 0.019228
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From 57567944cfd29c30605831a52ad72bbdb09cea46 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 02:34:45 +0000
Subject: [PATCH 1689/2012] chore(sync): update OpenRouter model catalog
 (#7525)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c435d69ac6..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.018396
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 42e4245ad89..af6134d3fef 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.018396
+input = 0.57684
+output = 1.73052
+cache_read = 0.019228
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From a680f9b5bf4021da44987fc5080f840a5982323b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 03:30:48 +0000
Subject: [PATCH 1690/2012] chore(sync): update OpenRouter model catalog
 (#7526)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index d635c80b599..0d037ac480a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03724
-output = 0.07448
-cache_read = 0.007448
+input = 0.03696
+output = 0.07392
+cache_read = 0.007392
 
 [limit]
 context = 1_048_576

From fb1319b931674274adef728d69ed497edcf4f052 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 03:31:07 +0000
Subject: [PATCH 1691/2012] chore(sync): update Kilo model catalog (#7527)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/baidu/ernie-4.5-vl-424b-a47b.toml         | 2 +-
 .../kilo/models/deepseek/deepseek-r1-distill-llama-70b.toml     | 2 +-
 providers/kilo/models/deepseek/deepseek-v3.1-terminus.toml      | 2 +-
 providers/kilo/models/deepseek/deepseek-v3.2-exp.toml           | 2 +-
 4 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/baidu/ernie-4.5-vl-424b-a47b.toml b/providers/kilo/models/baidu/ernie-4.5-vl-424b-a47b.toml
index a962a373363..ef65a04722c 100644
--- a/providers/kilo/models/baidu/ernie-4.5-vl-424b-a47b.toml
+++ b/providers/kilo/models/baidu/ernie-4.5-vl-424b-a47b.toml
@@ -1,4 +1,4 @@
-name = "Baidu: ERNIE 4.5 VL 424B A47B "
+name = "Baidu: ERNIE 4.5 VL 424B A47B  (retires Oct 8)"
 description = "Multimodal reasoning model for visual analysis, planning, and tool use"
 family = "ernie"
 release_date = "2025-06-30"
diff --git a/providers/kilo/models/deepseek/deepseek-r1-distill-llama-70b.toml b/providers/kilo/models/deepseek/deepseek-r1-distill-llama-70b.toml
index 3635f13d091..18d5e9a099b 100644
--- a/providers/kilo/models/deepseek/deepseek-r1-distill-llama-70b.toml
+++ b/providers/kilo/models/deepseek/deepseek-r1-distill-llama-70b.toml
@@ -1,4 +1,4 @@
-name = "DeepSeek: R1 Distill Llama 70B"
+name = "DeepSeek: R1 Distill Llama 70B (retires Sep 28)"
 description = "DeepSeek reasoning model for multi-step analysis, math, coding, and tools"
 family = "deepseek"
 release_date = "2025-01-23"
diff --git a/providers/kilo/models/deepseek/deepseek-v3.1-terminus.toml b/providers/kilo/models/deepseek/deepseek-v3.1-terminus.toml
index 79663dcfa0f..0715bd02e00 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.1-terminus.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.1-terminus.toml
@@ -1,4 +1,4 @@
-name = "DeepSeek: DeepSeek V3.1 Terminus"
+name = "DeepSeek: DeepSeek V3.1 Terminus (retires Sep 28)"
 description = "DeepSeek chat model for instruction following, coding, and analysis"
 family = "deepseek"
 release_date = "2025-09-22"
diff --git a/providers/kilo/models/deepseek/deepseek-v3.2-exp.toml b/providers/kilo/models/deepseek/deepseek-v3.2-exp.toml
index 7e9c21db938..0ac45019600 100644
--- a/providers/kilo/models/deepseek/deepseek-v3.2-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v3.2-exp.toml
@@ -1,4 +1,4 @@
-name = "DeepSeek: DeepSeek V3.2 Exp"
+name = "DeepSeek: DeepSeek V3.2 Exp (retires Sep 28)"
 description = "DeepSeek chat model for instruction following, coding, and analysis"
 family = "deepseek"
 release_date = "2025-09-29"

From e582fd85652879f0222385b6e994cd2d46e9037d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 04:29:15 +0000
Subject: [PATCH 1692/2012] chore(sync): update Kilo model catalog (#7528)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731:free.toml     | 15 ---------------
 1 file changed, 15 deletions(-)
 delete mode 100644 providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml
deleted file mode 100644
index 73a49138fbe..00000000000
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-0731:free.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-0731"
-name = "DeepSeek: DeepSeek V4 Flash 0731 (free)"
-description = "DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows...."
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "high", "max"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 1_048_576
-output = 393_216

From 9fc144896fb2b84a5c19dc36d6442b5de1c16c55 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 04:29:25 +0000
Subject: [PATCH 1693/2012] chore(sync): update OpenRouter model catalog
 (#7529)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek/deepseek-v4-flash-0731:free.toml | 20 -------------------
 .../models/deepseek/deepseek-v4-flash.toml    |  6 +++---
 providers/openrouter/models/z-ai/glm-5.2.toml |  6 +++---
 3 files changed, 6 insertions(+), 26 deletions(-)
 delete mode 100644 providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml
deleted file mode 100644
index 1f431a902ab..00000000000
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731:free.toml
+++ /dev/null
@@ -1,20 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-base_model = "deepseek/deepseek-v4-flash-0731"
-name = "DeepSeek V4 Flash 0731 (free)"
-description = "Fast DeepSeek model for efficient chat, coding help, and agent loops"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0d037ac480a..bd86f5a8f95 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03696
-output = 0.07392
-cache_read = 0.007392
+input = 0.03668
+output = 0.07336
+cache_read = 0.007336
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.2.toml b/providers/openrouter/models/z-ai/glm-5.2.toml
index 12ed56b2bd6..e1246ed1b0a 100644
--- a/providers/openrouter/models/z-ai/glm-5.2.toml
+++ b/providers/openrouter/models/z-ai/glm-5.2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.5544
-output = 1.7424
-cache_read = 0.10296
+input = 0.6496
+output = 2.0416
+cache_read = 0.12064
 
 [limit]
 context = 1_048_576

From 4a018cd92f582e01dfcd86fbdc5178d58907f0c4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 05:25:15 +0000
Subject: [PATCH 1694/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7530)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../vercel/models/stepfun/step-5-preview.toml | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/vercel/models/stepfun/step-5-preview.toml

diff --git a/providers/vercel/models/stepfun/step-5-preview.toml b/providers/vercel/models/stepfun/step-5-preview.toml
new file mode 100644
index 00000000000..8db9a1d1622
--- /dev/null
+++ b/providers/vercel/models/stepfun/step-5-preview.toml
@@ -0,0 +1,22 @@
+name = "Step 5 Preview"
+description = "StepFun flash model for efficient multimodal reasoning, coding, and tool use"
+family = "step"
+release_date = "2026-09-20"
+last_updated = "2026-09-20"
+attachment = true
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 1
+output = 2.7
+cache_read = 0.05
+
+[limit]
+context = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 65fe1856216db31941d962ca5895a047d0d2aca4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 06:37:31 +0000
Subject: [PATCH 1695/2012] chore(sync): update Kilo model catalog (#7533)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta/muse-glimmer-30b.toml            | 2 +-
 .../models/mistralai/mistral-small-3.1-24b-instruct.toml    | 2 +-
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/meta/muse-glimmer-30b.toml b/providers/kilo/models/meta/muse-glimmer-30b.toml
index 2c352c0d845..aa01201020a 100644
--- a/providers/kilo/models/meta/muse-glimmer-30b.toml
+++ b/providers/kilo/models/meta/muse-glimmer-30b.toml
@@ -11,7 +11,7 @@ output = 1.1
 cache_read = 0.04
 
 [limit]
-output = 117_964
+output = 16_384
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/mistralai/mistral-small-3.1-24b-instruct.toml b/providers/kilo/models/mistralai/mistral-small-3.1-24b-instruct.toml
index e5a92d90a1c..cddc0eb9495 100644
--- a/providers/kilo/models/mistralai/mistral-small-3.1-24b-instruct.toml
+++ b/providers/kilo/models/mistralai/mistral-small-3.1-24b-instruct.toml
@@ -6,7 +6,7 @@ last_updated = "2025-03-17"
 attachment = true
 reasoning = false
 temperature = true
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = false
 
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 384684585d6..5dd589c4cbf 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57684
-output = 1.73052
-cache_read = 0.019228
+input = 0.5742
+output = 1.7226
+cache_read = 0.01914
 
 [limit]
 context = 1_024_000

From aea958a0052aa3480685b872fea93bded65bbff7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 06:38:05 +0000
Subject: [PATCH 1696/2012] chore(sync): update OpenRouter model catalog
 (#7532)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/meta/muse-glimmer-30b.toml      | 6 +++---
 .../models/mistralai/mistral-small-3.1-24b-instruct.toml    | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/meta/muse-glimmer-30b.toml b/providers/openrouter/models/meta/muse-glimmer-30b.toml
index 3b46ee2987a..d4077f81891 100644
--- a/providers/openrouter/models/meta/muse-glimmer-30b.toml
+++ b/providers/openrouter/models/meta/muse-glimmer-30b.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.35
-output = 1.5
+input = 0.3
+output = 1.2
 cache_read = 0.04
 
 [limit]
-output = 117_964
+output = 16_384
diff --git a/providers/openrouter/models/mistralai/mistral-small-3.1-24b-instruct.toml b/providers/openrouter/models/mistralai/mistral-small-3.1-24b-instruct.toml
index 45893fba68d..bc32aea52eb 100644
--- a/providers/openrouter/models/mistralai/mistral-small-3.1-24b-instruct.toml
+++ b/providers/openrouter/models/mistralai/mistral-small-3.1-24b-instruct.toml
@@ -6,7 +6,7 @@ last_updated = "2025-03-17"
 attachment = true
 reasoning = false
 temperature = true
-tool_call = false
+tool_call = true
 structured_output = false
 knowledge = "2023-10-31"
 open_weights = true
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index af6134d3fef..d54f33b7b90 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57684
-output = 1.73052
-cache_read = 0.019228
+input = 0.5742
+output = 1.7226
+cache_read = 0.01914
 
 [limit]
 context = 1_048_576

From a24c1c5bd9edd972bd942e83fd64cb4444289d60 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 07:26:49 +0000
Subject: [PATCH 1697/2012] chore(sync): update Kilo model catalog (#7536)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 5dd589c4cbf..2b08b36b516 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5742
-output = 1.7226
-cache_read = 0.01914
+input = 0.57288
+output = 1.71864
+cache_read = 0.018228
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]

From 9c86b872b37895567458adabb0298ec512c8b4fa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 07:26:54 +0000
Subject: [PATCH 1698/2012] chore(sync): update OpenRouter model catalog
 (#7537)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d54f33b7b90..6f7ab19b7ee 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5742
-output = 1.7226
-cache_read = 0.01914
+input = 0.57288
+output = 1.71864
+cache_read = 0.018228
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From e82469bebab52d539d292fe1aab14f3a54b3fc00 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 08:31:07 +0000
Subject: [PATCH 1699/2012] chore(sync): update OpenRouter model catalog
 (#7538)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 providers/openrouter/models/~z-ai/glm-flash-latest.toml     | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 6f7ab19b7ee..cbf8d0161f9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57288
-output = 1.71864
-cache_read = 0.018228
+input = 0.57024
+output = 1.71072
+cache_read = 0.018144
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index c32fca67d4b..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From bd94b6ab639374b576d40c6ce78774b7742101ef Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 08:31:17 +0000
Subject: [PATCH 1700/2012] chore(sync): update Kilo model catalog (#7539)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-flash-latest.toml        | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 2b08b36b516..b30e8d73cb1 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57288
-output = 1.71864
-cache_read = 0.018228
+input = 0.57024
+output = 1.71072
+cache_read = 0.018144
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index e4ce4e5dfe6..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]

From 688ef859c6005b6c46c527b3ef7dc2b70c09dd0d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 09:25:40 +0000
Subject: [PATCH 1701/2012] chore(sync): update Kilo model catalog (#7540)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b30e8d73cb1..493f2b1a807 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57024
-output = 1.71072
-cache_read = 0.018144
+input = 0.5676
+output = 1.7028
+cache_read = 0.01806
 
 [limit]
 context = 1_048_576

From ff0fd62ef4dab82f27ea16545974a944d54a2864 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 09:25:44 +0000
Subject: [PATCH 1702/2012] chore(sync): update NanoGPT model catalog (#7541)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/TEE/deepseek-v4-flash.toml      | 13 -------------
 1 file changed, 13 deletions(-)
 delete mode 100644 providers/nano-gpt/models/TEE/deepseek-v4-flash.toml

diff --git a/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml b/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml
deleted file mode 100644
index e9d130676e5..00000000000
--- a/providers/nano-gpt/models/TEE/deepseek-v4-flash.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash"
-name = "DeepSeek V4 Flash TEE"
-reasoning_options = []
-
-[cost]
-input = 0.2
-output = 0.4
-cache_read = 0.04
-
-[limit]
-context = 1_048_576
-input = 1_048_576
-output = 393_216

From 5e6d15218298092d8310dff46544c2e72f3641c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 09:26:11 +0000
Subject: [PATCH 1703/2012] chore(sync): update OpenRouter model catalog
 (#7542)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index bd86f5a8f95..a3b41c6546f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03668
-output = 0.07336
-cache_read = 0.007336
+input = 0.0364
+output = 0.0728
+cache_read = 0.00728
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index cbf8d0161f9..2b68d4267da 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57024
-output = 1.71072
-cache_read = 0.018144
+input = 0.5676
+output = 1.7028
+cache_read = 0.01806
 
 [limit]
 context = 1_048_576

From 18dd8a6b1fd7820a785ba26e498aab90c7422372 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 10:24:23 +0000
Subject: [PATCH 1704/2012] chore(sync): update Kilo model catalog (#7547)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 493f2b1a807..0afbdafc83a 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5676
-output = 1.7028
-cache_read = 0.01806
+input = 0.56628
+output = 1.69884
+cache_read = 0.018876
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From 551c5ef299e818b0f2dc9de5177eac4426a7f8f5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 10:24:27 +0000
Subject: [PATCH 1705/2012] chore(sync): update NanoGPT model catalog (#7545)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml |  6 +++---
 .../models/qwen/qwen3.8-27b-uncensored:thinking.toml |  6 +++---
 providers/nano-gpt/models/typesafe/jev-latest.toml   | 12 ++++++++++++
 3 files changed, 18 insertions(+), 6 deletions(-)
 create mode 100644 providers/nano-gpt/models/typesafe/jev-latest.toml

diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
index 612900717c2..3dc414edb50 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.2
-output = 1.7
-cache_read = 0.18
+input = 0.15
+output = 1.2
+cache_read = 0.125
 
 [limit]
 context = 524_288
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
index 46cc8830338..d4ee170d540 100644
--- a/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-uncensored:thinking.toml
@@ -11,9 +11,9 @@ open_weights = true
 reasoning_options = []
 
 [cost]
-input = 0.2
-output = 1.7
-cache_read = 0.18
+input = 0.15
+output = 1.2
+cache_read = 0.125
 
 [limit]
 context = 524_288
diff --git a/providers/nano-gpt/models/typesafe/jev-latest.toml b/providers/nano-gpt/models/typesafe/jev-latest.toml
new file mode 100644
index 00000000000..49c0acf91f0
--- /dev/null
+++ b/providers/nano-gpt/models/typesafe/jev-latest.toml
@@ -0,0 +1,12 @@
+base_model = "typesafe/jev-latest"
+name = "Jev Latest"
+structured_output = false
+
+[cost]
+input = 0.042
+output = 0
+cache_read = 0.021
+
+[limit]
+context = 32_000
+input = 32_000

From ea28e08347e6d191b2596c285a5494fb6398bad4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 10:24:30 +0000
Subject: [PATCH 1706/2012] chore(sync): update CrossModel model catalog
 (#7548)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/qwen/qwen3.8-omni-flash.toml         | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/crossmodel/models/qwen/qwen3.8-omni-flash.toml

diff --git a/providers/crossmodel/models/qwen/qwen3.8-omni-flash.toml b/providers/crossmodel/models/qwen/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..b70235cd8f0
--- /dev/null
+++ b/providers/crossmodel/models/qwen/qwen3.8-omni-flash.toml
@@ -0,0 +1,17 @@
+base_model = "alibaba/qwen3.8-omni-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.13
+output = 0.43
+cache_read = 0.016
+cache_write = 0.13

From 085f4d737d63f33e68f36a281a38b961c80b0e9f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 10:24:38 +0000
Subject: [PATCH 1707/2012] chore(sync): update OpenRouter model catalog
 (#7546)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index d857fa53dab..91402212177 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.13
-output = 0.52
+input = 0.2
+output = 0.7
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 2b68d4267da..b85f4a83354 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5676
-output = 1.7028
-cache_read = 0.01806
+input = 0.56628
+output = 1.69884
+cache_read = 0.018876
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From ea88df45a8e84fb7e67023a7458b14fad41c7a2e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 11:23:08 +0000
Subject: [PATCH 1708/2012] chore(sync): update Kilo model catalog (#7560)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 0afbdafc83a..6565f5a8e9b 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.56628
-output = 1.69884
-cache_read = 0.018876
+input = 0.56364
+output = 1.69092
+cache_read = 0.018788
 
 [limit]
 context = 1_024_000

From 38ef8bdc93357492d429dcbab47058e66eb6365c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 11:23:12 +0000
Subject: [PATCH 1709/2012] chore(sync): update OpenRouter model catalog
 (#7561)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index b85f4a83354..ff538327fb0 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56628
-output = 1.69884
-cache_read = 0.018876
+input = 0.56364
+output = 1.69092
+cache_read = 0.018788
 
 [limit]
 context = 1_048_576

From 9062eaf44335489e1a31c53ea90d1f7de214010b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 11:23:15 +0000
Subject: [PATCH 1710/2012] chore(sync): update NanoGPT model catalog (#7563)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/qwen/qwen3-14b.toml     | 2 +-
 providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml | 1 -
 providers/nano-gpt/models/qwen/qwen3-8b.toml      | 2 +-
 3 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/qwen/qwen3-14b.toml b/providers/nano-gpt/models/qwen/qwen3-14b.toml
index 21b6ecf5bc8..0ec7e2615e0 100644
--- a/providers/nano-gpt/models/qwen/qwen3-14b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-14b.toml
@@ -6,7 +6,7 @@ release_date = "2024-01-01"
 last_updated = "2024-01-01"
 attachment = false
 reasoning = false
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = true
 
diff --git a/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml b/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
index ef583b4ad7d..3fe31cbc4c1 100644
--- a/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
@@ -1,7 +1,6 @@
 # Included in subscription
 base_model = "alibaba/qwen3-30b-a3b"
 reasoning = false
-tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/nano-gpt/models/qwen/qwen3-8b.toml b/providers/nano-gpt/models/qwen/qwen3-8b.toml
index 5dd994f1ea3..96cac74c86b 100644
--- a/providers/nano-gpt/models/qwen/qwen3-8b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-8b.toml
@@ -5,7 +5,7 @@ release_date = "2024-01-01"
 last_updated = "2024-01-01"
 attachment = false
 reasoning = false
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = true
 

From 5e24bc6d4dcbc10d6c6702070349cb64c5208677 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 12:33:21 +0000
Subject: [PATCH 1711/2012] chore(sync): update OpenRouter model catalog
 (#7565)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ff538327fb0..e56641ce63e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56364
-output = 1.69092
-cache_read = 0.018788
+input = 0.56232
+output = 1.68696
+cache_read = 0.017892
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From f717efa8a69901691a5590d774ad5f43aab14754 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 12:33:59 +0000
Subject: [PATCH 1712/2012] chore(sync): update Kilo model catalog (#7566)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6565f5a8e9b..c178a343e37 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.56364
-output = 1.69092
-cache_read = 0.018788
+input = 0.56232
+output = 1.68696
+cache_read = 0.017892
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]

From b8fb6496f17b8058260ac12d67ca16af8e835b41 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 13:23:36 +0000
Subject: [PATCH 1713/2012] chore(sync): update NanoGPT model catalog (#7569)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/stepfun/step-5-preview.toml        | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/nano-gpt/models/stepfun/step-5-preview.toml

diff --git a/providers/nano-gpt/models/stepfun/step-5-preview.toml b/providers/nano-gpt/models/stepfun/step-5-preview.toml
new file mode 100644
index 00000000000..e6b122e57eb
--- /dev/null
+++ b/providers/nano-gpt/models/stepfun/step-5-preview.toml
@@ -0,0 +1,28 @@
+name = "Step 5 Preview"
+description = "Step 5 Preview is StepFun's 600B sparse MoE frontier model for production-scale agents, activating 27B parameters per token. It is built for software engineering, long-horizon tool use, research, professional knowledge work, and finance, with native text, image, and video understanding and a 1M-token context window. ⚠️ Note: This model routes through StepFun, so privacy and logging guarantees may be limited."
+family = "step"
+release_date = "2026-09-20"
+last_updated = "2026-09-20"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1
+output = 2.7
+cache_read = 0.05
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

From 3fdcb5a3f3d3180e76f7a105da5d2602a710f522 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 13:23:41 +0000
Subject: [PATCH 1714/2012] chore(sync): update Kilo model catalog (#7568)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-27b.toml            |  1 -
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 820f69fb194..9b18fb98b33 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,5 +12,4 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
-context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index c178a343e37..0a7f6cc12c4 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.56232
-output = 1.68696
-cache_read = 0.017892
+input = 0.55836
+output = 1.67508
+cache_read = 0.018612
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From 5f6b993ea8252aabb2a6fd298bd198ff5240081a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 13:23:48 +0000
Subject: [PATCH 1715/2012] chore(sync): update OpenRouter model catalog
 (#7570)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-27b.toml         | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a3b41c6546f..0ab3f377da2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.0364
-output = 0.0728
-cache_read = 0.00728
+input = 0.03612
+output = 0.07224
+cache_read = 0.007224
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index d1ddcffa2ed..87b348fc2c9 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,8 +11,8 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.42
-output = 3
+input = 0.2
+output = 2.55
 cache_read = 0.085
 
 [limit]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index e56641ce63e..b7cb08f5927 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56232
-output = 1.68696
-cache_read = 0.017892
+input = 0.55836
+output = 1.67508
+cache_read = 0.018612
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From a85f89fbf19e18a9fa6121f77a90dcea5510104a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 14:23:34 +0000
Subject: [PATCH 1716/2012] chore(sync): update Kilo model catalog (#7572)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
+context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 0a7f6cc12c4..42c36b3bc07 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.55836
-output = 1.67508
-cache_read = 0.018612
+input = 0.55044
+output = 1.65132
+cache_read = 0.018348
 
 [limit]
 context = 1_024_000

From 70277a8a17439d36460a6d12741021e771910aa5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 14:23:42 +0000
Subject: [PATCH 1717/2012] chore(sync): update OpenRouter model catalog
 (#7573)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml               | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 0ab3f377da2..af806ba67d3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03612
-output = 0.07224
-cache_read = 0.007224
+input = 0.03584
+output = 0.07168
+cache_read = 0.007168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..679f96c65ab 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.55044
+output = 1.65132
+cache_read = 0.018348
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 7e314ece82e..452bead80a9 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.896
-output = 2.816
-cache_read = 0.1664
+input = 0.882
+output = 2.772
+cache_read = 0.1638
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index b7cb08f5927..21b1ebe291f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.55836
-output = 1.67508
-cache_read = 0.018612
+input = 0.55044
+output = 1.65132
+cache_read = 0.018348
 
 [limit]
 context = 1_048_576

From ad95e11fcdf4df38c5f7336b925e4d99baacb541 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 14:23:43 +0000
Subject: [PATCH 1718/2012] chore(sync): update NanoGPT model catalog (#7571)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml | 6 +++---
 .../models/deepseek/deepseek-v4.1-flash:thinking.toml       | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
index 8e61a5cbf70..c7745ff077e 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.4
-cache_read = 0.003
+input = 0.13
+output = 0.52
+cache_read = 0.006
 
 [limit]
 input = 1_000_000
diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
index 90c16dc648c..c26ac296879 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4.1-flash:thinking.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.4
-cache_read = 0.003
+input = 0.13
+output = 0.52
+cache_read = 0.006
 
 [limit]
 input = 1_000_000

From e35cc0fb47862c2c383b47fdc9feb1e22459dca3 Mon Sep 17 00:00:00 2001
From: Pedro Santos <18473317+pc-gs@users.noreply.github.com>
Date: Sun, 20 Sep 2026 15:04:58 +0000
Subject: [PATCH 1719/2012] fix(opencode-go): correct Go docs URL and MiniMax
 cache pricing

Fixes #2097
---
 providers/opencode-go/models/minimax-m2.5.toml | 3 ++-
 providers/opencode-go/models/minimax-m2.7.toml | 1 +
 providers/opencode-go/provider.toml            | 2 +-
 3 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/providers/opencode-go/models/minimax-m2.5.toml b/providers/opencode-go/models/minimax-m2.5.toml
index 81fcd79363a..f6973ebf170 100644
--- a/providers/opencode-go/models/minimax-m2.5.toml
+++ b/providers/opencode-go/models/minimax-m2.5.toml
@@ -15,7 +15,8 @@ status = "deprecated"
 [cost]
 input = 0.3
 output = 1.2
-cache_read = 0.03
+cache_read = 0.06
+cache_write = 0.375
 
 [limit]
 context = 204_800
diff --git a/providers/opencode-go/models/minimax-m2.7.toml b/providers/opencode-go/models/minimax-m2.7.toml
index 311a9c4af6f..33faddac7d9 100644
--- a/providers/opencode-go/models/minimax-m2.7.toml
+++ b/providers/opencode-go/models/minimax-m2.7.toml
@@ -15,6 +15,7 @@ open_weights = true
 input = 0.3
 output = 1.2
 cache_read = 0.06
+cache_write = 0.375
 
 [limit]
 context = 204_800
diff --git a/providers/opencode-go/provider.toml b/providers/opencode-go/provider.toml
index 833a7cb0002..7ddf38e1f5d 100644
--- a/providers/opencode-go/provider.toml
+++ b/providers/opencode-go/provider.toml
@@ -8,4 +8,4 @@ npm = "@ai-sdk/openai-compatible"
 # by the public Zen page.
 # https://opencode.ai/docs/zen#endpoints
 api = "https://opencode.ai/zen/go/v1"
-doc = "https://opencode.ai/docs/zen"
+doc = "https://opencode.ai/docs/go"

From dc365bb0098e0fb32f8c12012c8a0a7e99ecd9b7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 15:23:28 +0000
Subject: [PATCH 1720/2012] chore(sync): update LLM Gateway model catalog
 (#7578)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/baidu/deepseek-v4-flash.toml           |  2 +-
 .../models/baidu/deepseek-v4-pro.toml             |  2 +-
 .../models/baidu/deepseek-v4.1-flash.toml         | 15 +++++++++++++++
 3 files changed, 17 insertions(+), 2 deletions(-)
 create mode 100644 providers/llmgateway-providers/models/baidu/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/baidu/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/baidu/deepseek-v4-flash.toml
index 95bedd8498e..40383b0961a 100644
--- a/providers/llmgateway-providers/models/baidu/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/baidu/deepseek-v4-flash.toml
@@ -12,7 +12,7 @@ values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 [cost]
 input = 0.44
 output = 1.32
-cache_read = 0.044
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway-providers/models/baidu/deepseek-v4-pro.toml b/providers/llmgateway-providers/models/baidu/deepseek-v4-pro.toml
index bab1be34eeb..e9463e0fe2a 100644
--- a/providers/llmgateway-providers/models/baidu/deepseek-v4-pro.toml
+++ b/providers/llmgateway-providers/models/baidu/deepseek-v4-pro.toml
@@ -12,7 +12,7 @@ values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 [cost]
 input = 1.32
 output = 3.96
-cache_read = 0.132
+cache_read = 0.042
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway-providers/models/baidu/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/baidu/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..a22c803e3f5
--- /dev/null
+++ b/providers/llmgateway-providers/models/baidu/deepseek-v4.1-flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (Baidu)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.006
+
+[limit]
+context = 1_048_576
+output = 393_216

From 1801d72511dc39c04b33b77ef1315643938b433d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 15:23:38 +0000
Subject: [PATCH 1721/2012] chore(sync): update Kilo model catalog (#7577)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 42c36b3bc07..e4d3acd84bf 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.55044
-output = 1.65132
-cache_read = 0.018348
+input = 0.5478
+output = 1.6434
+cache_read = 0.01826
 
 [limit]
 context = 1_024_000

From e20b0c573ab134b6c55ec001f06463342f40d4e8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 15:23:41 +0000
Subject: [PATCH 1722/2012] chore(sync): update OpenRouter model catalog
 (#7576)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml               | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index af806ba67d3..98ee0445042 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03584
-output = 0.07168
-cache_read = 0.007168
+input = 0.03556
+output = 0.07112
+cache_read = 0.007112
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 679f96c65ab..7191760d710 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.55044
-output = 1.65132
-cache_read = 0.018348
+input = 0.5478
+output = 1.6434
+cache_read = 0.01826
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 452bead80a9..6ca3ad0c4f6 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.882
-output = 2.772
-cache_read = 0.1638
+input = 0.91
+output = 2.86
+cache_read = 0.169
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 21b1ebe291f..78806039bab 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.55044
-output = 1.65132
-cache_read = 0.018348
+input = 0.5478
+output = 1.6434
+cache_read = 0.01826
 
 [limit]
 context = 1_048_576

From 6947a51e1b8198ae8bedb4b13ca839433fe14b63 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Sun, 20 Sep 2026 10:59:09 -0500
Subject: [PATCH 1723/2012] feat: add model type filtering (#7579)

* feat: add model type filtering

* fix: preserve model types across catalog paths

---------

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 README.md                             |  12 +++
 bun.lock                              |   3 +
 models/typesafe/jev-latest.toml       |   1 +
 packages/core/src/filter.ts           |  83 ++++++++++++++++
 packages/core/src/index.ts            |   1 +
 packages/core/src/schema.ts           |   5 +
 packages/core/src/sync/index.ts       |   1 +
 packages/core/test/filter.test.ts     | 103 +++++++++++++++++++
 packages/core/test/schema.test.ts     |   6 ++
 packages/core/test/sync.test.ts       |  22 ++++
 packages/function/package.json        |   3 +
 packages/function/src/worker.ts       |  98 ++++++++++++++++--
 packages/function/test/worker.test.ts | 138 ++++++++++++++++++++++++++
 packages/sdk/src/client.ts            |  14 ++-
 packages/sdk/src/effect.ts            |   2 +-
 packages/sdk/src/effect/client.ts     |  28 ++++--
 packages/sdk/src/types.ts             |   5 +
 packages/sdk/test/client.test.ts      |  11 ++
 packages/sdk/test/effect.test.ts      |  14 +++
 packages/web/script/build.ts          |  28 ++++--
 packages/web/src/render.tsx           |   9 +-
 packages/web/src/server.ts            |  56 +++++++----
 22 files changed, 590 insertions(+), 53 deletions(-)
 create mode 100644 packages/core/src/filter.ts
 create mode 100644 packages/core/test/filter.test.ts
 create mode 100644 packages/function/test/worker.test.ts

diff --git a/README.md b/README.md
index f26ba19ab83..8a6c27e736a 100644
--- a/README.md
+++ b/README.md
@@ -22,6 +22,17 @@ You can access this data through an API.
 curl https://models.dev/api.json
 ```
 
+Specialized model types are omitted by default. Filter by one or more
+comma-separated model types, or use `all` for the complete catalog:
+
+```bash
+curl "https://models.dev/api.json?type=decision"
+curl "https://models.dev/api.json?type=all"
+```
+
+The currently supported model type is `decision`. The `type` parameter is also
+available on `models.json`, `catalog.json`, and `model-schema.json`.
+
 Use the **Model ID** field to do a lookup on any model; it's the identifier used by [AI SDK](https://ai-sdk.dev/).
 
 Provider-agnostic model metadata is available separately:
@@ -270,6 +281,7 @@ Models must conform to the following schema, as defined in `packages/core/src/sc
 **Model Schema:**
 
 - `name`: String — Display name of the model
+- `type` _(optional)_: String — Specialized model behavior; currently supports `decision`
 - `attachment`: Boolean — Supports file attachments
 - `reasoning`: Boolean — Supports reasoning / chain-of-thought
 - `tool_call`: Boolean - Supports tool calling
diff --git a/bun.lock b/bun.lock
index 2ba689570d3..2e4ee0e36cd 100644
--- a/bun.lock
+++ b/bun.lock
@@ -24,6 +24,9 @@
     },
     "packages/function": {
       "name": "@models.dev/function",
+      "dependencies": {
+        "@models.dev/core": "workspace:*",
+      },
       "devDependencies": {
         "@cloudflare/workers-types": "4.20250522.0",
         "@tsconfig/bun": "catalog:",
diff --git a/models/typesafe/jev-latest.toml b/models/typesafe/jev-latest.toml
index 9b2f3a75896..6b54d1bd25f 100644
--- a/models/typesafe/jev-latest.toml
+++ b/models/typesafe/jev-latest.toml
@@ -4,6 +4,7 @@
 # - https://typesafe.ai/blog/introducing-system-one-models-and-jev
 # The 64k context limit covers the state plus all questions in one request.
 name = "Jev"
+type = "decision"
 description = "System One model for fast, typed probabilistic decisions over text or structured state"
 release_date = "2026-09-15"
 last_updated = "2026-09-15"
diff --git a/packages/core/src/filter.ts b/packages/core/src/filter.ts
new file mode 100644
index 00000000000..3cba07d6cae
--- /dev/null
+++ b/packages/core/src/filter.ts
@@ -0,0 +1,83 @@
+export const MODEL_TYPES = ["decision"] as const;
+
+export type ModelTypeValue = (typeof MODEL_TYPES)[number];
+export type ModelTypeFilter = "default" | "all" | ModelTypeValue[];
+
+interface TypedModel {
+  type?: ModelTypeValue;
+}
+
+interface TypedProvider {
+  models: Record<string, TypedModel>;
+}
+
+export class InvalidModelTypeError extends Error {
+  constructor(value: string) {
+    super(`Invalid type value: ${value}`);
+    this.name = "InvalidModelTypeError";
+  }
+}
+
+export function parseModelTypes(value: string | null): ModelTypeFilter {
+  if (value === null || value === "") return "default";
+  if (value === "all") return "all";
+
+  const types = value.split(",");
+  if (
+    types.length === 0 ||
+    types.includes("all") ||
+    types.some((type) => !MODEL_TYPES.includes(type as ModelTypeValue))
+  ) {
+    throw new InvalidModelTypeError(value);
+  }
+
+  return [...new Set(types)] as ModelTypeValue[];
+}
+
+function includesModel(model: TypedModel, filter: ModelTypeFilter) {
+  if (filter === "all") return true;
+  if (filter === "default") return model.type === undefined;
+  return model.type !== undefined && filter.includes(model.type);
+}
+
+export function filterProvidersByModelType<
+  T extends Record<string, TypedProvider>,
+>(providers: T, filter: ModelTypeFilter): T {
+  if (filter === "all") return providers;
+
+  return Object.fromEntries(
+    Object.entries(providers).flatMap(([providerID, provider]) => {
+      const models = Object.fromEntries(
+        Object.entries(provider.models).filter(([, model]) =>
+          includesModel(model, filter),
+        ),
+      );
+      return Object.keys(models).length === 0
+        ? []
+        : [[providerID, { ...provider, models }]];
+    }),
+  ) as T;
+}
+
+export function filterModelsByModelType<T extends Record<string, TypedModel>>(
+  models: T,
+  filter: ModelTypeFilter,
+): T {
+  if (filter === "all") return models;
+  return Object.fromEntries(
+    Object.entries(models).filter(([, model]) => includesModel(model, filter)),
+  ) as T;
+}
+
+export function filterCatalogByModelType<
+  TProviders extends Record<string, TypedProvider>,
+  TModels extends Record<string, TypedModel>,
+>(
+  catalog: { providers: TProviders; models: TModels },
+  filter: ModelTypeFilter,
+): { providers: TProviders; models: TModels } {
+  return {
+    providers: filterProvidersByModelType(catalog.providers, filter),
+    models: filterModelsByModelType(catalog.models, filter),
+  };
+}
diff --git a/packages/core/src/index.ts b/packages/core/src/index.ts
index 69f4e893dcb..b8be1ed0f80 100644
--- a/packages/core/src/index.ts
+++ b/packages/core/src/index.ts
@@ -2,3 +2,4 @@ export * from "./schema.js";
 export * from "./generate.js";
 export * from "./describe.js";
 export * from "./family.js";
+export * from "./filter.js";
diff --git a/packages/core/src/schema.ts b/packages/core/src/schema.ts
index c6c8b8f38b2..5c2d7b06681 100644
--- a/packages/core/src/schema.ts
+++ b/packages/core/src/schema.ts
@@ -1,6 +1,7 @@
 import { z } from "zod";
 
 import { ModelFamily } from "./family";
+import { MODEL_TYPES } from "./filter";
 
 type JsonValue =
   | string
@@ -148,6 +149,8 @@ const DateString = z
 
 const Modality = z.enum(["text", "audio", "image", "video", "pdf"]);
 
+export const ModelType = z.enum(MODEL_TYPES);
+
 const Modalities = z
   .object({
     input: z.array(Modality),
@@ -219,6 +222,7 @@ export const BenchmarkResult = z
 
 const ModelMetadataBase = z.object({
   id: z.string(),
+  type: ModelType.optional(),
   name: z.string().min(1, "Model name cannot be empty"),
   description: z.string().min(1, "Model description cannot be empty"),
   family: ModelFamily.optional(),
@@ -245,6 +249,7 @@ export type ModelMetadata = z.infer<typeof ModelMetadata>;
 
 const ModelBase = z.object({
   id: z.string(),
+  type: ModelType.optional(),
   name: z.string().min(1, "Model name cannot be empty"),
   description: z.string().min(1, "Model description cannot be empty"),
   family: ModelFamily.optional(),
diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index 5095f4124ef..031341b56cd 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -1011,6 +1011,7 @@ export function formatToml(model: z.infer<typeof SyncedAuthoredModel>) {
   if ("base_model_omit" in model && model.base_model_omit !== undefined) {
     lines.push(`base_model_omit = [${model.base_model_omit.map(quote).join(", ")}]`);
   }
+  if (model.type !== undefined) lines.push(`type = ${quote(model.type)}`);
   if (model.name !== undefined) lines.push(`name = ${quote(model.name)}`);
   if (model.description !== undefined) lines.push(`description = ${quote(model.description)}`);
   if (model.family !== undefined) lines.push(`family = ${quote(model.family)}`);
diff --git a/packages/core/test/filter.test.ts b/packages/core/test/filter.test.ts
new file mode 100644
index 00000000000..ea14f1588fb
--- /dev/null
+++ b/packages/core/test/filter.test.ts
@@ -0,0 +1,103 @@
+import { describe, expect, test } from "bun:test";
+
+import {
+  filterCatalogByModelType,
+  generateCatalog,
+  InvalidModelTypeError,
+  parseModelTypes,
+} from "../src/index.js";
+import type { ModelMetadata, Provider } from "../src/index.js";
+import path from "node:path";
+
+describe("model type filtering", () => {
+  test("defaults to untyped models and supports specific types and all", () => {
+    expect(parseModelTypes(null)).toBe("default");
+    expect(parseModelTypes("")).toBe("default");
+    expect(parseModelTypes("decision")).toEqual(["decision"]);
+    expect(parseModelTypes("all")).toBe("all");
+  });
+
+  test("rejects unknown types and combining all with a type", () => {
+    expect(() => parseModelTypes("unknown")).toThrow(InvalidModelTypeError);
+    expect(() => parseModelTypes("all,decision")).toThrow(
+      InvalidModelTypeError,
+    );
+  });
+
+  test("omits typed models by default", () => {
+    const catalog = fixture();
+    const filtered = filterCatalogByModelType(catalog, "default");
+
+    expect(Object.keys(filtered.models)).toEqual(["standard"]);
+    expect(Object.keys(filtered.providers.example!.models)).toEqual([
+      "standard",
+    ]);
+    expect(filtered.providers.decisionOnly).toBeUndefined();
+  });
+
+  test("filters canonical and provider models by requested type", () => {
+    const catalog = fixture();
+    const filtered = filterCatalogByModelType(catalog, ["decision"]);
+
+    expect(Object.keys(filtered.models)).toEqual(["decision"]);
+    expect(Object.keys(filtered.providers.example!.models)).toEqual([
+      "decision",
+    ]);
+    expect(Object.keys(filtered.providers.decisionOnly!.models)).toEqual([
+      "decision",
+    ]);
+    expect(filterCatalogByModelType(catalog, "all")).toEqual(catalog);
+  });
+
+  test("every repository Jev model inherits decision and is omitted by default", async () => {
+    const root = path.join(import.meta.dir, "..", "..", "..");
+    const catalog = await generateCatalog(root);
+    const jevModels = Object.values(catalog.providers).flatMap((provider) =>
+      Object.values(provider.models).filter((model) =>
+        model.id.toLowerCase().includes("jev"),
+      ),
+    );
+
+    expect(jevModels.length).toBeGreaterThan(0);
+    expect(jevModels.every((model) => model.type === "decision")).toBe(true);
+
+    const defaults = filterCatalogByModelType(catalog, "default");
+    expect(
+      Object.values(defaults.providers).some((provider) =>
+        Object.values(provider.models).some((model) => model.type !== undefined),
+      ),
+    ).toBe(false);
+    expect(defaults.models["typesafe/jev-latest"]).toBeUndefined();
+    expect(defaults.providers.vivgrid?.models.jev).toBeUndefined();
+
+    const decisions = filterCatalogByModelType(catalog, ["decision"]);
+    expect(decisions.models["typesafe/jev-latest"]?.type).toBe("decision");
+    expect(
+      Object.values(decisions.providers).flatMap((provider) =>
+        Object.values(provider.models),
+      ).length,
+    ).toBe(jevModels.length);
+  });
+});
+
+function fixture() {
+  const standard = model("standard-model");
+  const decision = model("decision-model", "decision");
+  return {
+    models: { standard, decision },
+    providers: {
+      example: {
+        id: "example",
+        models: { standard, decision },
+      } as unknown as Provider,
+      decisionOnly: {
+        id: "decision-only",
+        models: { decision },
+      } as unknown as Provider,
+    },
+  };
+}
+
+function model(id: string, type?: "decision") {
+  return { id, type } as unknown as ModelMetadata;
+}
diff --git a/packages/core/test/schema.test.ts b/packages/core/test/schema.test.ts
index f343c50b615..97a6d602032 100644
--- a/packages/core/test/schema.test.ts
+++ b/packages/core/test/schema.test.ts
@@ -70,6 +70,12 @@ describe("model schema", () => {
     expect(AuthoredModel.safeParse(model).success).toBe(false);
   });
 
+  test("accepts decision model types", () => {
+    const model = baseModel({ type: "decision" });
+
+    expect(AuthoredModel.safeParse(model).success).toBe(true);
+  });
+
   test("accepts calendar-valid model dates", () => {
     for (const field of dateFields) {
       for (const value of [
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 3557ba66127..d0609631969 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -2806,6 +2806,28 @@ test("resolves Eden AI aliases to the model they point at", () => {
   ).toBe("anthropic/claude-opus-5");
 });
 
+test("preserves model type when formatting synced TOML", () => {
+  const content = formatToml({
+    id: "typesafe/jev-latest",
+    type: "decision",
+    name: "Jev",
+    description: "System One model for typed decisions",
+    release_date: "2026-09-15",
+    last_updated: "2026-09-15",
+    attachment: false,
+    reasoning: false,
+    tool_call: false,
+    open_weights: false,
+    limit: { context: 64_000, output: 0 },
+    modalities: { input: ["text"], output: ["text"] },
+  });
+
+  expect(Bun.TOML.parse(content)).toMatchObject({
+    type: "decision",
+    name: "Jev",
+  });
+});
+
 test("formats interleaved as a root field before reasoning option tables", () => {
   const content = formatToml({
     id: "example/model",
diff --git a/packages/function/package.json b/packages/function/package.json
index 1d1c7c9f5c1..e1fd3d8260a 100644
--- a/packages/function/package.json
+++ b/packages/function/package.json
@@ -3,6 +3,9 @@
   "name": "@models.dev/function",
   "private": true,
   "type": "module",
+  "dependencies": {
+    "@models.dev/core": "workspace:*"
+  },
   "devDependencies": {
     "@cloudflare/workers-types": "4.20250522.0",
     "@tsconfig/bun": "catalog:"
diff --git a/packages/function/src/worker.ts b/packages/function/src/worker.ts
index 9ccccdd42e4..d616805d8f9 100644
--- a/packages/function/src/worker.ts
+++ b/packages/function/src/worker.ts
@@ -1,3 +1,13 @@
+import {
+  filterCatalogByModelType,
+  filterModelsByModelType,
+  filterProvidersByModelType,
+  InvalidModelTypeError,
+  MODEL_TYPES,
+  parseModelTypes,
+} from "@models.dev/core/src/filter.js";
+import type { ModelTypeValue } from "@models.dev/core/src/filter.js";
+
 export interface Env {
   ASSETS: any;
   PosthogToken: string;
@@ -64,11 +74,8 @@ export default {
     }
 
     if (url.pathname === "/model-schema.json") {
-      const apiUrl = new URL(url);
-      apiUrl.pathname = "/_api.json";
-      const apiResponse = await env.ASSETS.fetch(
-        new Request(apiUrl.toString(), request),
-      );
+      const apiResponse = await catalogResponse(url, request, env, "api");
+      if (!apiResponse.ok) return apiResponse;
       const providers = (await apiResponse.json()) as Record<
         string,
         { models: Record<string, unknown> }
@@ -102,11 +109,11 @@ export default {
     }
 
     if (url.pathname === "/api.json") {
-      url.pathname = "/_api.json";
+      return catalogResponse(url, request, env, "api");
     } else if (url.pathname === "/models.json") {
-      url.pathname = "/_models.json";
+      return catalogResponse(url, request, env, "models");
     } else if (url.pathname === "/catalog.json") {
-      url.pathname = "/_catalog.json";
+      return catalogResponse(url, request, env, "catalog");
     } else if (
       url.pathname === "/" ||
       url.pathname === "/index.html" ||
@@ -143,6 +150,81 @@ export default {
   },
 };
 
+type CatalogEndpoint = "api" | "models" | "catalog";
+
+async function catalogResponse(
+  url: URL,
+  request: Request,
+  env: Env,
+  endpoint: CatalogEndpoint,
+) {
+  let filter;
+  try {
+    filter = parseModelTypes(url.searchParams.get("type"));
+  } catch (error) {
+    if (!(error instanceof InvalidModelTypeError)) throw error;
+    return Response.json(
+      {
+        error: error.message,
+        allowed: [...MODEL_TYPES, "all"],
+      },
+      {
+        status: 400,
+        headers: { "Access-Control-Allow-Origin": "*" },
+      },
+    );
+  }
+
+  const assetUrl = new URL(url);
+  const suffix = filter === "default"
+    ? ""
+    : filter === "all"
+      ? "-all"
+      : filter.length === 1
+        ? `-${filter[0]}`
+        : undefined;
+  assetUrl.pathname = `/_${endpoint}${suffix ?? "-all"}.json`;
+  assetUrl.search = "";
+  const assetResponse = await env.ASSETS.fetch(
+    new Request(assetUrl.toString(), request),
+  );
+  if (!assetResponse.ok || suffix !== undefined) return assetResponse;
+
+  const value = await assetResponse.json();
+  const filtered = endpoint === "api"
+    ? filterProvidersByModelType(
+        value as Record<string, CatalogProvider>,
+        filter,
+      )
+    : endpoint === "models"
+      ? filterModelsByModelType(
+          value as Record<string, CatalogModel>,
+          filter,
+        )
+      : filterCatalogByModelType(
+          value as {
+            providers: Record<string, CatalogProvider>;
+            models: Record<string, CatalogModel>;
+          },
+          filter,
+        );
+
+  const headers = new Headers(assetResponse.headers);
+  headers.delete("Content-Length");
+  headers.delete("ETag");
+  headers.set("Content-Type", "application/json");
+  headers.set("Cache-Control", "public, max-age=3600");
+  return new Response(JSON.stringify(filtered), { headers });
+}
+
+interface CatalogModel {
+  type?: ModelTypeValue;
+}
+
+interface CatalogProvider {
+  models: Record<string, CatalogModel>;
+}
+
 function isHtmlRoute(pathname: string) {
   return (
     pathname === "/models" ||
diff --git a/packages/function/test/worker.test.ts b/packages/function/test/worker.test.ts
new file mode 100644
index 00000000000..26d9e0fb04a
--- /dev/null
+++ b/packages/function/test/worker.test.ts
@@ -0,0 +1,138 @@
+import { describe, expect, test } from "bun:test";
+
+import worker, { type Env } from "../src/worker.js";
+
+const textModel = {
+  id: "text-model",
+  modalities: { input: ["text"], output: ["text"] },
+};
+const decisionModel = {
+  id: "decision-model",
+  type: "decision",
+  modalities: { input: ["text"], output: ["text"] },
+};
+const providers = {
+  example: {
+    id: "example",
+    models: { text: textModel, decision: decisionModel },
+  },
+};
+const models = { text: textModel, decision: decisionModel };
+
+describe("catalog API model type filtering", () => {
+  test("omits typed models from api.json by default", async () => {
+    const response = await request("/api.json");
+    const body = await response.json();
+
+    expect(Object.keys(body.example.models)).toEqual(["text"]);
+  });
+
+  test("omits typed models from models.json by default", async () => {
+    const response = await request("/models.json");
+    const body = await response.json();
+
+    expect(Object.keys(body)).toEqual(["text"]);
+  });
+
+  test("omits typed models from catalog.json by default", async () => {
+    const response = await request("/catalog.json");
+    const body = await response.json();
+
+    expect(Object.keys(body.models)).toEqual(["text"]);
+    expect(Object.keys(body.providers.example.models)).toEqual(["text"]);
+  });
+
+  test("returns explicitly requested decision models", async () => {
+    const response = await request("/catalog.json?type=decision");
+    const body = await response.json();
+
+    expect(Object.keys(body.models)).toEqual(["decision"]);
+    expect(Object.keys(body.providers.example.models)).toEqual(["decision"]);
+  });
+
+  test("returns the complete static catalog for all", async () => {
+    const response = await request("/models.json?type=all");
+    const body = await response.json();
+
+    expect(Object.keys(body)).toEqual(["text", "decision"]);
+  });
+
+  test("omits typed models from model-schema.json by default", async () => {
+    const response = await request("/model-schema.json");
+    const body = await response.json();
+
+    expect(body.$defs.Model.enum).toEqual(["example/text"]);
+  });
+
+  test("includes typed models in model-schema.json when requested", async () => {
+    const response = await request("/model-schema.json?type=all");
+    const body = await response.json();
+
+    expect(body.$defs.Model.enum).toEqual([
+      "example/decision",
+      "example/text",
+    ]);
+  });
+
+  test("rejects unknown model types", async () => {
+    const response = await request("/api.json?type=unknown");
+
+    expect(response.status).toBe(400);
+  });
+});
+
+async function request(path: string) {
+  const env = {
+    ASSETS: {
+      fetch(input: Request) {
+        const pathname = new URL(input.url).pathname;
+        if (pathname === "/_api.json") {
+          return Response.json({
+            example: { ...providers.example, models: { text: textModel } },
+          });
+        }
+        if (pathname === "/_api-all.json") return Response.json(providers);
+        if (pathname === "/_api-decision.json") {
+          return Response.json({
+            example: { ...providers.example, models: { decision: decisionModel } },
+          });
+        }
+        if (pathname === "/_models.json") {
+          return Response.json({ text: textModel });
+        }
+        if (pathname === "/_models-all.json") return Response.json(models);
+        if (pathname === "/_models-decision.json") {
+          return Response.json({ decision: decisionModel });
+        }
+        if (pathname === "/_catalog.json") {
+          return Response.json({
+            providers: {
+              example: { ...providers.example, models: { text: textModel } },
+            },
+            models: { text: textModel },
+          });
+        }
+        if (pathname === "/_catalog-all.json") {
+          return Response.json({ providers, models });
+        }
+        if (pathname === "/_catalog-decision.json") {
+          return Response.json({
+            providers: {
+              example: { ...providers.example, models: { decision: decisionModel } },
+            },
+            models: { decision: decisionModel },
+          });
+        }
+        return new Response(null, { status: 404 });
+      },
+    },
+  } as unknown as Env;
+
+  return worker.fetch(
+    new Request(`https://models.dev${path}`, {
+      headers: { "user-agent": "test" },
+    }),
+    env,
+    { waitUntil() {} } as unknown as ExecutionContext,
+  );
+}
diff --git a/packages/sdk/src/client.ts b/packages/sdk/src/client.ts
index f63aa5c3660..ee3848e4e3f 100644
--- a/packages/sdk/src/client.ts
+++ b/packages/sdk/src/client.ts
@@ -1,5 +1,5 @@
 import { ModelsDevError } from "./error.js"
-import type { Catalog, ModelMetadataMap, ProviderMap } from "./types.js"
+import type { Catalog, ModelMetadataMap, ModelType, ProviderMap } from "./types.js"
 
 /** Accepted anywhere headers can be passed. Same shapes as the standard `HeadersInit`. */
 export type HeadersInput = Headers | Record<string, string> | Array<[string, string]>
@@ -21,6 +21,8 @@ export interface RequestOptions {
   readonly signal?: AbortSignal
   /** Extra headers for this request. Overrides client-level headers. */
   readonly headers?: HeadersInput
+  /** Specialized model types to include. Omit for standard models; use `"all"` for the complete catalog. */
+  readonly modelTypes?: "all" | readonly ModelType[]
 }
 
 /**
@@ -41,7 +43,15 @@ export function make(options: ClientOptions = {}) {
 
     let response: Response
     try {
-      response = await fetch(new URL(path, base), {
+      const url = new URL(path, base)
+      const modelTypes = requestOptions?.modelTypes
+      if (modelTypes === "all") {
+        url.searchParams.set("type", "all")
+      } else if (modelTypes && modelTypes.length > 0) {
+        url.searchParams.set("type", modelTypes.join(","))
+      }
+
+      response = await fetch(url, {
         method: "GET",
         headers,
         signal: requestOptions?.signal,
diff --git a/packages/sdk/src/effect.ts b/packages/sdk/src/effect.ts
index cd1780d182e..24827173e48 100644
--- a/packages/sdk/src/effect.ts
+++ b/packages/sdk/src/effect.ts
@@ -1,4 +1,4 @@
 // Effect-native client. Requires the optional peer dependency `effect`.
 export * as Models from "./effect/client.js"
-export { ModelsDevError, type ClientOptions, type ModelsClient } from "./effect/client.js"
+export { ModelsDevError, type ClientOptions, type ModelsClient, type RequestOptions } from "./effect/client.js"
 export type * from "./types.js"
diff --git a/packages/sdk/src/effect/client.ts b/packages/sdk/src/effect/client.ts
index 0eb98835d55..ea005cf8f82 100644
--- a/packages/sdk/src/effect/client.ts
+++ b/packages/sdk/src/effect/client.ts
@@ -1,6 +1,6 @@
 import { Context, Effect, Layer, Schema } from "effect"
 import { HttpClient, HttpClientResponse } from "effect/unstable/http"
-import type { Catalog, ModelMetadataMap, ProviderMap } from "../types.js"
+import type { Catalog, ModelMetadataMap, ModelType, ProviderMap } from "../types.js"
 
 /** The only error in the failure channel of client methods. Wraps the underlying `HttpClientError` as `cause`. */
 export class ModelsDevError extends Schema.TaggedErrorClass<ModelsDevError>()("ModelsDevError", {
@@ -14,6 +14,11 @@ export interface ClientOptions {
   readonly headers?: Record<string, string>
 }
 
+export interface RequestOptions {
+  /** Specialized model types to include. Omit for standard models; use `"all"` for the complete catalog. */
+  readonly modelTypes?: "all" | readonly ModelType[]
+}
+
 /**
  * Creates a stateless models.dev client on top of the `HttpClient` service
  * from the environment (`FetchHttpClient.layer`, `NodeHttpClient.layer`, or a
@@ -26,9 +31,17 @@ export const make = (options?: ClientOptions) =>
     const baseUrl = options?.baseUrl ?? "https://models.dev"
     const base = baseUrl.endsWith("/") ? baseUrl : baseUrl + "/"
 
-    const get = <A>(path: string): Effect.Effect<A, ModelsDevError> =>
-      http
-        .get(new URL(path, base), {
+    const get = <A>(path: string, requestOptions?: RequestOptions): Effect.Effect<A, ModelsDevError> => {
+      const url = new URL(path, base)
+      const modelTypes = requestOptions?.modelTypes
+      if (modelTypes === "all") {
+        url.searchParams.set("type", "all")
+      } else if (modelTypes && modelTypes.length > 0) {
+        url.searchParams.set("type", modelTypes.join(","))
+      }
+
+      return http
+        .get(url, {
           headers: options?.headers,
         })
         .pipe(
@@ -37,14 +50,15 @@ export const make = (options?: ClientOptions) =>
           Effect.map((data) => data as A),
           Effect.mapError((cause) => new ModelsDevError({ cause })),
         )
+    }
 
     return {
       /** All providers with their models, pricing, and limits (`/api.json`). */
-      providers: () => get<ProviderMap>("api.json"),
+      providers: (options?: RequestOptions) => get<ProviderMap>("api.json", options),
       /** Provider-agnostic model metadata (`/models.json`). */
-      models: () => get<ModelMetadataMap>("models.json"),
+      models: (options?: RequestOptions) => get<ModelMetadataMap>("models.json", options),
       /** Providers and model metadata in a single request (`/catalog.json`). */
-      catalog: () => get<Catalog>("catalog.json"),
+      catalog: (options?: RequestOptions) => get<Catalog>("catalog.json", options),
     }
   })
 
diff --git a/packages/sdk/src/types.ts b/packages/sdk/src/types.ts
index 92999e2dea7..250551d5e7a 100644
--- a/packages/sdk/src/types.ts
+++ b/packages/sdk/src/types.ts
@@ -80,6 +80,9 @@ export interface ModelCost extends Cost {
 /** Input/output data types a model supports. */
 export type Modality = "text" | "audio" | "image" | "video" | "pdf"
 
+/** A model's specialized behavioral contract. Omitted for standard generative models. */
+export type ModelType = "decision"
+
 export interface Modalities {
   input: Modality[]
   output: Modality[]
@@ -143,6 +146,7 @@ export interface BenchmarkResult {
 export interface ModelMetadata {
   /** Canonical model ID, e.g. "anthropic/claude-opus-4-6". */
   id: string
+  type?: ModelType
   name: string
   description: string
   family?: ModelFamily
@@ -208,6 +212,7 @@ export interface ModelProviderConfig {
 export interface Model {
   /** Provider-scoped model ID, e.g. "claude-opus-4-6". */
   id: string
+  type?: ModelType
   name: string
   description: string
   family?: ModelFamily
diff --git a/packages/sdk/test/client.test.ts b/packages/sdk/test/client.test.ts
index 3f898ef9157..61bc8fda104 100644
--- a/packages/sdk/test/client.test.ts
+++ b/packages/sdk/test/client.test.ts
@@ -40,6 +40,17 @@ test("models() and catalog() hit their endpoints", async () => {
   expect(calls.map((call) => call.url.href)).toEqual(["https://models.dev/models.json", "https://models.dev/catalog.json"])
 })
 
+test("catalog endpoints encode model type filters", async () => {
+  const { calls, fetch } = stub({})
+  const client = Models.make({ fetch })
+  await client.providers({ modelTypes: ["decision"] })
+  await client.models({ modelTypes: "all" })
+  expect(calls.map((call) => call.url.href)).toEqual([
+    "https://models.dev/api.json?type=decision",
+    "https://models.dev/models.json?type=all",
+  ])
+})
+
 test("baseUrl with subpath is preserved, with or without trailing slash", async () => {
   const { calls, fetch } = stub({})
   await Models.make({ fetch, baseUrl: "https://example.com/mirror" }).providers()
diff --git a/packages/sdk/test/effect.test.ts b/packages/sdk/test/effect.test.ts
index 943acedbeb7..025e4a190e3 100644
--- a/packages/sdk/test/effect.test.ts
+++ b/packages/sdk/test/effect.test.ts
@@ -42,6 +42,20 @@ test("models() and catalog() hit their endpoints, baseUrl subpath preserved", as
   ])
 })
 
+test("catalog endpoints encode model type filters", async () => {
+  const { requests, layer } = stub({})
+  const program = Effect.gen(function* () {
+    const client = yield* Models.make()
+    yield* client.providers({ modelTypes: ["decision"] })
+    yield* client.catalog({ modelTypes: "all" })
+  })
+  await program.pipe(Effect.provide(layer), Effect.runPromise)
+  expect(requests.map((request) => request.url)).toEqual([
+    "https://models.dev/api.json?type=decision",
+    "https://models.dev/catalog.json?type=all",
+  ])
+})
+
 test("custom headers are sent", async () => {
   const { requests, layer } = stub({})
   const program = Effect.gen(function* () {
diff --git a/packages/web/script/build.ts b/packages/web/script/build.ts
index f387bccd848..8866aa8e435 100755
--- a/packages/web/script/build.ts
+++ b/packages/web/script/build.ts
@@ -1,6 +1,11 @@
 #!/usr/bin/env bun
 
 import { RenderedPages, Providers, Models, renderDocument } from "../src/render";
+import {
+  filterCatalogByModelType,
+  MODEL_TYPES,
+  type ModelTypeFilter,
+} from "@models.dev/core";
 import fs from "fs/promises";
 import path from "path";
 
@@ -74,15 +79,18 @@ for (const [route, rendered] of RenderedPages) {
   await Bun.write(filePath, renderDocument(template, rendered));
 }
 
-await Bun.write("./dist/api.json", JSON.stringify(Providers));
-await Bun.write(
-  "./dist/catalog.json",
-  JSON.stringify({ models: Models, providers: Providers }),
-);
-await Bun.write("./dist/models.json", JSON.stringify(Models));
-
-await fs.rename("./dist/api.json", "./dist/_api.json");
-await fs.rename("./dist/catalog.json", "./dist/_catalog.json");
-await fs.rename("./dist/models.json", "./dist/_models.json");
+const catalog = { models: Models, providers: Providers };
+const variants: Array<[suffix: string, filter: ModelTypeFilter]> = [
+  ["", "default"],
+  ["-all", "all"],
+  ...MODEL_TYPES.map((type) => [`-${type}`, [type]] as const),
+];
+
+for (const [suffix, filter] of variants) {
+  const filtered = filterCatalogByModelType(catalog, filter);
+  await Bun.write(`./dist/_api${suffix}.json`, JSON.stringify(filtered.providers));
+  await Bun.write(`./dist/_models${suffix}.json`, JSON.stringify(filtered.models));
+  await Bun.write(`./dist/_catalog${suffix}.json`, JSON.stringify(filtered));
+}
 
 await fs.rm("./dist/index.html", { force: true });
diff --git a/packages/web/src/render.tsx b/packages/web/src/render.tsx
index 93c73c6e66e..34632f57b1b 100644
--- a/packages/web/src/render.tsx
+++ b/packages/web/src/render.tsx
@@ -1497,21 +1497,22 @@ function HelpDialog() {
         <h2>API</h2>
         <p>
           You can access provider data, provider-agnostic model metadata, or the
-          combined catalog through JSON endpoints.
+          combined catalog through JSON endpoints. Specialized model types are
+          omitted by default; the site includes all model types.
         </p>
         <div class="code-block">
           <code>
-            curl <a href="/api.json">https://models.dev/api.json</a>
+            curl <a href="/api.json?type=all">https://models.dev/api.json?type=all</a>
           </code>
         </div>
         <div class="code-block">
           <code>
-            curl <a href="/models.json">https://models.dev/models.json</a>
+            curl <a href="/models.json?type=all">https://models.dev/models.json?type=all</a>
           </code>
         </div>
         <div class="code-block">
           <code>
-            curl <a href="/catalog.json">https://models.dev/catalog.json</a>
+            curl <a href="/catalog.json?type=all">https://models.dev/catalog.json?type=all</a>
           </code>
         </div>
         <h2>Logos</h2>
diff --git a/packages/web/src/server.ts b/packages/web/src/server.ts
index 5adb9e9fc60..f43e6ce830a 100644
--- a/packages/web/src/server.ts
+++ b/packages/web/src/server.ts
@@ -1,5 +1,12 @@
 import Index from "../index.html";
 import { getRenderedPage, Models, Providers, renderDocument } from "./render";
+import {
+  filterCatalogByModelType,
+  filterModelsByModelType,
+  filterProvidersByModelType,
+  InvalidModelTypeError,
+  parseModelTypes,
+} from "@models.dev/core";
 import path from "path";
 
 const assetPort = Number(Bun.env.ASSET_PORT ?? 16000);
@@ -99,30 +106,37 @@ Bun.serve({
         },
       });
     },
-    "/api.json": () =>
-      Response.json(Providers, {
-        headers: {
-          "Cache-Control": "public, max-age=3600",
-        },
-      }),
-    "/models.json": () =>
-      Response.json(Models, {
-        headers: {
-          "Cache-Control": "public, max-age=3600",
-        },
-      }),
-    "/catalog.json": () =>
-      Response.json(
-        { models: Models, providers: Providers },
-        {
-          headers: {
-            "Cache-Control": "public, max-age=3600",
-          },
-        },
-      ),
+    "/api.json": (req) => catalogResponse(req, "api"),
+    "/models.json": (req) => catalogResponse(req, "models"),
+    "/catalog.json": (req) => catalogResponse(req, "catalog"),
   },
 });
 
+function catalogResponse(req: Request, endpoint: "api" | "models" | "catalog") {
+  let filter;
+  try {
+    filter = parseModelTypes(new URL(req.url).searchParams.get("type"));
+  } catch (error) {
+    if (!(error instanceof InvalidModelTypeError)) throw error;
+    return Response.json({ error: error.message }, { status: 400 });
+  }
+
+  const value = endpoint === "api"
+    ? filterProvidersByModelType(Providers, filter)
+    : endpoint === "models"
+      ? filterModelsByModelType(Models, filter)
+      : filterCatalogByModelType(
+          { models: Models, providers: Providers },
+          filter,
+        );
+
+  return Response.json(value, {
+    headers: {
+      "Cache-Control": "public, max-age=3600",
+    },
+  });
+}
+
 const server = Bun.serve({
   development: true,
   hostname: "0.0.0.0",

From 79ed8a751dc10c55f982ab1d233c59a48abc755c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 16:25:49 +0000
Subject: [PATCH 1724/2012] chore(sync): update Kilo model catalog (#7580)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index e4d3acd84bf..4a153e157f4 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5478
-output = 1.6434
-cache_read = 0.01826
+input = 0.54252
+output = 1.62756
+cache_read = 0.018084
 
 [limit]
 context = 1_024_000

From 8540388b7fff1a0ae57f3de053ff7befb1f2125e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 16:26:06 +0000
Subject: [PATCH 1725/2012] chore(sync): update OpenRouter model catalog
 (#7581)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml   | 4 ++--
 providers/openrouter/models/tencent/hy3.toml                | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 4 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7191760d710..9b249382289 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5478
-output = 1.6434
-cache_read = 0.01826
+input = 0.54252
+output = 1.62756
+cache_read = 0.018084
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
index 91402212177..d857fa53dab 100644
--- a/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
+++ b/providers/openrouter/models/qwen/qwen3-vl-30b-a3b-instruct.toml
@@ -12,8 +12,8 @@ knowledge = "2025-03-31"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.7
+input = 0.13
+output = 0.52
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 78806039bab..70af1d85157 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5478
-output = 1.6434
-cache_read = 0.01826
+input = 0.54252
+output = 1.62756
+cache_read = 0.018084
 
 [limit]
 context = 1_048_576

From bd3424ccb06184edcfc14356cbd460200b2dc1f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 17:22:05 +0000
Subject: [PATCH 1726/2012] chore(sync): update OpenRouter model catalog
 (#7583)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 6 +++---
 3 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 9b249382289..fe38fdd1760 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.54252
-output = 1.62756
-cache_read = 0.018084
+input = 0.53856
+output = 1.61568
+cache_read = 0.017136
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 70af1d85157..809399be81d 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.54252
-output = 1.62756
-cache_read = 0.018084
+input = 0.53856
+output = 1.61568
+cache_read = 0.017136
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index f42b1197f93..c7cc1981beb 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8442
-output = 2.6532
-cache_read = 0.15678
+input = 0.8316
+output = 2.6136
+cache_read = 0.15444
 
 [limit]
 context = 1_310_720

From 2d25c71b1ab3000b8492798e0108cfbe8e2a4c10 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 17:22:20 +0000
Subject: [PATCH 1727/2012] chore(sync): update Kilo model catalog (#7582)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 ++-
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 providers/kilo/models/~z-ai/glm-latest.toml            |  6 +++---
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 4a153e157f4..f1d4ad4c3b8 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.54252
-output = 1.62756
-cache_read = 0.018084
+input = 0.53856
+output = 1.61568
+cache_read = 0.017136
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index c5af7ff24a4..4331c62d136 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8442
-output = 2.6532
-cache_read = 0.15678
+input = 0.8316
+output = 2.6136
+cache_read = 0.15444
 
 [limit]
 context = 1_048_576

From ba313562117db54326bcc8b3933024f92aa61fc9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 18:28:15 +0000
Subject: [PATCH 1728/2012] chore(sync): update Kilo model catalog (#7585)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3.5-lightning.toml | 2 +-
 providers/kilo/models/qwen/qwen3.6-35b-a3b.toml          | 4 ++--
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index 6759948a30e..5244dc1ac02 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -6,7 +6,7 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.04
+input = 0.065
 output = 0.18
 
 [limit]
diff --git a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
index 1385363812a..21f9408b4a8 100644
--- a/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.1
-output = 0.9
+input = 0.15
+output = 1
 cache_read = 0.05
 
 [limit]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index f1d4ad4c3b8..c995568e34b 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.53856
-output = 1.61568
-cache_read = 0.017136
+input = 0.53328
+output = 1.59984
+cache_read = 0.016968
 
 [limit]
 context = 1_048_576

From e755d2ad97947686bb6b8221db177a53132fe435 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 18:28:19 +0000
Subject: [PATCH 1729/2012] chore(sync): update OpenRouter model catalog
 (#7584)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml       | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index fe38fdd1760..e3772508842 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53856
-output = 1.61568
-cache_read = 0.017136
+input = 0.53328
+output = 1.59984
+cache_read = 0.016968
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
index ac95b282785..a223e74df58 100644
--- a/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-35b-a3b.toml
@@ -6,8 +6,8 @@ base_model = "alibaba/qwen3.6-35b-a3b"
 type = "toggle"
 
 [cost]
-input = 0.1
-output = 0.9
+input = 0.15
+output = 1
 cache_read = 0.05
 
 [limit]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 809399be81d..45e0c0ee8cc 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53856
-output = 1.61568
-cache_read = 0.017136
+input = 0.53328
+output = 1.59984
+cache_read = 0.016968
 
 [limit]
 context = 1_048_576

From cf992a84f6132bdfc2ad6b8f408c3e1b36536857 Mon Sep 17 00:00:00 2001
From: Frank <frank@anoma.ly>
Date: Sun, 20 Sep 2026 14:59:35 -0400
Subject: [PATCH 1730/2012] Delete jev-latest.toml

---
 providers/opencode/models/jev-latest.toml | 6 ------
 1 file changed, 6 deletions(-)
 delete mode 100644 providers/opencode/models/jev-latest.toml

diff --git a/providers/opencode/models/jev-latest.toml b/providers/opencode/models/jev-latest.toml
deleted file mode 100644
index 409cf1a8931..00000000000
--- a/providers/opencode/models/jev-latest.toml
+++ /dev/null
@@ -1,6 +0,0 @@
-# Pricing source: https://typesafe.ai/blog/introducing-system-one-models-and-jev
-base_model = "typesafe/jev-latest"
-
-[cost]
-input = 0.042
-output = 0

From 546157ba4e6e59956ed56ef16545e86709a76055 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 19:20:59 +0000
Subject: [PATCH 1731/2012] chore(sync): update OpenRouter model catalog
 (#7588)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index e3772508842..0db1bf12100 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53328
-output = 1.59984
-cache_read = 0.016968
+input = 0.528
+output = 1.584
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 45e0c0ee8cc..813379229ea 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53328
-output = 1.59984
-cache_read = 0.016968
+input = 0.528
+output = 1.584
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576

From f9bd1ec77589641514cdf17a4f358f3f15d842f6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 19:21:07 +0000
Subject: [PATCH 1732/2012] chore(sync): update Kilo model catalog (#7589)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index c995568e34b..7c557b67a56 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.53328
-output = 1.59984
-cache_read = 0.016968
+input = 0.528
+output = 1.584
+cache_read = 0.0168
 
 [limit]
 context = 1_048_576

From 4a583e6170b8dabd19cac0ab602a91920ef47b5b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 20:24:51 +0000
Subject: [PATCH 1733/2012] chore(sync): update OpenRouter model catalog
 (#7590)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 2 +-
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml     | 8 ++++----
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 2 +-
 5 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 5e9065ff206..a96d85eba1b 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -11,7 +11,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.08
+output = 0.12
 cache_read = 0.016
 
 [limit]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 0417e3027d5..2186c60b6fb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -11,10 +11,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.2156
-output = 0.6468
-cache_read = 0.00686
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
-output = 262_144
+output = 943_718
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 0db1bf12100..6d3a571d213 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.528
-output = 1.584
-cache_read = 0.0168
+input = 0.52668
+output = 1.58004
+cache_read = 0.017556
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 813379229ea..6f16213302e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.528
-output = 1.584
-cache_read = 0.0168
+input = 0.52668
+output = 1.58004
+cache_read = 0.017556
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index cd84b32afe3..f83290ea281 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.08
+output = 0.12
 cache_read = 0.016
 
 [limit]

From 3118014fb461b48f37d26b0bd7ab5a908408d98d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 20:24:58 +0000
Subject: [PATCH 1734/2012] chore(sync): update Kilo model catalog (#7591)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml  |  2 +-
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  2 +-
 4 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
index 41e3e4d74b4..a280435f49f 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -12,4 +12,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
-output = 262_144
+output = 943_718
diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 7c557b67a56..aac53c4f859 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.528
-output = 1.584
-cache_read = 0.0168
+input = 0.52668
+output = 1.58004
+cache_read = 0.017556
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 2c01541878c..0662392bcff 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.08
+output = 0.12
 cache_read = 0.016
 
 [limit]

From 2b269a637f367f958b0788ad6cca8f75b4fe68da Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 21:23:14 +0000
Subject: [PATCH 1735/2012] chore(sync): update OpenRouter model catalog
 (#7593)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml      | 2 +-
 .../openrouter/models/~deepseek/deepseek-v4-flash-latest.toml   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index a96d85eba1b..be2f907d844 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -11,7 +11,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.12
+output = 0.16
 cache_read = 0.016
 
 [limit]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index f83290ea281..c98a00af617 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.12
+output = 0.16
 cache_read = 0.016
 
 [limit]

From e2b20e7430e1d42fbda3dab37e72fa1d4a5182e6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 21:23:21 +0000
Subject: [PATCH 1736/2012] chore(sync): update NanoGPT model catalog (#7594)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/z-ai/glm-4.7-flash-original.toml   | 25 -------------------
 .../z-ai/glm-4.7-flash-original:thinking.toml | 25 -------------------
 2 files changed, 50 deletions(-)
 delete mode 100644 providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
 delete mode 100644 providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml

diff --git a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
deleted file mode 100644
index 3491643442e..00000000000
--- a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "GLM 4.7 Flash Original"
-description = "GLM-4.7-Flash is a lightweight 30B model optimized for coding and agentic tasks. Balances high performance with efficiency, perfect for local deployment."
-family = "glm-flash"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.07
-output = 0.4
-cache_read = 0.035
-
-[limit]
-context = 200_000
-input = 200_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml b/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
deleted file mode 100644
index 65480647fea..00000000000
--- a/providers/nano-gpt/models/z-ai/glm-4.7-flash-original:thinking.toml
+++ /dev/null
@@ -1,25 +0,0 @@
-name = "GLM 4.7 Flash Original Thinking"
-description = "GLM-4.7-Flash with extended thinking capabilities for complex reasoning. Lightweight 30B model optimized for coding and agentic tasks."
-family = "glm-flash"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-tool_call = true
-structured_output = false
-open_weights = true
-reasoning_options = []
-
-[cost]
-input = 0.07
-output = 0.4
-cache_read = 0.035
-
-[limit]
-context = 200_000
-input = 200_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

From 3c1aed35ba863dde45f0233a6d78a48cc8fb9913 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 21:23:25 +0000
Subject: [PATCH 1737/2012] chore(sync): update Kilo model catalog (#7595)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 0662392bcff..7ead984259a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.12
+output = 0.16
 cache_read = 0.016
 
 [limit]

From 34293334cff48dca9f9b914243b8a51aa9ba0efa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 22:24:00 +0000
Subject: [PATCH 1738/2012] chore(sync): update Kilo model catalog (#7597)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml                 | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index aac53c4f859..02af10ccee9 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.52668
-output = 1.58004
-cache_read = 0.017556
+input = 0.52404
+output = 1.57212
+cache_read = 0.017468
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 7ead984259a..b586ef3f991 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,12 +16,12 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.16
-cache_read = 0.016
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 4331c62d136..2237628e2fc 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8316
-output = 2.6136
-cache_read = 0.15444
+input = 0.7728
+output = 2.4288
+cache_read = 0.14352
 
 [limit]
 context = 1_048_576

From 28edbfdc1a374a69095558a98cd03ba9fe1a4c22 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 22:24:07 +0000
Subject: [PATCH 1739/2012] chore(sync): update OpenRouter model catalog
 (#7598)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml           | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 6d3a571d213..46ca79925cf 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52668
-output = 1.58004
-cache_read = 0.017556
+input = 0.52404
+output = 1.57212
+cache_read = 0.017468
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 6f16213302e..b3b53f82fcd 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52668
-output = 1.58004
-cache_read = 0.017556
+input = 0.52404
+output = 1.57212
+cache_read = 0.017468
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index c98a00af617..12c99d28b38 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,12 +21,12 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.16
-cache_read = 0.016
+output = 0.1
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index c7cc1981beb..709b8881ffa 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.8316
-output = 2.6136
-cache_read = 0.15444
+input = 0.7728
+output = 2.4288
+cache_read = 0.14352
 
 [limit]
 context = 1_310_720

From 27efbe920eea10c2c411f2620f58f2170f3b8a5f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 23:23:01 +0000
Subject: [PATCH 1740/2012] chore(sync): update OpenRouter model catalog
 (#7600)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 6 +++---
 3 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 46ca79925cf..f4c4b33d3cc 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52404
-output = 1.57212
-cache_read = 0.017468
+input = 0.52272
+output = 1.56816
+cache_read = 0.016632
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index b3b53f82fcd..23ba1c5d915 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52404
-output = 1.57212
-cache_read = 0.017468
+input = 0.52272
+output = 1.56816
+cache_read = 0.016632
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 12c99d28b38..c98a00af617 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,12 +21,12 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.1
-cache_read = 0.01
+output = 0.16
+cache_read = 0.016
 
 [limit]
 context = 1_310_720
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From 9c20af0e283a3dea2136291dbc7b059acbd820d4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Sun, 20 Sep 2026 23:23:05 +0000
Subject: [PATCH 1741/2012] chore(sync): update Kilo model catalog (#7601)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 ++-
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  6 +++---
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 02af10ccee9..6bd317a45ab 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.52404
-output = 1.57212
-cache_read = 0.017468
+input = 0.52272
+output = 1.56816
+cache_read = 0.016632
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index b586ef3f991..7ead984259a 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,12 +16,12 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.1
-cache_read = 0.01
+output = 0.16
+cache_read = 0.016
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 943_718
 
 [modalities]
 input = ["text"]

From 91c2e781204dc9fb57c5c50a086b7c941748af2c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 01:04:42 +0000
Subject: [PATCH 1742/2012] chore(sync): update Kilo model catalog (#7603)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml  | 1 -
 .../kilo/models/meta-llama/llama-3.1-70b-instruct.toml    | 2 +-
 providers/kilo/models/meta-llama/llama-4-maverick.toml    | 2 +-
 providers/kilo/models/tencent/hy3.toml                    | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml               | 8 ++++----
 6 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..6fc67901197 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index fab1dd0fc03..537ab8cd5fd 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -15,7 +15,7 @@ input = 0.1875
 output = 0.6525
 
 [limit]
-context = 128_000
+context = 1_048_576
 output = 16_384
 
 [modalities]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..2bdfa3cbd7b 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6bd317a45ab..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.52272
-output = 1.56816
-cache_read = 0.016632
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2237628e2fc..b6ea0fea56d 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7728
-output = 2.4288
-cache_read = 0.14352
+input = 0.7735
+output = 2.431
+cache_read = 0.127075
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From c44d7ad670bb33f55d4e8955049112ae30c5b69f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 01:04:47 +0000
Subject: [PATCH 1743/2012] chore(sync): update OpenRouter model catalog
 (#7604)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/ibm-granite/granite-4.2-8b.toml     | 6 +++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml         | 6 +++---
 .../openrouter/models/meta-llama/llama-4-maverick.toml    | 4 ++--
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 9 files changed, 28 insertions(+), 29 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 98ee0445042..57d98266566 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.03556
-output = 0.07112
-cache_read = 0.007112
+input = 0.089866
+output = 0.179732
+cache_read = 0.017973
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index f4c4b33d3cc..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52272
-output = 1.56816
-cache_read = 0.016632
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 5bc792c3ae0..9be6b267131 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.422298
-output = 0.844596
-cache_read = 0.035192
+input = 0.9483
+output = 1.8966
+cache_read = 0.079025
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
index 62943cfaadb..1b683dd536e 100644
--- a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
+++ b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.06
-output = 0.25
-cache_read = 0.015
+input = 0.1
+output = 0.15
+cache_read = 0.05
 
 [limit]
 context = 131_072
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 881a90e0d15..1352c1ee599 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.4
-output = 0.4
+input = 0.72
+output = 0.72
 
 [limit]
 context = 131_072
-output = 16_384
+output = 8_192
diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index b9b42749c83..de6e5240ff9 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -12,8 +12,8 @@ knowledge = "2024-08-31"
 open_weights = true
 
 [cost]
-input = 0.1875
-output = 0.6525
+input = 0.2
+output = 0.8
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 23ba1c5d915..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.52272
-output = 1.56816
-cache_read = 0.016632
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 709b8881ffa..25e0814a971 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7728
-output = 2.4288
-cache_read = 0.14352
+input = 0.7735
+output = 2.431
+cache_read = 0.127075
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 34e14cb7160470457fec09c934a2497ee4f4e337 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 01:04:54 +0000
Subject: [PATCH 1744/2012] chore(sync): update NanoGPT model catalog (#7602)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From fdee7705c5d1254d4f169c3a27585d07264c163f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 01:41:41 +0000
Subject: [PATCH 1745/2012] chore(sync): update OpenRouter model catalog
 (#7606)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-27b.toml         | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index 87b348fc2c9..64a9a2f6128 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -12,8 +12,8 @@ values = ["low", "medium", "xhigh"]
 
 [cost]
 input = 0.2
-output = 2.55
-cache_read = 0.085
+output = 2.5
+cache_read = 0.05
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..dfb8b85c369 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 76488ffa1fb42e92850d7bba321815492fefb462 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 01:41:49 +0000
Subject: [PATCH 1746/2012] chore(sync): update Kilo model catalog (#7605)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..46e3c60f672 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 62c3e418a832af3bfe6d721be8b8014feef550aa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 02:38:35 +0000
Subject: [PATCH 1747/2012] chore(sync): update Kilo model catalog (#7608)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml              | 6 +++---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml         | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index fd81e087ca6..31838f83a2c 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.675
+output = 9.38
+cache_read = 0.1943
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 9e0f74cdd43..2193a017b17 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.13
-output = 0.52
-cache_read = 0.0026
+input = 0.12
+output = 0.48
+cache_read = 0.0036
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 631c1758bcd..71dba53f553 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.675
+output = 9.38
+cache_read = 0.1943
 
 [limit]
 context = 1_048_576

From 52eff77bbe3c9b75a83adf769120305150c359c3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 02:38:39 +0000
Subject: [PATCH 1748/2012] chore(sync): update OpenRouter model catalog
 (#7609)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml         | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 5 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 57d98266566..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.089866
-output = 0.179732
-cache_read = 0.017973
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 9be6b267131..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.9483
-output = 1.8966
-cache_read = 0.079025
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 22b1407542a..3b2924c9c2d 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.675
+output = 9.38
+cache_read = 0.1943
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 643ffd542be..56426182daf 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.13
-output = 0.52
-cache_read = 0.0026
+input = 0.12
+output = 0.48
+cache_read = 0.0036
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 7031b748f98..731915d3eeb 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.675
+output = 9.38
+cache_read = 0.1943
 
 [limit]
 context = 1_048_576

From b44a3432f47d0ec1d137eaa26bad06ee99897539 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 03:33:34 +0000
Subject: [PATCH 1749/2012] chore(sync): update OpenRouter model catalog
 (#7611)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/moonshotai/kimi-k3.toml      | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 3b2924c9c2d..22b1407542a 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.675
-output = 9.38
-cache_read = 0.1943
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 731915d3eeb..7031b748f98 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.675
-output = 9.38
-cache_read = 0.1943
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 context = 1_048_576

From 3764983dc09fe350e6e768a2867a531476ff7154 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 03:33:41 +0000
Subject: [PATCH 1750/2012] chore(sync): update Kilo model catalog (#7610)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml      | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index 31838f83a2c..fd81e087ca6 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.675
-output = 9.38
-cache_read = 0.1943
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 71dba53f553..631c1758bcd 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.675
-output = 9.38
-cache_read = 0.1943
+input = 1.7
+output = 8.5
+cache_read = 0.17
 
 [limit]
 context = 1_048_576

From 1c43b109c5d2d1f538069005bd8f6e6f4939bb40 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 04:32:58 +0000
Subject: [PATCH 1751/2012] chore(sync): update NanoGPT model catalog (#7613)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 4e14674202195bc855e571cebc4d1829c9e54e93 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 04:33:04 +0000
Subject: [PATCH 1752/2012] chore(sync): update Kilo model catalog (#7612)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 46e3c60f672..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 2a91d70292280963d53f3df71fd48611a031e8a8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 04:33:08 +0000
Subject: [PATCH 1753/2012] chore(sync): update OpenRouter model catalog
 (#7614)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index dfb8b85c369..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From cfc5669256db6c46f77060f8b0f56b53e4d8beb9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 05:32:43 +0000
Subject: [PATCH 1754/2012] chore(sync): update Eden AI model catalog (#7616)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml  | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml b/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
index 33e52d4c7a4..be84a18d071 100644
--- a/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
+++ b/providers/edenai/models/cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct.toml
@@ -1,7 +1,7 @@
 base_model = "alibaba/qwen2.5-coder-32b-instruct"
 name = "Qwen2.5-Coder-32B-Instruct (Cloudflare)"
 tool_call = false
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.66

From 5ec54bd903bfe2dc13e1a654dd450a68463e652d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 06:52:10 +0000
Subject: [PATCH 1755/2012] chore(sync): update OpenRouter model catalog
 (#7618)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..dfb8b85c369 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From b8647dc5b85a2bd941c20340a5c90f0a56b7a522 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 06:52:13 +0000
Subject: [PATCH 1756/2012] chore(sync): update NanoGPT model catalog (#7621)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 5cd0647e7143f4716f1c3363f4685e5aea599b24 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 06:52:18 +0000
Subject: [PATCH 1757/2012] chore(sync): update LLM Gateway model catalog
 (#7619)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/scx-ai-gp/deepseek-v4.1-flash.toml      | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/scx-ai-gp/deepseek-v4.1-flash.toml

diff --git a/providers/llmgateway-providers/models/scx-ai-gp/deepseek-v4.1-flash.toml b/providers/llmgateway-providers/models/scx-ai-gp/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..9ea85540b29
--- /dev/null
+++ b/providers/llmgateway-providers/models/scx-ai-gp/deepseek-v4.1-flash.toml
@@ -0,0 +1,14 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash (SCX.ai)"
+attachment = false
+reasoning = false
+tool_call = false
+structured_output = false
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.01
+
+[modalities]
+input = ["text"]

From d8d98a4ab0ed3f25d24c50e7c28e06217379ffe1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 06:52:24 +0000
Subject: [PATCH 1758/2012] chore(sync): update Kilo model catalog (#7620)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..46e3c60f672 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.7
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 962a7ceec798b906e2fb3a027b207cb3581ae42b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 08:37:44 +0000
Subject: [PATCH 1759/2012] chore(sync): update NanoGPT model catalog (#7623)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/qwen/qwen3-30b-a3b.toml   |  1 +
 .../models/qwen/qwen3.8-27b-hemmingway.toml   | 25 +++++++++++++++++++
 2 files changed, 26 insertions(+)
 create mode 100644 providers/nano-gpt/models/qwen/qwen3.8-27b-hemmingway.toml

diff --git a/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml b/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
index 3fe31cbc4c1..ef583b4ad7d 100644
--- a/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-30b-a3b.toml
@@ -1,6 +1,7 @@
 # Included in subscription
 base_model = "alibaba/qwen3-30b-a3b"
 reasoning = false
+tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/nano-gpt/models/qwen/qwen3.8-27b-hemmingway.toml b/providers/nano-gpt/models/qwen/qwen3.8-27b-hemmingway.toml
new file mode 100644
index 00000000000..e8e4e1ce33f
--- /dev/null
+++ b/providers/nano-gpt/models/qwen/qwen3.8-27b-hemmingway.toml
@@ -0,0 +1,25 @@
+name = "Qwen 3.8 27B Hemingway"
+description = "Qwen 3.8 27B Hemingway is an open-weight NVFP4 multimodal creative finetune for long-form prose, character dialogue, storytelling, and roleplay."
+family = "qwen"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = false
+open_weights = true
+reasoning_options = []
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.125
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 25c48bf4e195c5b7da7466969976df06c0e3c5c9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 08:38:17 +0000
Subject: [PATCH 1760/2012] chore(sync): update Cortecs model catalog (#7624)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/cortecs/models/qwen3-32b.toml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/cortecs/models/qwen3-32b.toml b/providers/cortecs/models/qwen3-32b.toml
index d6ad27de61b..32225e598f0 100644
--- a/providers/cortecs/models/qwen3-32b.toml
+++ b/providers/cortecs/models/qwen3-32b.toml
@@ -4,9 +4,8 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 0.089
-output = 0.312
+input = 0.179
+output = 0.697
 
 [limit]
-context = 32_000
-output = 32_000
+context = 16_384

From 60efdd3429eb7d396331e8988c327216c1e52393 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 10:28:21 +0000
Subject: [PATCH 1761/2012] chore(sync): update NanoGPT model catalog (#7628)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 5f15e38c6e8221f63afacb019ea7bfe7b84d4d27 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 10:28:24 +0000
Subject: [PATCH 1762/2012] chore(sync): update OpenRouter model catalog
 (#7629)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index dfb8b85c369..ea09a0d12b6 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From b3c181144b4157b1e68b1bb8d8d547267f0ca9ac Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 10:28:32 +0000
Subject: [PATCH 1763/2012] chore(sync): update Kilo model catalog (#7627)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 46e3c60f672..b14064bf897 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 1c56caef93a8e78727d6e859a1ee115cf133e3c8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 11:27:19 +0000
Subject: [PATCH 1764/2012] chore(sync): update NanoGPT model catalog (#7630)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml   | 2 +-
 .../models/qwen3-vl-235b-a22b-instruct-original.toml        | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
index a648f070f2e..8b97c180ac4 100644
--- a/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
+++ b/providers/nano-gpt/models/qwen/qwen3-vl-235b-a22b-instruct.toml
@@ -4,7 +4,7 @@ structured_output = false
 
 [cost]
 input = 0.3
-output = 1.2
+output = 1.9
 cache_read = 0.15
 
 [limit]
diff --git a/providers/nano-gpt/models/qwen3-vl-235b-a22b-instruct-original.toml b/providers/nano-gpt/models/qwen3-vl-235b-a22b-instruct-original.toml
index 2f39269ea3f..851b51ce84f 100644
--- a/providers/nano-gpt/models/qwen3-vl-235b-a22b-instruct-original.toml
+++ b/providers/nano-gpt/models/qwen3-vl-235b-a22b-instruct-original.toml
@@ -9,9 +9,9 @@ structured_output = false
 open_weights = false
 
 [cost]
-input = 0.5
-output = 1.2
-cache_read = 0.25
+input = 0.3
+output = 1.9
+cache_read = 0.15
 
 [limit]
 context = 32_768

From 48e0aaa37303ccd9da5b9c13cf2e7e3a6e039b13 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:27:49 +0000
Subject: [PATCH 1765/2012] chore(sync): update Kilo model catalog (#7632)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b14064bf897..653d45fd9d1 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,12 +15,12 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65472
+output = 1.96416
+cache_read = 0.021824
 
 [limit]
-context = 1_048_576
+context = 1_024_000
 output = 384_000
 
 [modalities]

From cfcc2393f3eebc260859697f7a5addc3f2f07e56 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:28:04 +0000
Subject: [PATCH 1766/2012] chore(sync): update OpenRouter model catalog
 (#7631)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..b8c1d29ced8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.05852
+output = 0.11704
+cache_read = 0.011704
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..2a431cd5648 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.951432
+output = 1.902864
+cache_read = 0.079286
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ea09a0d12b6..dd01ad1ecf9 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.65472
+output = 1.96416
+cache_read = 0.021824
 
 [limit]
 context = 1_048_576

From e379bc9252c87075160ecd47fc894bd4e74f098a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 14:29:39 +0000
Subject: [PATCH 1767/2012] chore(sync): update OpenRouter model catalog
 (#7634)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../models/meta-llama/llama-3.1-70b-instruct.toml         | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 6 files changed, 21 insertions(+), 20 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b8c1d29ced8..b044540752e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.05852
-output = 0.11704
-cache_read = 0.011704
+input = 0.05698
+output = 0.11396
+cache_read = 0.011396
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..c6c87510247 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 2a431cd5648..ae6a4dbd88c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.951432
-output = 1.902864
-cache_read = 0.079286
+input = 0.948126
+output = 1.896252
+cache_read = 0.079011
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
index 1352c1ee599..881a90e0d15 100644
--- a/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/openrouter/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -3,9 +3,9 @@ description = "Open Llama instruction model for multilingual chat, reasoning, an
 structured_output = true
 
 [cost]
-input = 0.72
-output = 0.72
+input = 0.4
+output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index dd01ad1ecf9..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.65472
-output = 1.96416
-cache_read = 0.021824
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 25e0814a971..709b8881ffa 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7735
-output = 2.431
-cache_read = 0.127075
+input = 0.7728
+output = 2.4288
+cache_read = 0.14352
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 074421e6fe8dbe32f450c3c27f1933ae181ffde0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 14:29:55 +0000
Subject: [PATCH 1768/2012] chore(sync): update Kilo model catalog (#7633)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 +
 .../kilo/models/meta-llama/llama-3.1-70b-instruct.toml |  2 +-
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 providers/kilo/models/~z-ai/glm-latest.toml            |  8 ++++----
 4 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
index 6fc67901197..881a90e0d15 100644
--- a/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
+++ b/providers/kilo/models/meta-llama/llama-3.1-70b-instruct.toml
@@ -8,4 +8,4 @@ output = 0.4
 
 [limit]
 context = 131_072
-output = 8_192
+output = 16_384
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 653d45fd9d1..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.65472
-output = 1.96416
-cache_read = 0.021824
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index b6ea0fea56d..2237628e2fc 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7735
-output = 2.431
-cache_read = 0.127075
+input = 0.7728
+output = 2.4288
+cache_read = 0.14352
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 88bb3d960afafee3ef107fac71100acb51ef1201 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:21:48 +0800
Subject: [PATCH 1769/2012] fix(deepseek): align V4 Pro effort with GA
 documentation (#7564)

* fix(deepseek): align V4 Pro effort with GA documentation

* chore: retrigger review workflow

The previous review run failed while installing opencode
("Failed to fetch version information"), so no code review ran.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/deepseek/models/deepseek-v4-pro.toml | 13 +++++++++----
 providers/deepseek/provider.toml               |  6 +++---
 2 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/providers/deepseek/models/deepseek-v4-pro.toml b/providers/deepseek/models/deepseek-v4-pro.toml
index 4d5aff0281b..6a81c8584d2 100644
--- a/providers/deepseek/models/deepseek-v4-pro.toml
+++ b/providers/deepseek/models/deepseek-v4-pro.toml
@@ -1,18 +1,23 @@
+# Toggle: thinking.type = enabled|disabled on /chat/completions.
+# Effort: reasoning_effort = low|high|max; default high.
+# Anthropic: output_config.effort = low|high|max; budget ignored.
+# The 2026-08-13 V4-Pro GA announcement explicitly introduces all three levels;
+# the pricing page maps API model deepseek-v4-pro to DeepSeek-V4-Pro-0813.
+# https://api-docs.deepseek.com/news/news260813/ (accessed 2026-09-20)
+# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-09-20)
+# https://api-docs.deepseek.com/quick_start/pricing/ (accessed 2026-09-20)
 # Reasoning tokens are billed at the output rate (no separate CoT price).
 # `completion_tokens_details.reasoning_tokens` is a subset of completion_tokens.
 # https://api-docs.deepseek.com/quick_start/pricing/ (accessed 2026-08-12)
 base_model = "deepseek/deepseek-v4-pro-0813"
 name = "DeepSeek V4 Pro"
-# OpenAI: `thinking.type = enabled|disabled`, `reasoning_effort = high|max`.
-# Anthropic: `thinking.type`, `output_config.effort = high|max`; budget ignored.
-# https://api-docs.deepseek.com/api/create-chat-completion (accessed 2026-06-25)
 
 [[reasoning_options]]
 type = "toggle"
 
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepseek/provider.toml b/providers/deepseek/provider.toml
index 11731946bfd..76274ae9de4 100644
--- a/providers/deepseek/provider.toml
+++ b/providers/deepseek/provider.toml
@@ -2,11 +2,11 @@ name = "DeepSeek"
 env = ["DEEPSEEK_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
 # OpenAI Chat is POST `/chat/completions`: `thinking.type = enabled|disabled`
-# and `reasoning_effort = low|high|max`. Flash maps low→low; Pro maps low→high;
-# xhigh maps to high (flash) or max (pro).
+# and `reasoning_effort = low|high|max`. Both current models map minimal→low,
+# medium/xhigh→high, and ultra→max; low/high/max remain distinct native levels.
 # Anthropic Messages is POST `/anthropic/v1/messages`: `thinking.type` and
 # `output_config.effort = low|high|max`; `thinking.budget_tokens` is ignored.
-# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-08-02)
+# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-09-20)
 # https://api-docs.deepseek.com/guides/anthropic_api (accessed 2026-06-25)
 doc = "https://api-docs.deepseek.com/quick_start/pricing"
 api = "https://api.deepseek.com"

From 843685c20f25423b52f5a85d15f340c2d2099c9a Mon Sep 17 00:00:00 2001
From: Pedro Santos <pedrocarlos.ti@gmail.com>
Date: Mon, 21 Sep 2026 16:22:26 +0100
Subject: [PATCH 1770/2012] fix(zai): add missing glm-4.6v-flash provider model
 (#7592)

Resolves broken zhipuai symlink to zai/glm-4.6v-flash.toml.

Fixes #1470

Co-authored-by: Pedro Santos <18473317+pc-gs@users.noreply.github.com>
---
 providers/zai/models/glm-4.6v-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/zai/models/glm-4.6v-flash.toml

diff --git a/providers/zai/models/glm-4.6v-flash.toml b/providers/zai/models/glm-4.6v-flash.toml
new file mode 100644
index 00000000000..d9793eab5e8
--- /dev/null
+++ b/providers/zai/models/glm-4.6v-flash.toml
@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-4.6v-flash"
+# Free Flash tier on Z.AI; cost shape mirrors providers/zai/models/glm-4.5-flash.toml.
+# Lab metadata (limits/modalities) lives in models/zhipuai/glm-4.6v-flash.toml.
+# https://z.ai/blog/glm-4.6v
+# https://huggingface.co/zai-org/GLM-4.6V-Flash
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From d0b52220faa55db8fdf0c9d3ede02d79a78dabc5 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:22:58 +0800
Subject: [PATCH 1771/2012] feat(aihubmix): add step-3.7-flash (#7559)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/step-3.7-flash.toml | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/aihubmix/models/step-3.7-flash.toml

diff --git a/providers/aihubmix/models/step-3.7-flash.toml b/providers/aihubmix/models/step-3.7-flash.toml
new file mode 100644
index 00000000000..d6dcca8c632
--- /dev/null
+++ b/providers/aihubmix/models/step-3.7-flash.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://platform.stepfun.com/docs/zh/guides/models/step-3.7-flash
+# Effort: reasoning_effort = low|medium|high on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "stepfun/step-3.7-flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.22
+output = 1.32
+cache_read = 0.044

From 137714d0fe5c9e0af9a18a5bb1f03359289e8cee Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:23:11 +0800
Subject: [PATCH 1772/2012] feat(aihubmix): add minimax-m3 (#7558)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/minimax-m3.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/aihubmix/models/minimax-m3.toml

diff --git a/providers/aihubmix/models/minimax-m3.toml b/providers/aihubmix/models/minimax-m3.toml
new file mode 100644
index 00000000000..2103f7586f0
--- /dev/null
+++ b/providers/aihubmix/models/minimax-m3.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://platform.minimaxi.com/docs/api-reference/text-chat-openai.md
+# Toggle: enable_thinking = true|false on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "minimax/MiniMax-M3"
+structured_output = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.288
+output = 1.152

From 1da5b84d0bb6d17235ff9235d44222c528bad1d5 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:23:46 +0800
Subject: [PATCH 1773/2012] feat(aihubmix): add gemini-3.1-flash-image (#7556)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 .../aihubmix/models/gemini-3.1-flash-image.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/aihubmix/models/gemini-3.1-flash-image.toml

diff --git a/providers/aihubmix/models/gemini-3.1-flash-image.toml b/providers/aihubmix/models/gemini-3.1-flash-image.toml
new file mode 100644
index 00000000000..1c51818cb89
--- /dev/null
+++ b/providers/aihubmix/models/gemini-3.1-flash-image.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-image
+# Effort: reasoning_effort = minimal|high on /v1/chat/completions.
+# Gemini native: generationConfig.thinkingConfig.thinkingLevel uses the corresponding uppercase levels.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "google/gemini-3.1-flash-image"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "high"]
+
+[cost]
+input = 0.5
+output = 3

From 1fc9f0f6b4cd03ce9bd8bb95e382da88fcdc8ad4 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:23:53 +0800
Subject: [PATCH 1774/2012] feat(aihubmix): add hy3 (#7555)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/hy3.toml | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/aihubmix/models/hy3.toml

diff --git a/providers/aihubmix/models/hy3.toml b/providers/aihubmix/models/hy3.toml
new file mode 100644
index 00000000000..a5f55cbba88
--- /dev/null
+++ b/providers/aihubmix/models/hy3.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://huggingface.co/tencent/Hy3
+# Effort: reasoning_effort = none|low|high on /v1/chat/completions.
+# Off is effort=none; AIHubMix no_think is normalized to none, so no separate toggle.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "tencent/hy3"
+structured_output = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high"]
+
+[cost]
+input = 0.1562
+output = 0.6248
+cache_read = 0.03905

From 4c5584c0449f80108f32c15b74272aee55f1bd2e Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:24:04 +0800
Subject: [PATCH 1775/2012] feat(aihubmix): add muse-spark-1.1 (#7554)

* feat(aihubmix): add muse-spark-1.1

* docs(aihubmix): cite Muse Spark 1.1 audio input declaration

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/muse-spark-1.1.toml | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/aihubmix/models/muse-spark-1.1.toml

diff --git a/providers/aihubmix/models/muse-spark-1.1.toml b/providers/aihubmix/models/muse-spark-1.1.toml
new file mode 100644
index 00000000000..f0c00f709bf
--- /dev/null
+++ b/providers/aihubmix/models/muse-spark-1.1.toml
@@ -0,0 +1,20 @@
+# Sources (accessed 2026-09-20):
+# AIHubMix catalog row model_id=muse-spark-1.1 declares
+# input_modalities="text,image,video,audio,pdf". Audio is a host-catalog override
+# relative to the lab entry; this is a catalog declaration, not an inference test.
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://dev.meta.ai/docs/models/
+# Effort: reasoning_effort = minimal|low|medium|high|xhigh on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "meta/muse-spark-1.1"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.375
+output = 4.675
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]

From d5fb44d21e599898970deb7e3c6851df089082c5 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:24:12 +0800
Subject: [PATCH 1776/2012] feat(aihubmix): add gemini-3.5-flash-lite (#7553)

* feat(aihubmix): add gemini-3.5-flash-lite

* docs(aihubmix): cite exact catalog price for Gemini 3.5 Flash Lite

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 .../models/gemini-3.5-flash-lite.toml         | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/aihubmix/models/gemini-3.5-flash-lite.toml

diff --git a/providers/aihubmix/models/gemini-3.5-flash-lite.toml b/providers/aihubmix/models/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..19c43bfae2f
--- /dev/null
+++ b/providers/aihubmix/models/gemini-3.5-flash-lite.toml
@@ -0,0 +1,19 @@
+# Sources (accessed 2026-09-20):
+# AIHubMix catalog row model_id=gemini-3.5-flash-lite publishes
+# pricing={"input":0.3,"output":2.499999,"cache_read":0.03}.
+# Preserve the host's exact published price; Google's 2.50 is a different host.
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://ai.google.dev/gemini-api/docs/thinking
+# Effort: reasoning_effort = minimal|low|medium|high on /v1/chat/completions.
+# Gemini native: generationConfig.thinkingConfig.thinkingLevel uses the corresponding uppercase levels.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "google/gemini-3.5-flash-lite"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.3
+output = 2.499999
+cache_read = 0.03

From 5d137a902e47a1f1de8df73832ee9a7d1b160f1b Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:24:19 +0800
Subject: [PATCH 1777/2012] feat(aihubmix): add gemini-3.1-flash-lite-image
 (#7552)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 .../models/gemini-3.1-flash-lite-image.toml    | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 providers/aihubmix/models/gemini-3.1-flash-lite-image.toml

diff --git a/providers/aihubmix/models/gemini-3.1-flash-lite-image.toml b/providers/aihubmix/models/gemini-3.1-flash-lite-image.toml
new file mode 100644
index 00000000000..14141e81cc6
--- /dev/null
+++ b/providers/aihubmix/models/gemini-3.1-flash-lite-image.toml
@@ -0,0 +1,18 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite-image
+# Effort: reasoning_effort = minimal|high on /v1/chat/completions.
+# Gemini native: generationConfig.thinkingConfig.thinkingLevel uses the corresponding uppercase levels.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "google/gemini-3.1-flash-lite-image"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "high"]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025

From 90db5bf0fc3c2af0a4460570a3be94a48939ac29 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:24:29 +0800
Subject: [PATCH 1778/2012] feat(aihubmix): add muse-spark-1.2 (#7551)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/muse-spark-1.2.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/aihubmix/models/muse-spark-1.2.toml

diff --git a/providers/aihubmix/models/muse-spark-1.2.toml b/providers/aihubmix/models/muse-spark-1.2.toml
new file mode 100644
index 00000000000..01d1fa8c78a
--- /dev/null
+++ b/providers/aihubmix/models/muse-spark-1.2.toml
@@ -0,0 +1,14 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://dev.meta.ai/docs/overview/
+# Effort: reasoning_effort = minimal|low|medium|high|xhigh on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "meta/muse-spark-1.2"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.375
+output = 4.675

From 21e40fe94fc66aa22261400a90096b493bc1d9dd Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:24:37 +0800
Subject: [PATCH 1779/2012] feat(aihubmix): add qwen3.8-omni-flash (#7550)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 .../aihubmix/models/qwen3.8-omni-flash.toml   | 23 +++++++++++++++++++
 1 file changed, 23 insertions(+)
 create mode 100644 providers/aihubmix/models/qwen3.8-omni-flash.toml

diff --git a/providers/aihubmix/models/qwen3.8-omni-flash.toml b/providers/aihubmix/models/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..8dea0f811e5
--- /dev/null
+++ b/providers/aihubmix/models/qwen3.8-omni-flash.toml
@@ -0,0 +1,23 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://www.alibabacloud.com/help/en/model-studio/qwen-omni
+# Toggle: enable_thinking = true|false on /v1/chat/completions.
+# Effort: reasoning_effort = low|medium|xhigh on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "alibaba/qwen3.8-omni-flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.1126
+output = 0.380025
+cache_read = 0.014075
+cache_write = 0.175937

From 0578dc285fd5a2d1d929b3a698e1c406e77c455d Mon Sep 17 00:00:00 2001
From: Pessimist228 <artemburej5@gmail.com>
Date: Mon, 21 Sep 2026 17:24:50 +0200
Subject: [PATCH 1780/2012] llmtech: rename model to nvidia/Qwen3.8-27B-NVFP4
 (#7549)

* llmtech: canonical model id is nvidia/Qwen3.8-27B-NVFP4

The served build is NVIDIA's own NVFP4 checkpoint, not the community one.
The previous id stays accepted by the API as a synonym, so nothing breaks
for anyone who has not switched yet.

* llmtech: drop the old unsloth/ model path
---
 .../llmtech/models/{unsloth => nvidia}/Qwen3.8-27B-NVFP4.toml     | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename providers/llmtech/models/{unsloth => nvidia}/Qwen3.8-27B-NVFP4.toml (100%)

diff --git a/providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml b/providers/llmtech/models/nvidia/Qwen3.8-27B-NVFP4.toml
similarity index 100%
rename from providers/llmtech/models/unsloth/Qwen3.8-27B-NVFP4.toml
rename to providers/llmtech/models/nvidia/Qwen3.8-27B-NVFP4.toml

From ba1ee716d838a37b985a48b6805e7237e88fe755 Mon Sep 17 00:00:00 2001
From: "C.C." <fanweixiao@gmail.com>
Date: Mon, 21 Sep 2026 23:25:03 +0800
Subject: [PATCH 1781/2012] feat(vivgrid): add viv-fast and deepseek-v4.1-flash
 models (#7626)

- viv-fast: new Vivgrid first-party SLM (1M context, 256K output,
  text+image input, reasoning effort low/high/max), lab entry under
  models/vivgrid/ plus provider pricing at $0.13/$0.40 with $0.05
  cache read
- deepseek-v4.1-flash: base_model override on deepseek/deepseek-v4.1-flash
  with Vivgrid pricing ($0.31/$1.23, $0.01 cache read), DeepSeek-native
  reasoning controls (thinking toggle + effort low/high/max) and
  reasoning_content interleaved, matching the existing deepseek-v4-flash
  entry on the same surface
---
 models/vivgrid/viv-fast.toml                  | 26 +++++++++++++++++++
 .../vivgrid/models/deepseek-v4.1-flash.toml   | 20 ++++++++++++++
 providers/vivgrid/models/viv-fast.toml        | 10 +++++++
 3 files changed, 56 insertions(+)
 create mode 100644 models/vivgrid/viv-fast.toml
 create mode 100644 providers/vivgrid/models/deepseek-v4.1-flash.toml
 create mode 100644 providers/vivgrid/models/viv-fast.toml

diff --git a/models/vivgrid/viv-fast.toml b/models/vivgrid/viv-fast.toml
new file mode 100644
index 00000000000..9bed45cd435
--- /dev/null
+++ b/models/vivgrid/viv-fast.toml
@@ -0,0 +1,26 @@
+# Sources:
+# - https://docs.vivgrid.com/models/viv-fast
+# - https://docs.vivgrid.com/api/model-api
+name = "Viv Fast"
+description = "Fast coding model"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+reasoning = true
+knowledge = "2026-09"
+temperature = false
+open_weights = false
+tool_call = true
+structured_output = true
+attachment = true
+
+[limit]
+context = 1_000_000
+output = 256_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[links]]
+label = "Model docs"
+url = "https://docs.vivgrid.com/models/viv-fast"
diff --git a/providers/vivgrid/models/deepseek-v4.1-flash.toml b/providers/vivgrid/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..276f6360406
--- /dev/null
+++ b/providers/vivgrid/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,20 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/deepseek-v4.1-flash (accessed 2026-09-21)
+base_model = "deepseek/deepseek-v4.1-flash"
+
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.31
+output = 1.23
+cache_read = 0.01
diff --git a/providers/vivgrid/models/viv-fast.toml b/providers/vivgrid/models/viv-fast.toml
new file mode 100644
index 00000000000..354c69bf873
--- /dev/null
+++ b/providers/vivgrid/models/viv-fast.toml
@@ -0,0 +1,10 @@
+# Pricing, context window and max output:
+# https://docs.vivgrid.com/models/viv-fast (accessed 2026-09-21)
+base_model = "vivgrid/viv-fast"
+# Effort: reasoning_effort = low|high|max
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[cost]
+input = 0.13
+output = 0.40
+cache_read = 0.05

From 5d395d498aa4352ae6ee6ad40da45cfd8049ba1e Mon Sep 17 00:00:00 2001
From: bukowskiiiiiiiiiiiii <ruijiazhang76@gmail.com>
Date: Mon, 21 Sep 2026 23:25:52 +0800
Subject: [PATCH 1782/2012] feat(stepfun-step-plan): add step-5-preview (#7544)

StepFun's new flagship base model, available on the Step Plan gateway
(api.stepfun.com/step_plan/v1). Lab metadata: 1M-token context/input/
output, text/image/video input, reasoning + tool calling + structured
output, closed weights. Provider override: reasoning_effort accepts
low/medium/high; reasoning side channel is reasoning_content.

Sources:
- https://platform.stepfun.com/docs/zh/guides/models/step-5-preview
- https://api.stepfun.com/step_plan/v1/models (accessed 2026-09-20)

bun validate passes.

Co-authored-by: BUKOWSKIREAL <BUKOWSKIREAL@users.noreply.github.com>
---
 models/stepfun/step-5-preview.toml            | 20 +++++++++++++++++++
 .../models/step-5-preview.toml                | 11 ++++++++++
 providers/stepfun-step-plan/provider.toml     |  2 +-
 3 files changed, 32 insertions(+), 1 deletion(-)
 create mode 100644 models/stepfun/step-5-preview.toml
 create mode 100644 providers/stepfun-step-plan/models/step-5-preview.toml

diff --git a/models/stepfun/step-5-preview.toml b/models/stepfun/step-5-preview.toml
new file mode 100644
index 00000000000..e4718be0938
--- /dev/null
+++ b/models/stepfun/step-5-preview.toml
@@ -0,0 +1,20 @@
+name = "Step 5 Preview"
+description = "StepFun's next-generation flagship base model for coding and professional knowledge work, with native text, image, and video input and a 1M-token context window"
+# release_date is StepFun's API catalog first-appearance timestamp (created=1789564127,
+# 2026-09-16 13:08 UTC) — no dated launch announcement was found on the model docs page.
+release_date = "2026-09-16"
+last_updated = "2026-09-20"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 1_000_000
+input = 1_000_000
+output = 1_000_000
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/stepfun-step-plan/models/step-5-preview.toml b/providers/stepfun-step-plan/models/step-5-preview.toml
new file mode 100644
index 00000000000..684d9f67224
--- /dev/null
+++ b/providers/stepfun-step-plan/models/step-5-preview.toml
@@ -0,0 +1,11 @@
+# Chat `reasoning_effort` accepts low/medium/high (accessed 2026-09-20).
+# https://platform.stepfun.com/docs/zh/guides/models/step-5-preview
+# Step Plan is subscription/credit based with no public per-token price, so no [cost].
+base_model = "stepfun/step-5-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/stepfun-step-plan/provider.toml b/providers/stepfun-step-plan/provider.toml
index ebcf390c442..1c0546449a3 100644
--- a/providers/stepfun-step-plan/provider.toml
+++ b/providers/stepfun-step-plan/provider.toml
@@ -4,7 +4,7 @@ npm = "@ai-sdk/openai-compatible"
 # Reasoning HTTP format (accessed 2026-06-25):
 # Step Plan exposes POST /step_plan/v1/chat/completions with top-level
 # `reasoning_effort` and POST /step_plan/v1/messages with
-# `output_config.effort`. step-3.7-flash accepts low/medium/high;
+# `output_config.effort`. step-3.7-flash and step-5-preview accept low/medium/high;
 # step-3.5-flash and step-3.5-flash-2603 accept low/high. No plan Responses
 # endpoint is listed.
 # Source:

From 9b23514872a602c5edbd2ff05de796f636b2e59d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 15:26:20 +0000
Subject: [PATCH 1783/2012] chore(sync): update Kilo model catalog (#7637)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..a92bf5149a1 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.57816
+output = 1.73448
+cache_read = 0.019272
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From c49c7b1f831ce90e0ca964206a11fca47034988b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 15:26:41 +0000
Subject: [PATCH 1784/2012] chore(sync): update OpenRouter model catalog
 (#7635)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 4 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index b044540752e..f9ab26fd006 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.05698
-output = 0.11396
-cache_read = 0.011396
+input = 0.05544
+output = 0.11088
+cache_read = 0.011088
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c6c87510247..32fc69b9875 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.57816
+output = 1.73448
+cache_read = 0.019272
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ae6a4dbd88c..b7feb0289f9 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.948126
-output = 1.896252
-cache_read = 0.079011
+input = 0.942906
+output = 1.885812
+cache_read = 0.078576
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..9b9fc44533c 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.57816
+output = 1.73448
+cache_read = 0.019272
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From 1e4a5a889fb69680b00a315afa65e0518a3977f1 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:26:47 +0800
Subject: [PATCH 1785/2012] feat(aihubmix): add longcat-2.0 (#7557)

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
---
 providers/aihubmix/models/longcat-2.0.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/aihubmix/models/longcat-2.0.toml

diff --git a/providers/aihubmix/models/longcat-2.0.toml b/providers/aihubmix/models/longcat-2.0.toml
new file mode 100644
index 00000000000..4f0b9ae311d
--- /dev/null
+++ b/providers/aihubmix/models/longcat-2.0.toml
@@ -0,0 +1,17 @@
+# Sources (accessed 2026-09-20):
+# https://aihubmix.com/api/v1/models?type=llm (USD per million tokens)
+# https://github.com/meituan-longcat/LongCat-2.0
+# Toggle: enable_thinking = true|false on /v1/chat/completions.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "meituan/longcat-2.0"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.7746
+output = 3.0984
+cache_read = 0.015492

From 12465a3b1719412cc202a20b1d02a5f9f61672af Mon Sep 17 00:00:00 2001
From: NoahChen <noahchen2002@gmail.com>
Date: Mon, 21 Sep 2026 23:26:55 +0800
Subject: [PATCH 1786/2012] feat(siliconflow): sync model catalog with
 SiliconFlow international site (#7625)

* feat(siliconflow): sync catalog with SiliconFlow international site (2026-09)

Add 13 models: DeepSeek-V4-Flash-Vision-Exp, DeepSeek-V4-Pro-0813,
DeepSeek-V4-Flash-0731, Qwen3.8-2.4T-A95B, GLM-5.3, GLM-5.3-Flash,
Kimi-K3, Kimi-K2.7-Code, MiniMax-M3, gemma-4-12B-it, Nex-N2-Pro,
meituan-longcat/LongCat-2.0, tencent/Hy3 (replaces Hy3-preview).

Remove 5 models no longer listed on SiliconFlow international:
ERNIE-4.5-300B-A47B, Qwen3-VL-235B-A22B-Instruct/Thinking,
Qwen3-235B-A22B-Thinking-2507, Hy3-preview.

Update pricing to current rates for DeepSeek-V3.2, DeepSeek-V4-Flash,
DeepSeek-V4-Pro, Kimi-K2.6, GLM-5.1, GLM-5.2.

Add lab metadata for google/gemma-4-12b-it and nex-agi/nex-n2-pro.

Sources: https://www.siliconflow.com/models,
https://www.siliconflow.com/pricing,
https://docs.siliconflow.com/en/release-notes/overview.md

* fix(siliconflow): address auto-review items on reasoning options

- LongCat-2.0: drop toggle (SF does not enumerate enable_thinking for
  LongCat), keep budget_tokens with leading wire-path comment
- MiniMax-M3, Qwen3.8-2.4T-A95B: add leading wire-path comments citing
  SF's thinking_budget surface; SF exposes no reasoning_effort param
- Hy3: drop stale open_weights=false override inherited from Hy3-preview
  (lab entry tencent/hy3 is open weights)
- nex-agi/nex-n2-pro lab: attachment=true, consistent with image input
  modality

* fix(siliconflow): address round-2 auto-review items

- Narrow inherited modalities to SF's actual surface: MiniMax-M3,
  Kimi-K3, Kimi-K2.7-Code, GLM-5.3-Flash now serve text+image only
  (per each model page's "Support image input: Yes"; no video/pdf on SF)
- Hy3: drop redundant [modalities] (identical to lab), document that
  262K/262K limits are SF-advertised deltas vs lab 256K/128K
- GLM-5.1: cite exact price row (input 1.19 / cached 0.6 / output 3.74
  from the GLM-5.1 model page) - cache_read=0.6 is SF's published price
- V4-Pro / K2.7-Code / GLM-5.2: note costs are published USD list
  prices from siliconflow.com/pricing, not FX conversions

* fix(siliconflow): address round-3 auto-review items

- GLM-5.3 / GLM-5.3-Flash: effort low|high|max, matching the first-party
  zhipuai GLM-5.3 entry (5.2's high|max was generation-specific)
- gemma-4-12B-it: drop [limit].context (identical to lab), keep only the
  real output delta

---------

Co-authored-by: NoahChen <288864333+noahchen2002@users.noreply.github.com>
---
 models/google/gemma-4-12b-it.toml             | 23 ++++++++++++++++
 models/nex-agi/nex-n2-pro.toml                | 23 ++++++++++++++++
 .../models/MiniMaxAI/MiniMax-M3.toml          | 26 +++++++++++++++++++
 .../Qwen/Qwen3-235B-A22B-Thinking-2507.toml   | 24 -----------------
 .../Qwen/Qwen3-VL-235B-A22B-Instruct.toml     | 23 ----------------
 .../Qwen/Qwen3-VL-235B-A22B-Thinking.toml     | 24 -----------------
 .../models/Qwen/Qwen3.8-2.4T-A95B.toml        | 16 ++++++++++++
 .../models/baidu/ERNIE-4.5-300B-A47B.toml     | 23 ----------------
 .../models/deepseek-ai/DeepSeek-V3.2.toml     |  1 +
 .../deepseek-ai/DeepSeek-V4-Flash-0731.toml   | 12 +++++++++
 .../DeepSeek-V4-Flash-Vision-Exp.toml         | 12 +++++++++
 .../models/deepseek-ai/DeepSeek-V4-Flash.toml |  2 +-
 .../deepseek-ai/DeepSeek-V4-Pro-0813.toml     | 12 +++++++++
 .../models/deepseek-ai/DeepSeek-V4-Pro.toml   |  8 +++---
 .../models/google/gemma-4-12B-it.toml         | 16 ++++++++++++
 .../models/meituan-longcat/LongCat-2.0.toml   | 18 +++++++++++++
 .../models/moonshotai/Kimi-K2.6.toml          |  4 +--
 .../models/moonshotai/Kimi-K2.7-Code.toml     | 21 +++++++++++++++
 .../models/moonshotai/Kimi-K3.toml            | 22 ++++++++++++++++
 .../models/nex-agi/Nex-N2-Pro.toml            | 15 +++++++++++
 .../models/tencent/Hy3-preview.toml           | 18 -------------
 providers/siliconflow/models/tencent/Hy3.toml | 15 +++++++++++
 .../siliconflow/models/zai-org/GLM-5.1.toml   | 11 +++++---
 .../siliconflow/models/zai-org/GLM-5.2.toml   |  7 +++--
 .../models/zai-org/GLM-5.3-Flash.toml         | 24 +++++++++++++++++
 .../siliconflow/models/zai-org/GLM-5.3.toml   | 16 ++++++++++++
 26 files changed, 292 insertions(+), 124 deletions(-)
 create mode 100644 models/google/gemma-4-12b-it.toml
 create mode 100644 models/nex-agi/nex-n2-pro.toml
 create mode 100644 providers/siliconflow/models/MiniMaxAI/MiniMax-M3.toml
 delete mode 100644 providers/siliconflow/models/Qwen/Qwen3-235B-A22B-Thinking-2507.toml
 delete mode 100644 providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Instruct.toml
 delete mode 100644 providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Thinking.toml
 create mode 100644 providers/siliconflow/models/Qwen/Qwen3.8-2.4T-A95B.toml
 delete mode 100644 providers/siliconflow/models/baidu/ERNIE-4.5-300B-A47B.toml
 create mode 100644 providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
 create mode 100644 providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
 create mode 100644 providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
 create mode 100644 providers/siliconflow/models/google/gemma-4-12B-it.toml
 create mode 100644 providers/siliconflow/models/meituan-longcat/LongCat-2.0.toml
 create mode 100644 providers/siliconflow/models/moonshotai/Kimi-K2.7-Code.toml
 create mode 100644 providers/siliconflow/models/moonshotai/Kimi-K3.toml
 create mode 100644 providers/siliconflow/models/nex-agi/Nex-N2-Pro.toml
 delete mode 100644 providers/siliconflow/models/tencent/Hy3-preview.toml
 create mode 100644 providers/siliconflow/models/tencent/Hy3.toml
 create mode 100644 providers/siliconflow/models/zai-org/GLM-5.3-Flash.toml
 create mode 100644 providers/siliconflow/models/zai-org/GLM-5.3.toml

diff --git a/models/google/gemma-4-12b-it.toml b/models/google/gemma-4-12b-it.toml
new file mode 100644
index 00000000000..af2b5f0d680
--- /dev/null
+++ b/models/google/gemma-4-12b-it.toml
@@ -0,0 +1,23 @@
+name = "Gemma 4 12B IT"
+description = "Compact Gemma 4 instruction model for open, self-hosted chat and reasoning"
+family = "gemma"
+release_date = "2026-06-09"
+last_updated = "2026-06-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-4-12B-it"
diff --git a/models/nex-agi/nex-n2-pro.toml b/models/nex-agi/nex-n2-pro.toml
new file mode 100644
index 00000000000..3b134613e5f
--- /dev/null
+++ b/models/nex-agi/nex-n2-pro.toml
@@ -0,0 +1,23 @@
+# https://huggingface.co/nex-agi/Nex-N2-Pro
+name = "Nex-N2-Pro"
+description = "Open agentic MoE model (397B total, 17B active) for coding, tool use, and research workflows"
+release_date = "2026-06-02"
+last_updated = "2026-06-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/nex-agi/Nex-N2-Pro"
diff --git a/providers/siliconflow/models/MiniMaxAI/MiniMax-M3.toml b/providers/siliconflow/models/MiniMaxAI/MiniMax-M3.toml
new file mode 100644
index 00000000000..748a1991b69
--- /dev/null
+++ b/providers/siliconflow/models/MiniMaxAI/MiniMax-M3.toml
@@ -0,0 +1,26 @@
+# Budget: thinking_budget (integer reasoning tokens, 128..32768) — SiliconFlow documents
+# thinking_budget for all reasoning models; enable_thinking is not enumerated for MiniMax
+# models on this host, so no toggle.
+# Sources: https://www.siliconflow.com/models (2026-09-08),
+# https://docs.siliconflow.com/en/api-reference/chat-completions/chat-completions
+base_model = "minimax/MiniMax-M3"
+
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+output = 131_000
+
+# SiliconFlow serves image input but not video for this model
+# (model page "Support image input: Yes"): https://www.siliconflow.com/models/minimax-m3
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/siliconflow/models/Qwen/Qwen3-235B-A22B-Thinking-2507.toml b/providers/siliconflow/models/Qwen/Qwen3-235B-A22B-Thinking-2507.toml
deleted file mode 100644
index fb446e7d261..00000000000
--- a/providers/siliconflow/models/Qwen/Qwen3-235B-A22B-Thinking-2507.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Qwen/Qwen3-235B-A22B-Thinking-2507"
-description = "Qwen reasoning model for deliberate problem solving, math, and coding"
-family = "qwen"
-release_date = "2025-07-28"
-last_updated = "2025-11-25"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.13
-output = 0.6
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Instruct.toml b/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Instruct.toml
deleted file mode 100644
index 8c74b4bceca..00000000000
--- a/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Instruct.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "Qwen/Qwen3-VL-235B-A22B-Instruct"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-10-04"
-last_updated = "2025-11-25"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.3
-output = 1.5
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Thinking.toml b/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Thinking.toml
deleted file mode 100644
index 03bb1766276..00000000000
--- a/providers/siliconflow/models/Qwen/Qwen3-VL-235B-A22B-Thinking.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Qwen/Qwen3-VL-235B-A22B-Thinking"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-10-04"
-last_updated = "2025-11-25"
-attachment = true
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.45
-output = 3.5
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/siliconflow/models/Qwen/Qwen3.8-2.4T-A95B.toml b/providers/siliconflow/models/Qwen/Qwen3.8-2.4T-A95B.toml
new file mode 100644
index 00000000000..86601f6baa1
--- /dev/null
+++ b/providers/siliconflow/models/Qwen/Qwen3.8-2.4T-A95B.toml
@@ -0,0 +1,16 @@
+# Budget: thinking_budget (integer reasoning tokens, 128..32768) — SiliconFlow exposes no
+# reasoning_effort parameter; thinking_budget is documented for all reasoning models.
+# Sources: https://www.siliconflow.com/models (2026-09-08),
+# https://docs.siliconflow.com/en/api-reference/chat-completions/chat-completions
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[cost]
+input = 2.0
+output = 6.0
+cache_read = 0.25
+
+[limit]
+context = 1_049_000
+output = 131_000
diff --git a/providers/siliconflow/models/baidu/ERNIE-4.5-300B-A47B.toml b/providers/siliconflow/models/baidu/ERNIE-4.5-300B-A47B.toml
deleted file mode 100644
index 5309d4c0402..00000000000
--- a/providers/siliconflow/models/baidu/ERNIE-4.5-300B-A47B.toml
+++ /dev/null
@@ -1,23 +0,0 @@
-name = "baidu/ERNIE-4.5-300B-A47B"
-description = "Tool-capable chat model for instruction following and agentic application workflows"
-family = "ernie"
-release_date = "2025-07-02"
-last_updated = "2025-11-25"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.28
-output = 1.1
-
-[limit]
-context = 131_000
-output = 131_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
\ No newline at end of file
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V3.2.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V3.2.toml
index a05b736c528..84bd5c7b342 100644
--- a/providers/siliconflow/models/deepseek-ai/DeepSeek-V3.2.toml
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V3.2.toml
@@ -14,6 +14,7 @@ open_weights = false
 [cost]
 input = 0.27
 output = 0.42
+cache_read = 0.135
 
 [limit]
 context = 164_000
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
new file mode 100644
index 00000000000..be8c8bf0744
--- /dev/null
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml
@@ -0,0 +1,12 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.22
+output = 0.66
+cache_read = 0.014
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
new file mode 100644
index 00000000000..674c8102988
--- /dev/null
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp.toml
@@ -0,0 +1,12 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "deepseek/deepseek-v4-flash-vision-exp"
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.44
+output = 1.32
+cache_read = 0.028
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash.toml
index 744b022d3f0..14ac45494a9 100644
--- a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash.toml
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Flash.toml
@@ -6,6 +6,6 @@ reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
 field = "reasoning_content"
 
 [cost]
-input = 0.14
+input = 0.13
 output = 0.28
 cache_read = 0.028
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
new file mode 100644
index 00000000000..dd2c092e9dc
--- /dev/null
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml
@@ -0,0 +1,12 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "deepseek/deepseek-v4-pro-0813"
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.32
+output = 3.96
+cache_read = 0.044
diff --git a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro.toml b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro.toml
index a573cf5c2af..9994892360a 100644
--- a/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro.toml
+++ b/providers/siliconflow/models/deepseek-ai/DeepSeek-V4-Pro.toml
@@ -1,3 +1,5 @@
+# Cost values are the published USD list prices on SiliconFlow international
+# (https://www.siliconflow.com/pricing, accessed 2026-09-08), not FX conversions by this PR.
 base_model = "deepseek/deepseek-v4-pro"
 
 reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
@@ -6,6 +8,6 @@ reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
 field = "reasoning_content"
 
 [cost]
-input = 1.74
-output = 3.48
-cache_read = 0.145
+input = 1.50162
+output = 3.135
+cache_read = 0.135
diff --git a/providers/siliconflow/models/google/gemma-4-12B-it.toml b/providers/siliconflow/models/google/gemma-4-12B-it.toml
new file mode 100644
index 00000000000..05329524ad5
--- /dev/null
+++ b/providers/siliconflow/models/google/gemma-4-12B-it.toml
@@ -0,0 +1,16 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "google/gemma-4-12b-it"
+attachment = false
+reasoning = false
+open_weights = false
+
+[cost]
+input = 0.10
+output = 0.30
+
+[limit]
+output = 262_144
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/siliconflow/models/meituan-longcat/LongCat-2.0.toml b/providers/siliconflow/models/meituan-longcat/LongCat-2.0.toml
new file mode 100644
index 00000000000..dfe567a7db0
--- /dev/null
+++ b/providers/siliconflow/models/meituan-longcat/LongCat-2.0.toml
@@ -0,0 +1,18 @@
+# Budget: thinking_budget (integer reasoning tokens, 128..32768) — SiliconFlow documents
+# thinking_budget for all reasoning models and enumerates enable_thinking only for specific
+# models (LongCat not among them), so no toggle on this host.
+# Sources: https://www.siliconflow.com/models (2026-09-08),
+# https://docs.siliconflow.com/en/api-reference/chat-completions/chat-completions
+base_model = "meituan/longcat-2.0"
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.75
+output = 2.95
+cache_read = 0.015
+
+[limit]
+context = 1_049_000
diff --git a/providers/siliconflow/models/moonshotai/Kimi-K2.6.toml b/providers/siliconflow/models/moonshotai/Kimi-K2.6.toml
index 73e5e28af8b..48e6aaf149e 100644
--- a/providers/siliconflow/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/siliconflow/models/moonshotai/Kimi-K2.6.toml
@@ -16,8 +16,8 @@ field = "reasoning_content"
 
 [cost]
 input = 0.77
-output = 4.0
-cache_read = 0.2
+output = 3.4
+cache_read = 0.14
 
 [limit]
 context = 262_000
diff --git a/providers/siliconflow/models/moonshotai/Kimi-K2.7-Code.toml b/providers/siliconflow/models/moonshotai/Kimi-K2.7-Code.toml
new file mode 100644
index 00000000000..f70f197f92d
--- /dev/null
+++ b/providers/siliconflow/models/moonshotai/Kimi-K2.7-Code.toml
@@ -0,0 +1,21 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "moonshotai/kimi-k2.7-code"
+
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.85916
+output = 3.8
+cache_read = 0.17993
+
+# Cost values are the published USD list prices on SiliconFlow international
+# (https://www.siliconflow.com/pricing), not FX conversions by this PR.
+# SiliconFlow serves image input but not video for this model
+# (model page "Support image input: Yes"): https://www.siliconflow.com/models/kimi-k2-7-code
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/siliconflow/models/moonshotai/Kimi-K3.toml b/providers/siliconflow/models/moonshotai/Kimi-K3.toml
new file mode 100644
index 00000000000..daa54bb61f7
--- /dev/null
+++ b/providers/siliconflow/models/moonshotai/Kimi-K3.toml
@@ -0,0 +1,22 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "moonshotai/kimi-k3"
+
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.7
+output = 13.5
+cache_read = 0.27
+
+[limit]
+output = 262_000
+
+# SiliconFlow serves image input but not video for this model
+# (model page "Support image input: Yes"): https://www.siliconflow.com/models/kimi-k3
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/siliconflow/models/nex-agi/Nex-N2-Pro.toml b/providers/siliconflow/models/nex-agi/Nex-N2-Pro.toml
new file mode 100644
index 00000000000..f160fdf834d
--- /dev/null
+++ b/providers/siliconflow/models/nex-agi/Nex-N2-Pro.toml
@@ -0,0 +1,15 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "nex-agi/nex-n2-pro"
+
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.5
+output = 2.5
+cache_read = 0.25
+
+[limit]
+output = 256_000
diff --git a/providers/siliconflow/models/tencent/Hy3-preview.toml b/providers/siliconflow/models/tencent/Hy3-preview.toml
deleted file mode 100644
index fddc3766f5a..00000000000
--- a/providers/siliconflow/models/tencent/Hy3-preview.toml
+++ /dev/null
@@ -1,18 +0,0 @@
-base_model = "tencent/hy3-preview"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
-open_weights = false
-
-[cost]
-input = 0.066
-cache_read = 0.029
-output = 0.26
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/siliconflow/models/tencent/Hy3.toml b/providers/siliconflow/models/tencent/Hy3.toml
new file mode 100644
index 00000000000..f304cbd9060
--- /dev/null
+++ b/providers/siliconflow/models/tencent/Hy3.toml
@@ -0,0 +1,15 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+# Limits below are this host's real deltas vs the lab entry (256K/128K):
+# SiliconFlow advertises 262K total context and 262K max output
+# (https://www.siliconflow.com/models/hy3).
+base_model = "tencent/hy3"
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
+
+[cost]
+input = 0.132
+cache_read = 0.033
+output = 0.528
+
+[limit]
+context = 262_144
+output = 262_144
diff --git a/providers/siliconflow/models/zai-org/GLM-5.1.toml b/providers/siliconflow/models/zai-org/GLM-5.1.toml
index 5a341072c0d..248e165981b 100644
--- a/providers/siliconflow/models/zai-org/GLM-5.1.toml
+++ b/providers/siliconflow/models/zai-org/GLM-5.1.toml
@@ -1,3 +1,6 @@
+# Prices verified against SiliconFlow international, GLM-5.1 model page
+# (https://www.siliconflow.com/models/glm-5-1, accessed 2026-09-08):
+# input $1.19, cached input $0.6, output $3.74 per M tokens.
 name = "zai-org/GLM-5.1"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
 family = "glm"
@@ -15,14 +18,14 @@ open_weights = true
 field = "reasoning_content"
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.19
+output = 3.74
+cache_read = 0.6
 cache_write = 0
 
 [limit]
 context = 205_000
-output = 205_000
+output = 131_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/siliconflow/models/zai-org/GLM-5.2.toml b/providers/siliconflow/models/zai-org/GLM-5.2.toml
index 1a6b9e26746..4f4b4d198e8 100644
--- a/providers/siliconflow/models/zai-org/GLM-5.2.toml
+++ b/providers/siliconflow/models/zai-org/GLM-5.2.toml
@@ -1,3 +1,6 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+# Cost values are the published USD list prices on SiliconFlow international
+# (https://www.siliconflow.com/pricing, accessed 2026-09-08), not FX conversions by this PR.
 base_model = "zhipuai/glm-5.2"
 reasoning_options = [{ type = "effort", values = ["high", "max"] }]
 
@@ -5,8 +8,8 @@ reasoning_options = [{ type = "effort", values = ["high", "max"] }]
 field = "reasoning_content"
 
 [cost]
-input = 1.4
-output = 4.4
+input = 1.302
+output = 4.092
 cache_read = 0.26
 cache_write = 0
 
diff --git a/providers/siliconflow/models/zai-org/GLM-5.3-Flash.toml b/providers/siliconflow/models/zai-org/GLM-5.3-Flash.toml
new file mode 100644
index 00000000000..afb2e9853b8
--- /dev/null
+++ b/providers/siliconflow/models/zai-org/GLM-5.3-Flash.toml
@@ -0,0 +1,24 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "zhipuai/glm-5.3-flash"
+
+
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+cache_write = 0
+
+[limit]
+context = 1_049_000
+output = 262_000
+
+# SiliconFlow serves image input but not video/pdf for this model
+# (model page "Support image input: Yes"): https://www.siliconflow.com/models/glm-5-3-flash
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/siliconflow/models/zai-org/GLM-5.3.toml b/providers/siliconflow/models/zai-org/GLM-5.3.toml
new file mode 100644
index 00000000000..3de6268387c
--- /dev/null
+++ b/providers/siliconflow/models/zai-org/GLM-5.3.toml
@@ -0,0 +1,16 @@
+# Sources: https://www.siliconflow.com/models (2026-09-08)
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+cache_write = 0
+
+[limit]
+context = 1_049_000
+output = 262_000

From e64a2898d35f1f4b219fbd04b9d3017764e0298e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 15:27:31 +0000
Subject: [PATCH 1787/2012] chore(sync): update Eden AI model catalog (#7636)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/deepseek/deepseek-v4-pro.toml         | 2 +-
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 7 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/edenai/models/deepseek/deepseek-v4-pro.toml b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
index 043e1bb1f41..05723a127e4 100644
--- a/providers/edenai/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
@@ -2,7 +2,7 @@ base_model = "deepseek/deepseek-v4-pro"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "high", "max"]
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.66
diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index 9b2ca243154..beffb59e2ed 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.3438
-output = 0.4584
+input = 0.3447
+output = 0.4596
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index b5b5ca537eb..11dee7e0f74 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.7449
-output = 0.7449
+input = 0.74685
+output = 0.74685
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index fb25f513d77..2bf7b7f0f21 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.1719
-output = 0.7449
+input = 0.17235
+output = 0.74685
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index ed1db7ea77d..d761571f13a 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4584
-output = 0.9168
+input = 0.4596
+output = 0.9192
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index df8da020b1b..dbff07167cc 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.1719
-output = 0.6876
+input = 0.17235
+output = 0.6894
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index da84cd257a7..8d733529a6c 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.0314
-output = 1.0314
+input = 1.0341
+output = 1.0341

From 66635ce3d1efdad42d03046d1b58685956bc793d Mon Sep 17 00:00:00 2001
From: Kali Norby <knorby@users.noreply.github.com>
Date: Mon, 21 Sep 2026 08:30:37 -0700
Subject: [PATCH 1788/2012] Update neuralwatt catalog: add 8 models, deprecate
 retired GLM-5.2 family and V4 Pro (#7482)

---
 .../models/deepseek-v4-flash-flex.toml        |  5 +--
 .../models/deepseek-v4-flash-speed.toml       | 26 ++++++++++++++
 .../neuralwatt/models/deepseek-v4-flash.toml  |  4 ++-
 .../neuralwatt/models/deepseek-v4-pro.toml    | 11 +++---
 .../models/deepseek-v4.1-flash-flex.toml      | 26 ++++++++++++++
 .../models/deepseek-v4.1-flash.toml           | 25 +++++++++++++
 providers/neuralwatt/models/glm-5.2-fast.toml |  3 ++
 providers/neuralwatt/models/glm-5.2-flex.toml |  3 ++
 .../models/glm-5.2-short-fast-flex.toml       |  3 ++
 .../neuralwatt/models/glm-5.2-short-fast.toml |  3 ++
 .../neuralwatt/models/glm-5.2-short-flex.toml |  3 ++
 .../neuralwatt/models/glm-5.2-short.toml      |  3 ++
 providers/neuralwatt/models/glm-5.2.toml      |  3 ++
 .../neuralwatt/models/glm-5.3-flash-flex.toml | 34 ++++++++++++++++++
 .../neuralwatt/models/glm-5.3-flash.toml      | 35 +++++++++++++++++++
 providers/neuralwatt/models/glm-5.3-flex.toml | 30 ++++++++++++++++
 providers/neuralwatt/models/glm-5.3.toml      |  8 ++---
 .../neuralwatt/models/qwen-3.8-27b-flex.toml  | 33 +++++++++++++++++
 providers/neuralwatt/models/qwen-3.8-27b.toml | 11 +++---
 .../neuralwatt/models/qwen3.6-35b-flex.toml   | 32 +++++++++++++++++
 20 files changed, 278 insertions(+), 23 deletions(-)
 create mode 100644 providers/neuralwatt/models/deepseek-v4-flash-speed.toml
 create mode 100644 providers/neuralwatt/models/deepseek-v4.1-flash-flex.toml
 create mode 100644 providers/neuralwatt/models/deepseek-v4.1-flash.toml
 create mode 100644 providers/neuralwatt/models/glm-5.3-flash-flex.toml
 create mode 100644 providers/neuralwatt/models/glm-5.3-flash.toml
 create mode 100644 providers/neuralwatt/models/glm-5.3-flex.toml
 create mode 100644 providers/neuralwatt/models/qwen-3.8-27b-flex.toml
 create mode 100644 providers/neuralwatt/models/qwen3.6-35b-flex.toml

diff --git a/providers/neuralwatt/models/deepseek-v4-flash-flex.toml b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
index 090d4d2f3f7..fae8619652d 100644
--- a/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
+++ b/providers/neuralwatt/models/deepseek-v4-flash-flex.toml
@@ -1,5 +1,6 @@
 # Flex-tier variant of deepseek-v4-flash: same model, context window
-# (1,048,560), and output cap (65,536) as the standard tier; requested via the
+# (1,048,560), and output cap (393,216, raised from 65,536 per GET /v1/models
+# checked 2026-09-18) as the standard tier; requested via the
 # `-flex` model ID or service_tier="flex" and requires stream=true
 # (non-streaming requests fall through to the standard tier).
 # Cost is the flex rate: 0.65x of standard pricing (35% off) per the flex-tier
@@ -26,4 +27,4 @@ cache_read = 0.0182
 
 [limit]
 context = 1_048_560
-output = 65_536
+output = 393_216
diff --git a/providers/neuralwatt/models/deepseek-v4-flash-speed.toml b/providers/neuralwatt/models/deepseek-v4-flash-speed.toml
new file mode 100644
index 00000000000..adf0fbd952c
--- /dev/null
+++ b/providers/neuralwatt/models/deepseek-v4-flash-speed.toml
@@ -0,0 +1,26 @@
+# Speed variant serving the DeepSeek V4 Flash 0731 checkpoint: same pricing
+# as deepseek-v4-flash ($0.14/$0.28/$0.028) with a 393,216-token output cap,
+# per the public GET /v1/models metadata (checked 2026-09-18). Text-only
+# input on this host. Reasoning matches the standard tier: reasoning_effort
+# = none|high|max (default none, reasoning off by default; aliases
+# low/minimal/medium -> high, xhigh -> max). Off is effort=none, so no
+# separate toggle. thinking_token_budget is rejected with a 400 on this
+# generation, so it is not declared.
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/docs/api/models
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash (Speed)"
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high", "max"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.028
+
+[limit]
+context = 1_048_560
+output = 393_216
diff --git a/providers/neuralwatt/models/deepseek-v4-flash.toml b/providers/neuralwatt/models/deepseek-v4-flash.toml
index 5d30b1cbdca..4a32afaee6c 100644
--- a/providers/neuralwatt/models/deepseek-v4-flash.toml
+++ b/providers/neuralwatt/models/deepseek-v4-flash.toml
@@ -8,6 +8,8 @@
 # is rejected with a 400 on this model, so it is not declared.
 # https://portal.neuralwatt.com/docs/api/models
 # https://portal.neuralwatt.com/pricing
+# Output cap raised to 393,216 tokens per GET /v1/models (checked
+# 2026-09-18; was 65,536).
 # https://portal.neuralwatt.com/docs/api/chat-completions
 base_model = "deepseek/deepseek-v4-flash"
 interleaved = true
@@ -23,4 +25,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_560
-output = 65_536
+output = 393_216
diff --git a/providers/neuralwatt/models/deepseek-v4-pro.toml b/providers/neuralwatt/models/deepseek-v4-pro.toml
index 7869e9bb888..7b3f48e037a 100644
--- a/providers/neuralwatt/models/deepseek-v4-pro.toml
+++ b/providers/neuralwatt/models/deepseek-v4-pro.toml
@@ -1,8 +1,6 @@
-# Preview rollout: deepseek-v4-pro is in private preview on Neuralwatt —
-# visible to paid users after requesting access per model card, per the
-# Preview Models guide. Pricing, output cap, and cache-read from the
-# authenticated GET /v1/models metadata (checked 2026-08-27); V4-Pro is not
-# yet in the public catalog or token-pricing table.
+# Deprecated: Neuralwatt retired deepseek-v4-pro from GET /v1/models
+# (checked 2026-09-18); it never left private preview. Retained for pricing
+# history.
 # Reasoning per the chat-completions per-model table (docs verified against
 # the live API on 2026-08-26): reasoning_effort accepts none|low|high|max as
 # distinct levels with low as the default (reasoning on by default); aliases
@@ -10,9 +8,8 @@
 # accepted on this model (unlike deepseek-v4-flash, which rejects it); no
 # bounds or disable sentinel documented.
 # https://portal.neuralwatt.com/docs/api/chat-completions
-# https://portal.neuralwatt.com/docs/guides/preview-models
 base_model = "deepseek/deepseek-v4-pro"
-status = "beta"
+status = "deprecated"
 interleaved = true
 
 # Effort: reasoning_effort = none|low|high|max (off is effort=none; aliases
diff --git a/providers/neuralwatt/models/deepseek-v4.1-flash-flex.toml b/providers/neuralwatt/models/deepseek-v4.1-flash-flex.toml
new file mode 100644
index 00000000000..a3dcc0edcd1
--- /dev/null
+++ b/providers/neuralwatt/models/deepseek-v4.1-flash-flex.toml
@@ -0,0 +1,26 @@
+# Flex-tier variant of deepseek-v4.1-flash: same model, context (1,048,560),
+# vision, and effort ladder as the standard tier; requested via the `-flex`
+# model ID or service_tier="flex" and requires stream=true (non-streaming
+# requests fall through to the standard tier). Cost is the flex rate: 0.65x
+# of standard pricing (35% off) per the flex-tier guide, applied to the
+# standard rates advertised by GET /v1/models ($0.15/$0.6/$0.015, checked
+# 2026-09-18). thinking_token_budget is rejected with a 400 on the V4-Flash
+# generation and is not documented for V4.1-Flash, so it is not declared.
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash Flex"
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "xhigh", "max"]
+
+[cost]
+input = 0.0975
+output = 0.39
+cache_read = 0.00975
+
+[limit]
+context = 1_048_560
+output = 393_216
diff --git a/providers/neuralwatt/models/deepseek-v4.1-flash.toml b/providers/neuralwatt/models/deepseek-v4.1-flash.toml
new file mode 100644
index 00000000000..240f6508802
--- /dev/null
+++ b/providers/neuralwatt/models/deepseek-v4.1-flash.toml
@@ -0,0 +1,25 @@
+# Pricing, limits, vision, and effort ladder from the public GET /v1/models
+# metadata (checked 2026-09-18): $0.15/$0.6/$0.015, 1,048,560-token context,
+# 393,216-token output cap, 20-image cap. Effort: reasoning_effort =
+# none|low|high|xhigh|max (default high, reasoning on by default; aliases
+# medium -> high, minimal -> low). Off is effort=none, so no separate toggle.
+# thinking_token_budget is rejected with a 400 on the V4-Flash generation
+# (checked 2026-08-26) and is not documented for V4.1-Flash, so it is not
+# declared.
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/docs/api/models
+base_model = "deepseek/deepseek-v4.1-flash"
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.015
+
+[limit]
+context = 1_048_560
+output = 393_216
diff --git a/providers/neuralwatt/models/glm-5.2-fast.toml b/providers/neuralwatt/models/glm-5.2-fast.toml
index a61cf125702..241dd01adcd 100644
--- a/providers/neuralwatt/models/glm-5.2-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-fast.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Fast"
@@ -5,6 +7,7 @@ description = "Efficient GLM model for fast reasoning, coding, and agent workflo
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2-flex.toml b/providers/neuralwatt/models/glm-5.2-flex.toml
index e42b7fb1cc5..b8fce8e6ba2 100644
--- a/providers/neuralwatt/models/glm-5.2-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-flex.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Flex"
@@ -5,6 +7,7 @@ description = "Flagship GLM model for hybrid reasoning, coding, and agentic engi
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
index eea25d5f58e..f463c21d078 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast-flex.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Fast Flex"
@@ -5,6 +7,7 @@ description = "Efficient GLM model for fast reasoning, coding, and agent workflo
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2-short-fast.toml b/providers/neuralwatt/models/glm-5.2-short-fast.toml
index fa48e039e59..20da2db40c4 100644
--- a/providers/neuralwatt/models/glm-5.2-short-fast.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-fast.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Fast"
@@ -5,6 +7,7 @@ description = "Efficient GLM model for fast reasoning, coding, and agent workflo
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2-short-flex.toml b/providers/neuralwatt/models/glm-5.2-short-flex.toml
index 604df60b11d..0b1cc1ed110 100644
--- a/providers/neuralwatt/models/glm-5.2-short-flex.toml
+++ b/providers/neuralwatt/models/glm-5.2-short-flex.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short Flex"
@@ -5,6 +7,7 @@ description = "Flagship GLM model for hybrid reasoning, coding, and agentic engi
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2-short.toml b/providers/neuralwatt/models/glm-5.2-short.toml
index 26127431e9c..a324bae38c5 100644
--- a/providers/neuralwatt/models/glm-5.2-short.toml
+++ b/providers/neuralwatt/models/glm-5.2-short.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Short"
@@ -5,6 +7,7 @@ description = "Flagship GLM model for hybrid reasoning, coding, and agentic engi
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.2.toml b/providers/neuralwatt/models/glm-5.2.toml
index 0b7b1191535..b83cbca3c04 100644
--- a/providers/neuralwatt/models/glm-5.2.toml
+++ b/providers/neuralwatt/models/glm-5.2.toml
@@ -1,3 +1,5 @@
+# Deprecated: Neuralwatt retired the GLM-5.2 family from GET /v1/models
+# (checked 2026-09-18); retained for pricing history.
 # Budget: thinking_token_budget (integer reasoning tokens)
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2"
@@ -5,6 +7,7 @@ description = "Flagship GLM model for hybrid reasoning, coding, and agentic engi
 release_date = "2026-06-17"
 last_updated = "2026-06-17"
 structured_output = false
+status = "deprecated"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/glm-5.3-flash-flex.toml b/providers/neuralwatt/models/glm-5.3-flash-flex.toml
new file mode 100644
index 00000000000..502beae0882
--- /dev/null
+++ b/providers/neuralwatt/models/glm-5.3-flash-flex.toml
@@ -0,0 +1,34 @@
+# Flex-tier variant of glm-5.3-flash: same model, context (1,048,560),
+# vision, and effort ladder as the standard tier; requested via the `-flex`
+# model ID or service_tier="flex" and requires stream=true (non-streaming
+# requests fall through to the standard tier). Cost is the flex rate: 0.65x
+# of standard pricing (35% off) per the flex-tier guide, applied to the
+# standard rates advertised by GET /v1/models ($0.15/$0.5/$0.03, checked
+# 2026-09-18). JSON mode is off on this host, unlike the lab default.
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash Flex"
+structured_output = false
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.0975
+output = 0.325
+cache_read = 0.0195
+
+[limit]
+context = 1_048_560
+output = 1_048_560
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.3-flash.toml b/providers/neuralwatt/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..8f87fce7fcc
--- /dev/null
+++ b/providers/neuralwatt/models/glm-5.3-flash.toml
@@ -0,0 +1,35 @@
+# Exited preview 2026-09-11. Pricing, vision, and effort ladder from the
+# public GET /v1/models metadata (checked 2026-09-18): $0.15/$0.5/$0.03,
+# 1,048,560-token context, 20-image cap. Reasoning cannot be disabled
+# (reasoning.mandatory = true), so there is no none level and no toggle;
+# low, high and max are distinct efforts, with minimal/medium/xhigh accepted
+# as aliases onto them. Budget: thinking_token_budget (integer reasoning
+# tokens). Neuralwatt serves text+image input only (lab metadata also lists
+# video/pdf). JSON mode is off on this host, unlike the lab default.
+# https://portal.neuralwatt.com/docs/api/chat-completions
+# https://portal.neuralwatt.com/models/glm-5.3-flash
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash"
+structured_output = false
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_560
+output = 1_048_560
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/glm-5.3-flex.toml b/providers/neuralwatt/models/glm-5.3-flex.toml
new file mode 100644
index 00000000000..fa4a8aaef2c
--- /dev/null
+++ b/providers/neuralwatt/models/glm-5.3-flex.toml
@@ -0,0 +1,30 @@
+# Flex-tier variant of glm-5.3: same model, context (1,048,560), and effort
+# ladder as the standard tier; requested via the `-flex` model ID or
+# service_tier="flex" and requires stream=true (non-streaming requests fall
+# through to the standard tier). Cost is the flex rate: 0.65x of standard
+# pricing (35% off) per the flex-tier guide, applied to the standard rates
+# advertised by GET /v1/models ($1.45/$4.5/$0.145, checked 2026-09-18).
+# Budget: thinking_token_budget (integer reasoning tokens)
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "zhipuai/glm-5.3"
+name = "GLM 5.3 Flex"
+structured_output = false
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.9425
+output = 2.925
+cache_read = 0.09425
+
+[limit]
+context = 1_048_560
+output = 1_048_560
diff --git a/providers/neuralwatt/models/glm-5.3.toml b/providers/neuralwatt/models/glm-5.3.toml
index 700f54f026e..5e77d0a249d 100644
--- a/providers/neuralwatt/models/glm-5.3.toml
+++ b/providers/neuralwatt/models/glm-5.3.toml
@@ -1,17 +1,13 @@
-# Preview rollout: glm-5.3 is a gated preview on Neuralwatt, visible to
-# accounts granted access per the Preview Models guide. Its list price is
-# Neuralwatt's GLM 5.2 parity default and is marked for review at launch,
-# per the model description in GET /v1/models (checked 2026-08-30).
+# Exited preview 2026-09-10; the preview-parity list price ($1.45/$4.5/$0.145)
+# was confirmed as the launch price per GET /v1/models (checked 2026-09-18).
 # Reasoning cannot be disabled (reasoning.mandatory = true), so there is no
 # none level and no toggle. The API supports low, high and max as distinct
 # efforts; minimal, medium and xhigh are accepted as aliases onto them.
 # Budget: thinking_token_budget (integer reasoning tokens)
 # https://portal.neuralwatt.com/docs/api/chat-completions
-# https://portal.neuralwatt.com/docs/guides/preview-models
 base_model = "zhipuai/glm-5.3"
 name = "GLM 5.3"
 structured_output = false
-status = "beta"
 
 interleaved = true
 
diff --git a/providers/neuralwatt/models/qwen-3.8-27b-flex.toml b/providers/neuralwatt/models/qwen-3.8-27b-flex.toml
new file mode 100644
index 00000000000..7021462ac7f
--- /dev/null
+++ b/providers/neuralwatt/models/qwen-3.8-27b-flex.toml
@@ -0,0 +1,33 @@
+# Flex-tier variant of qwen-3.8-27b: same model, context (262,128), vision,
+# and effort ladder as the standard tier; requested via the `-flex` model ID
+# or service_tier="flex" and requires stream=true (non-streaming requests
+# fall through to the standard tier). Cost is the flex rate: 0.65x of
+# standard pricing (35% off) per the flex-tier guide, applied to the standard
+# rates advertised by GET /v1/models ($0.45/$3.2/$0.25, checked 2026-09-18).
+# Budget: thinking_token_budget (integer reasoning tokens)
+# Neuralwatt serves text+image input only (lab metadata also lists video).
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8 27B Flex"
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2925
+output = 2.08
+cache_read = 0.1625
+
+[limit]
+context = 262_128
+output = 131_072
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/neuralwatt/models/qwen-3.8-27b.toml b/providers/neuralwatt/models/qwen-3.8-27b.toml
index baa1ccb1e23..e2749dc1438 100644
--- a/providers/neuralwatt/models/qwen-3.8-27b.toml
+++ b/providers/neuralwatt/models/qwen-3.8-27b.toml
@@ -1,6 +1,5 @@
-# Preview rollout: qwen-3.8-27b is in preview on Neuralwatt — early-access
-# program, granted via the portal enroll page; absent from the public
-# /v1/models catalog until access is granted.
+# Left preview and joined the public GET /v1/models catalog; the output cap
+# was raised from 65,536 to 131,072 tokens (checked 2026-09-18).
 # Effort: reasoning_effort = none|low|medium|xhigh (default xhigh, reasoning
 # on by default); aliases max/xhigh/high -> xhigh and minimal -> low, per the
 # chat-completions per-model table (docs verified against the live API on
@@ -9,12 +8,10 @@
 # model.
 # Neuralwatt serves text+image input only (lab metadata also lists video).
 # Pricing, context, output cap, and effort ladder from the model page and the
-# authenticated GET /v1/models metadata (checked 2026-08-27):
+# GET /v1/models metadata (checked 2026-09-18):
 # https://portal.neuralwatt.com/models/qwen-3.8-27b
 # https://portal.neuralwatt.com/docs/api/chat-completions
-# https://portal.neuralwatt.com/docs/guides/preview-models
 base_model = "alibaba/qwen3.8-27b"
-status = "beta"
 interleaved = true
 
 [[reasoning_options]]
@@ -31,7 +28,7 @@ cache_read = 0.25
 
 [limit]
 context = 262_128
-output = 65_536
+output = 131_072
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/neuralwatt/models/qwen3.6-35b-flex.toml b/providers/neuralwatt/models/qwen3.6-35b-flex.toml
new file mode 100644
index 00000000000..3f8bd238eca
--- /dev/null
+++ b/providers/neuralwatt/models/qwen3.6-35b-flex.toml
@@ -0,0 +1,32 @@
+# Flex-tier variant of qwen3.6-35b: same model, context (131,056), vision,
+# and effort ladder as the standard tier; requested via the `-flex` model ID
+# or service_tier="flex" and requires stream=true (non-streaming requests
+# fall through to the standard tier). Cost is the flex rate: 0.65x of
+# standard pricing (35% off) per the flex-tier guide, applied to the standard
+# rates advertised by GET /v1/models ($0.29/$1.15/$0.029, checked 2026-09-18).
+# Budget: thinking_token_budget (integer reasoning tokens)
+# https://portal.neuralwatt.com/docs/guides/flex-tier
+# https://portal.neuralwatt.com/docs/api/chat-completions
+base_model = "alibaba/qwen3.6-35b-a3b"
+name = "Qwen3.6 35B Flex"
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.1885
+output = 0.7475
+cache_read = 0.01885
+
+[limit]
+context = 131_056
+output = 131_056
+
+[modalities]
+input = ["text", "image"]

From 6c97e53166f543afe27be8cf75d99fa64b171a0a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 10:30:43 -0500
Subject: [PATCH 1789/2012] chore(sync): update Hugging Face model catalog
 (#7562)

* chore(sync): update Hugging Face model catalog

* fix(huggingface): add Hy4 reasoning options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../huggingface/models/tencent/Hy4-preview.toml     | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/huggingface/models/tencent/Hy4-preview.toml

diff --git a/providers/huggingface/models/tencent/Hy4-preview.toml b/providers/huggingface/models/tencent/Hy4-preview.toml
new file mode 100644
index 00000000000..5a2ca915a5b
--- /dev/null
+++ b/providers/huggingface/models/tencent/Hy4-preview.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning_effort = none|high; high is the default.
+# https://huggingface.co/tencent/Hy4-preview#quickstart
+base_model = "tencent/hy4-preview"
+description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[cost]
+input = 0.834
+output = 2.501
+
+[limit]
+context = 1_000_000

From 7c89bbad28ce8171e9f4853fed74edd8c8bf47a6 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 10:36:03 -0500
Subject: [PATCH 1790/2012] fix: [missing-model] xai:
 grok-imagine-image-quality (#7391)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 models/xai/grok-imagine-image-quality.toml    | 26 +++++++++++++++++++
 .../models/grok-imagine-image-quality.toml    |  8 ++++++
 2 files changed, 34 insertions(+)
 create mode 100644 models/xai/grok-imagine-image-quality.toml
 create mode 100644 providers/xai/models/grok-imagine-image-quality.toml

diff --git a/models/xai/grok-imagine-image-quality.toml b/models/xai/grok-imagine-image-quality.toml
new file mode 100644
index 00000000000..46eb1539995
--- /dev/null
+++ b/models/xai/grok-imagine-image-quality.toml
@@ -0,0 +1,26 @@
+# Sources:
+# - https://docs.x.ai/docs/models
+# - https://docs.x.ai/developers/models/grok-imagine-image-quality
+# - https://docs.x.ai/developers/pricing
+# - https://docs.x.ai/docs/guides/image-generation
+# Pricing: $0.05/image (1K), $0.07/image (2K); image input $0.01/image (not token-based; no [cost] authored)
+# Release: dated alias grok-imagine-image-quality-20260403; also aliased as grok-imagine-image-quality-latest, grok-imagine-image-pro
+
+name = "Grok Imagine Image Quality"
+description = "Higher-fidelity Grok Imagine image model for prompt-driven generation, editing, and visual design workflows"
+family = "grok"
+release_date = "2026-04-03"
+last_updated = "2026-04-03"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 16_000
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/providers/xai/models/grok-imagine-image-quality.toml b/providers/xai/models/grok-imagine-image-quality.toml
new file mode 100644
index 00000000000..caf513b6e28
--- /dev/null
+++ b/providers/xai/models/grok-imagine-image-quality.toml
@@ -0,0 +1,8 @@
+# Sources:
+# - https://docs.x.ai/docs/models
+# - https://docs.x.ai/developers/models/grok-imagine-image-quality
+# - https://docs.x.ai/developers/pricing
+# Pricing: $0.05/image (1K), $0.07/image (2K); image input $0.01/image (not token-based; no [cost] authored)
+# Aliases: grok-imagine-image-quality-20260403, grok-imagine-image-quality-latest, grok-imagine-image-pro
+
+base_model = "xai/grok-imagine-image-quality"

From 3afe1fe9568450183252db8300b85c1de0887368 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:36:11 +0800
Subject: [PATCH 1791/2012] feat(aihubmix): add gpt-6-astra (#7422)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/openai/gpt-6-astra.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/gpt-6-astra.toml | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/aihubmix/models/gpt-6-astra.toml

diff --git a/providers/aihubmix/models/gpt-6-astra.toml b/providers/aihubmix/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..b005bff58b8
--- /dev/null
+++ b/providers/aihubmix/models/gpt-6-astra.toml
@@ -0,0 +1,22 @@
+# Effort: low|medium|high|xhigh|max
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25

From 7815e59aed1c50ab5eb312b47502894b7a56aea3 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:36:21 +0800
Subject: [PATCH 1792/2012] feat(aihubmix): add gemini-3.8-flash (#7423)

* feat(aihubmix): add gemini-3.8-flash

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/google/gemini-3.8-flash.toml rather than being
repeated here.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* aihubmix/gemini-3.8-flash: drop budget_tokens from reasoning_options

The bare budget_tokens entry came from a generic multi-protocol field
dump, not from model-level evidence: first-party
providers/google/models/gemini-3.8-flash.toml and the AIHubMix peer
gemini-3.7-flash are both effort-only, and Google's documented
reasoning-token budgets apply to the 2.5 family, not 3.8.

Effort low|medium|high is unchanged and is the model-level control.
The leading comment loses the budget field paths along with it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/gemini-3.8-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/aihubmix/models/gemini-3.8-flash.toml

diff --git a/providers/aihubmix/models/gemini-3.8-flash.toml b/providers/aihubmix/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..9657ff10785
--- /dev/null
+++ b/providers/aihubmix/models/gemini-3.8-flash.toml
@@ -0,0 +1,14 @@
+# Effort: low|medium|high
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "google/gemini-3.8-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075

From cd17c40386ecc955e977b534a3d1fa399b22668b Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:36:28 +0800
Subject: [PATCH 1793/2012] feat(aihubmix): add hy4-preview (#7424)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/tencent/hy4-preview.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/hy4-preview.toml | 28 ++++++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/aihubmix/models/hy4-preview.toml

diff --git a/providers/aihubmix/models/hy4-preview.toml b/providers/aihubmix/models/hy4-preview.toml
new file mode 100644
index 00000000000..dc0f2ad6ac7
--- /dev/null
+++ b/providers/aihubmix/models/hy4-preview.toml
@@ -0,0 +1,28 @@
+# Off is effort=none; graded levels — no toggle. The same off elsewhere:
+# $.enable_thinking = true|false on the OpenAI-compatible /v1/chat/completions path (verified live 2026-09-11);
+# $.thinking.type = "enabled"|"disabled"|"adaptive" on /v1/messages; $.generationConfig.thinkingConfig on the Gemini path.
+# Effort: none|high
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# Budget:
+# integer $.reasoning.max_tokens on /v1/chat/completions; $.thinking.budget_tokens >= 1024 on /v1/messages;
+# integer $.generationConfig.thinkingConfig.thinkingBudget on the Gemini path (-1 dynamic, 0 off where supported);
+# the Responses path carries effort but has no reasoning-token budget field.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "tencent/hy4-preview"
+structured_output = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.845
+output = 2.535
+cache_read = 0.04225

From 88245969a40fc5610915ffd9fcbb3f162d26d3e0 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:36:37 +0800
Subject: [PATCH 1794/2012] feat(aihubmix): add muse-spark-1.3 (#7425)

* feat(aihubmix): add muse-spark-1.3

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/meta/muse-spark-1.3.toml rather than being
repeated here.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* aihubmix/muse-spark-1.3: drop budget_tokens from reasoning_options

The bare budget_tokens entry came from a generic multi-protocol field
dump rather than a model-level control: first-party Meta and the
established relays of meta/muse-spark-1.3 are effort-only, and the
same-host AIHubMix effort-only peers omit it as well.

The effort set stays at minimal|low|medium|high|xhigh. `max` is
deliberately not listed: Meta's 1.3 launch post says "Previously
available reasoning modes are available today with max reasoning
coming shortly after we finish additional safety testing", so the
"Muse Spark 1.3 (max)" benchmark label is not an API availability
statement yet.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/muse-spark-1.3.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/aihubmix/models/muse-spark-1.3.toml

diff --git a/providers/aihubmix/models/muse-spark-1.3.toml b/providers/aihubmix/models/muse-spark-1.3.toml
new file mode 100644
index 00000000000..21edfd6c678
--- /dev/null
+++ b/providers/aihubmix/models/muse-spark-1.3.toml
@@ -0,0 +1,14 @@
+# Effort: minimal|low|medium|high|xhigh
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "meta/muse-spark-1.3"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.375
+output = 4.675
+cache_read = 0.165

From 96655bf2b6494f39d2203a91cc97e532a2ae3e27 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:36:44 +0800
Subject: [PATCH 1795/2012] feat(aihubmix): add deepseek-v4-flash-0731-fast
 (#7429)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/deepseek/deepseek-v4-flash-0731.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 .../models/deepseek-v4-flash-0731-fast.toml   | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/aihubmix/models/deepseek-v4-flash-0731-fast.toml

diff --git a/providers/aihubmix/models/deepseek-v4-flash-0731-fast.toml b/providers/aihubmix/models/deepseek-v4-flash-0731-fast.toml
new file mode 100644
index 00000000000..db3b1b7c1e9
--- /dev/null
+++ b/providers/aihubmix/models/deepseek-v4-flash-0731-fast.toml
@@ -0,0 +1,24 @@
+# Toggle:
+# $.enable_thinking = true|false on the OpenAI-compatible /v1/chat/completions path (verified live 2026-09-11);
+# $.thinking.type = "enabled"|"disabled"|"adaptive" on /v1/messages; $.generationConfig.thinkingConfig on the Gemini path.
+# Effort: low|high|max
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "deepseek/deepseek-v4-flash-0731"
+name = "DeepSeek V4 Flash 0731 Fast"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.28
+output = 1.4
+cache_read = 0.07

From 63cca5fee192c2f25b7876af96cb9a0093250d94 Mon Sep 17 00:00:00 2001
From: Kastan Day <kastanvday@gmail.com>
Date: Mon, 21 Sep 2026 08:37:35 -0700
Subject: [PATCH 1796/2012] fix(cloudflare): correct catalogue reasoning
 efforts & `base_model`s (#7454)

* fix(cloudflare): refresh catalogue reasoning and canonical links

* fix(cloudflare): sync GLM 4.7 toggle and expand reasoning tables

* fix(cloudflare): cap oversized context and output limits

* fix(cloudflare): inherit SEA-LION description and clarify reasoning sources

* docs(cloudflare): cite GLM thinking control at its catalogue entry

* docs(cloudflare): cite dated DeepSeek effort controls

* fix(cloudflare): omit fixed Kimi K2.7 effort control
---
 .../aisingapore/gemma-sea-lion-v4-27b-it.toml | 18 +-----------
 .../deepseek-ai/deepseek-v4-flash-0731.toml   | 15 +++++-----
 .../@cf/deepseek-ai/deepseek-v4-pro-0813.toml | 10 +++++--
 .../models/@cf/google/gemma-4-26b-a4b-it.toml | 12 +++++---
 .../@cf/ibm-granite/granite-4.0-h-micro.toml  | 13 +--------
 .../mistral-small-3.1-24b-instruct.toml       | 10 +------
 .../models/@cf/moonshotai/kimi-k2.6.toml      | 12 +++++---
 .../models/@cf/moonshotai/kimi-k2.7-code.toml |  4 ++-
 .../@cf/nvidia/nemotron-3-120b-a12b.toml      |  8 ++++-
 .../models/@cf/qwen/qwen3.8-27b.toml          |  8 ++++-
 .../models/@cf/zai-org/glm-4.7-flash.toml     | 29 ++++++-------------
 .../models/@cf/zai-org/glm-5.2.toml           |  7 ++---
 .../models/@cf/zai-org/glm-5.3-flash.toml     | 23 ++++++---------
 .../models/@cf/zai-org/glm-5.3.toml           | 13 +++++++--
 14 files changed, 82 insertions(+), 100 deletions(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml b/providers/cloudflare-workers-ai/models/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
index 8976b92b55a..3fe191a4d99 100644
--- a/providers/cloudflare-workers-ai/models/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/aisingapore/gemma-sea-lion-v4-27b-it.toml
@@ -1,23 +1,7 @@
+base_model = "aisingapore/gemma-sea-lion-v4-27b-it"
 name = "Gemma Sea Lion V4 27B It"
-description = "Open Gemma instruction model for efficient chat and self-hosted deployments"
-family = "gemma"
-release_date = "2025-09-23"
-last_updated = "2025-09-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = false
 structured_output = false
-open_weights = true
 
 [cost]
 input = 0.351
 output = 0.555
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
index b72f1489e71..2385a1bc60b 100644
--- a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
@@ -1,13 +1,14 @@
-# Toggle: thinking.type = enabled|disabled.
-# Effort: reasoning_effort = high|max.
+# Effort: reasoning_effort = none|low|high|max; none disables thinking.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# Context/output capped to Cloudflare Workers AI context window (1,048,576).
+# Search reports 1,310,720, exceeding AI Gateway max_completion_tokens.
+# This dated release adds a distinct low effort; the older preview does not.
+# Creator: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731#chat-template
 base_model = "deepseek/deepseek-v4-flash-0731"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["high", "max"]
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.44
@@ -15,5 +16,5 @@ output = 1.32
 cache_read = 0.014
 
 [limit]
-context = 1_310_720
+context = 1_048_576
 output = 1_048_576
diff --git a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-pro-0813.toml b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
index b28c06ffbe1..32c7c6708b0 100644
--- a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-pro-0813.toml
@@ -1,8 +1,12 @@
-# Toggle: thinking.type = enabled|disabled.
-# Effort: reasoning_effort = high|max.
+# Effort: reasoning_effort = none|low|high|max; none disables thinking.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# This dated release adds a distinct low effort; the older preview does not.
+# Creator: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813#chat-template
 base_model = "deepseek/deepseek-v4-pro-0813"
 
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 1.32
diff --git a/providers/cloudflare-workers-ai/models/@cf/google/gemma-4-26b-a4b-it.toml b/providers/cloudflare-workers-ai/models/@cf/google/gemma-4-26b-a4b-it.toml
index 817334b2cee..2c6a8e4386f 100644
--- a/providers/cloudflare-workers-ai/models/@cf/google/gemma-4-26b-a4b-it.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/google/gemma-4-26b-a4b-it.toml
@@ -1,11 +1,15 @@
-# Native `/ai/run` accepts `reasoning_effort = low|medium|high` and
-# `chat_template_kwargs.enable_thinking = true|false`; no budget is documented.
-# https://developers.cloudflare.com/workers-ai/models/gemma-4-26b-a4b-it/sync-input.json (accessed 2026-06-25)
+# Effort: reasoning_effort = none|high; none disables thinking.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# Per-model Search config normalizes low/medium to high; these are aliases,
+# not distinct levels despite the generic sync-input.json effort enum.
 
 base_model = "google/gemma-4-26b-a4b-it"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 interleaved = true
 
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
 [cost]
 input = 0.1
 output = 0.3
diff --git a/providers/cloudflare-workers-ai/models/@cf/ibm-granite/granite-4.0-h-micro.toml b/providers/cloudflare-workers-ai/models/@cf/ibm-granite/granite-4.0-h-micro.toml
index a43a661e427..1aa972501cf 100644
--- a/providers/cloudflare-workers-ai/models/@cf/ibm-granite/granite-4.0-h-micro.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/ibm-granite/granite-4.0-h-micro.toml
@@ -1,14 +1,7 @@
+base_model = "ibm/granite-4-h-micro"
 name = "Granite 4.0 H Micro"
 description = "Efficient model for low-latency assistance, extraction, and routine automation"
-family = "granite"
-release_date = "2025-10-07"
-last_updated = "2025-10-07"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
 
 [cost]
 input = 0.017
@@ -17,7 +10,3 @@ output = 0.112
 [limit]
 context = 131_000
 output = 131_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cloudflare-workers-ai/models/@cf/mistralai/mistral-small-3.1-24b-instruct.toml b/providers/cloudflare-workers-ai/models/@cf/mistralai/mistral-small-3.1-24b-instruct.toml
index e456a0e8625..7ed149c081e 100644
--- a/providers/cloudflare-workers-ai/models/@cf/mistralai/mistral-small-3.1-24b-instruct.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/mistralai/mistral-small-3.1-24b-instruct.toml
@@ -1,23 +1,15 @@
+base_model = "mistral/mistral-small-3-1-24b-instruct-2503"
 name = "Mistral Small 3.1 24B Instruct"
 description = "Efficient Mistral model for fast chat, extraction, and production assistants"
-family = "mistral-small"
-release_date = "2025-03-18"
-last_updated = "2025-03-18"
 attachment = false
-reasoning = false
-temperature = true
-tool_call = true
 structured_output = false
-open_weights = true
 
 [cost]
 input = 0.351
 output = 0.555
 
 [limit]
-context = 128_000
 output = 128_000
 
 [modalities]
 input = ["text"]
-output = ["text"]
diff --git a/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.6.toml b/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.6.toml
index 3c4c22a8305..6148d7218ca 100644
--- a/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.6.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.6.toml
@@ -1,9 +1,13 @@
-# Native `/ai/run` accepts `reasoning_effort = low|medium|high` and
-# `chat_template_kwargs.thinking = true|false`; no budget is documented.
-# https://developers.cloudflare.com/workers-ai/models/kimi-k2.6/sync-input.json (accessed 2026-06-25)
+# Effort: reasoning_effort = none|high; none selects instant mode.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# Per-model Search config normalizes low/medium to high; these are aliases,
+# not distinct levels despite the generic sync-input.json effort enum.
 
 base_model = "moonshotai/kimi-k2.6"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.7-code.toml b/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.7-code.toml
index a06c8bc9a29..43a8509e3a6 100644
--- a/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.7-code.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/moonshotai/kimi-k2.7-code.toml
@@ -1,6 +1,8 @@
+# Thinking is mandatory; high is the only effective effort, so aliases add no caller control.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
 base_model = "moonshotai/kimi-k2.7-code"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
+reasoning_options = []
 
 [cost]
 input = 0.95
diff --git a/providers/cloudflare-workers-ai/models/@cf/nvidia/nemotron-3-120b-a12b.toml b/providers/cloudflare-workers-ai/models/@cf/nvidia/nemotron-3-120b-a12b.toml
index 554decfc2e3..a75a99e22b1 100644
--- a/providers/cloudflare-workers-ai/models/@cf/nvidia/nemotron-3-120b-a12b.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/nvidia/nemotron-3-120b-a12b.toml
@@ -5,9 +5,15 @@
 base_model = "nvidia/nemotron-3-super-120b-a12b"
 name = "Nemotron 3 Super 120B"
 structured_output = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 interleaved = true
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.5
 output = 1.5
diff --git a/providers/cloudflare-workers-ai/models/@cf/qwen/qwen3.8-27b.toml b/providers/cloudflare-workers-ai/models/@cf/qwen/qwen3.8-27b.toml
index 307aa38fb79..f491dd473a3 100644
--- a/providers/cloudflare-workers-ai/models/@cf/qwen/qwen3.8-27b.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/qwen/qwen3.8-27b.toml
@@ -5,7 +5,13 @@
 
 base_model = "alibaba/qwen3.8-27b"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "xhigh"] }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
 
 [cost]
 input = 0.45
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-4.7-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-4.7-flash.toml
index 197fa7c94b5..37f0d8a203e 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-4.7-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-4.7-flash.toml
@@ -1,32 +1,21 @@
-# Native `/ai/run` accepts `reasoning_effort = low|medium|high` and
-# `chat_template_kwargs.enable_thinking = true|false`; no budget is documented.
-# https://developers.cloudflare.com/workers-ai/models/glm-4.7-flash/sync-input.json (accessed 2026-06-25)
-
-name = "GLM-4.7-Flash"
+# Toggle: chat_template_kwargs.enable_thinking = true|false (default: true).
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# Workers AI's per-model config exposes only the binary thinking toggle;
+# the generic sync-input.json low/medium/high enum is not an effort selector here.
+# Creator template: https://huggingface.co/zai-org/GLM-4.7-Flash/blob/main/chat_template.jinja
+base_model = "zhipuai/glm-4.7-flash"
 description = "Efficient GLM model for fast reasoning, coding, and agent workflows"
-family = "glm-flash"
-release_date = "2026-01-19"
-last_updated = "2026-01-19"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
 structured_output = true
-knowledge = "2025-04"
-open_weights = true
 
 [interleaved]
 field = "reasoning_content"
 
+[[reasoning_options]]
+type = "toggle"
+
 [cost]
 input = 0.0605
 output = 0.4
 
 [limit]
 context = 131_072
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.2.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.2.toml
index 5642ab6e4ea..48337ad55f9 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.2.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.2.toml
@@ -1,14 +1,13 @@
+# Effort: reasoning_effort = none|high|max; none disables thinking.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
 # Context and output limits verified against Cloudflare Workers AI's OpenAI-compatible endpoint.
 # https://developers.cloudflare.com/workers-ai/models/glm-5.2/
 base_model = "zhipuai/glm-5.2"
 name = "Glm 5.2"
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high"]
+values = ["none", "high", "max"]
 
 [cost]
 input = 1.4
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
index 000ebf8a49a..83a85d6c195 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -1,18 +1,14 @@
+# Effort: reasoning_effort = low|high|max; thinking is mandatory.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
 # Context/output capped to Cloudflare Workers AI context window (1,048,576).
-# Catalog/search previously reported 1,310,720, which exceeds AI Gateway max_completion_tokens.
-# https://developers.cloudflare.com/workers-ai/models/glm-5.3-flash/
+# Search reports 1,310,720, exceeding AI Gateway max_completion_tokens.
+base_model = "zhipuai/glm-5.3-flash"
 name = "Glm 5.3 Flash"
 description = "GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm-flash"
-release_date = "2026-08-26"
-last_updated = "2026-08-26"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
 input = 0.15
@@ -20,9 +16,8 @@ output = 0.5
 cache_read = 0.03
 
 [limit]
-context = 1_310_720
+context = 1_048_576
 output = 1_048_576
 
 [modalities]
 input = ["text", "image"]
-output = ["text"]
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
index a5f20bbaf80..a66154848f3 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
@@ -1,7 +1,14 @@
+# Effort: reasoning_effort = low|high|max; thinking is mandatory.
+# Source: Workers AI /ai/models/search?format=openrouter (2026-09-18).
+# Context/output capped to Cloudflare Workers AI context window (1,048,576).
+# Search reports 1,310,720, exceeding AI Gateway max_completion_tokens.
 base_model = "zhipuai/glm-5.3"
 name = "Glm 5.3"
 description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [cost]
 input = 1.4
@@ -9,5 +16,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_310_720
-output = 1_310_720
+context = 1_048_576
+output = 1_048_576

From e52dfc2abcc74fcb39c9a2bd7736f998517dc518 Mon Sep 17 00:00:00 2001
From: Daniel Chen <daniel85.chen@gmail.com>
Date: Mon, 21 Sep 2026 08:45:33 -0700
Subject: [PATCH 1797/2012] feat(opencode):added grok-4.7 to go and zen

---
 models/xai/grok-4.7.toml                   | 20 ++++++++++++++++++++
 providers/opencode-go/models/grok-4.7.toml | 19 +++++++++++++++++++
 2 files changed, 39 insertions(+)
 create mode 100644 models/xai/grok-4.7.toml
 create mode 100644 providers/opencode-go/models/grok-4.7.toml

diff --git a/models/xai/grok-4.7.toml b/models/xai/grok-4.7.toml
new file mode 100644
index 00000000000..1142f86cd13
--- /dev/null
+++ b/models/xai/grok-4.7.toml
@@ -0,0 +1,20 @@
+name = "Grok 4.7"
+description = "xAI's frontier model for long-running agents, coding, knowledge work, and visual projects"
+family = "grok"
+knowledge = "2026-02-01"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[limit]
+context = 500_000
+output = 500_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/opencode-go/models/grok-4.7.toml b/providers/opencode-go/models/grok-4.7.toml
new file mode 100644
index 00000000000..edde035f699
--- /dev/null
+++ b/providers/opencode-go/models/grok-4.7.toml
@@ -0,0 +1,19 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000  }
+input = 4
+output = 12
+cache_read = 1
+
+[provider]
+npm = "@ai-sdk/openai"

From 739a4e206b88ea0699c94bb57ea3b138bc23c743 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:16 +0000
Subject: [PATCH 1798/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#7640)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml          | 2 +-
 .../cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml | 2 +-
 providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
index 2385a1bc60b..48cdf638dfe 100644
--- a/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml
@@ -16,5 +16,5 @@ output = 1.32
 cache_read = 0.014
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 1_048_576
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
index 83a85d6c195..13c5aacae4c 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml
@@ -16,7 +16,7 @@ output = 0.5
 cache_read = 0.03
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 1_048_576
 
 [modalities]
diff --git a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
index a66154848f3..456f1fb0549 100644
--- a/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml
@@ -16,5 +16,5 @@ output = 4.4
 cache_read = 0.26
 
 [limit]
-context = 1_048_576
+context = 1_310_720
 output = 1_048_576

From 68310f550497b27d52e2f53c8f828fb63e3d42c4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:29 +0000
Subject: [PATCH 1799/2012] chore(sync): update xAI model catalog (#7644)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/xai/models/grok-imagine-image-quality.toml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/providers/xai/models/grok-imagine-image-quality.toml b/providers/xai/models/grok-imagine-image-quality.toml
index caf513b6e28..1be5068f356 100644
--- a/providers/xai/models/grok-imagine-image-quality.toml
+++ b/providers/xai/models/grok-imagine-image-quality.toml
@@ -4,5 +4,8 @@
 # - https://docs.x.ai/developers/pricing
 # Pricing: $0.05/image (1K), $0.07/image (2K); image input $0.01/image (not token-based; no [cost] authored)
 # Aliases: grok-imagine-image-quality-20260403, grok-imagine-image-quality-latest, grok-imagine-image-pro
-
 base_model = "xai/grok-imagine-image-quality"
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["image", "pdf"]

From 39e41897358590fdd14291475e0d3f4725f536d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:33 +0000
Subject: [PATCH 1800/2012] chore(sync): update OpenRouter model catalog
 (#7645)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-4.toml       | 31 -------------------
 .../models/anthropic/claude-sonnet-4.toml     |  2 +-
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++--
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 providers/openrouter/models/tencent/hy3.toml  |  6 ++--
 .../openrouter/models/x-ai/grok-4.7.toml      | 23 ++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 .../openrouter/models/~x-ai/grok-latest.toml  | 12 +++----
 8 files changed, 42 insertions(+), 50 deletions(-)
 delete mode 100644 providers/openrouter/models/anthropic/claude-opus-4.toml
 create mode 100644 providers/openrouter/models/x-ai/grok-4.7.toml

diff --git a/providers/openrouter/models/anthropic/claude-opus-4.toml b/providers/openrouter/models/anthropic/claude-opus-4.toml
deleted file mode 100644
index 094f7f59203..00000000000
--- a/providers/openrouter/models/anthropic/claude-opus-4.toml
+++ /dev/null
@@ -1,31 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Claude Opus 4"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-05-22"
-last_updated = "2025-05-22"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-knowledge = "2025-01-31"
-open_weights = false
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-cache_write = 18.75
-
-[limit]
-context = 200_000
-output = 32_000
-
-[modalities]
-input = ["image", "text", "pdf"]
-output = ["text"]
diff --git a/providers/openrouter/models/anthropic/claude-sonnet-4.toml b/providers/openrouter/models/anthropic/claude-sonnet-4.toml
index 7baa5f4f635..b7b54a299a9 100644
--- a/providers/openrouter/models/anthropic/claude-sonnet-4.toml
+++ b/providers/openrouter/models/anthropic/claude-sonnet-4.toml
@@ -30,7 +30,7 @@ cache_read = 0.6
 cache_write = 7.5
 
 [limit]
-context = 1_000_000
+context = 200_000
 output = 64_000
 
 [modalities]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 32fc69b9875..13dfedbfc61 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.019272
+input = 0.57288
+output = 1.71864
+cache_read = 0.019096
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b7feb0289f9..a705f18ca62 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.942906
-output = 1.885812
-cache_read = 0.078576
+input = 0.936294
+output = 1.872588
+cache_read = 0.078025
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/x-ai/grok-4.7.toml b/providers/openrouter/models/x-ai/grok-4.7.toml
new file mode 100644
index 00000000000..49d2235d6f9
--- /dev/null
+++ b/providers/openrouter/models/x-ai/grok-4.7.toml
@@ -0,0 +1,23 @@
+base_model = "xai/grok-4.7"
+description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.6
+output = 4.8
+cache_read = 0.4
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 3.2
+output = 9.6
+cache_read = 0.8
+
+[limit]
+output = 450_000
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 9b9fc44533c..3a21dc3b34a 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.019272
+input = 0.57288
+output = 1.71864
+cache_read = 0.019096
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~x-ai/grok-latest.toml b/providers/openrouter/models/~x-ai/grok-latest.toml
index ef7efee492e..2ce36baf40e 100644
--- a/providers/openrouter/models/~x-ai/grok-latest.toml
+++ b/providers/openrouter/models/~x-ai/grok-latest.toml
@@ -15,15 +15,15 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 2
-output = 6
-cache_read = 0.5
+input = 1.6
+output = 4.8
+cache_read = 0.4
 
 [[cost.tiers]]
 tier = { type = "context", size = 200_000 }
-input = 4
-output = 12
-cache_read = 1
+input = 3.2
+output = 9.6
+cache_read = 0.8
 
 [limit]
 context = 500_000

From b6a3717a04c18450eacb27fd10a5aa5c1d2bd01c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:37 +0000
Subject: [PATCH 1801/2012] chore(sync): update Kilo model catalog (#7647)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/anthropic/claude-opus-4.toml  | 29 -------------------
 providers/kilo/models/tencent/hy3.toml        |  6 ++--
 providers/kilo/models/x-ai/grok-4.7.toml      | 17 +++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 providers/kilo/models/~x-ai/grok-latest.toml  |  6 ++--
 5 files changed, 26 insertions(+), 38 deletions(-)
 delete mode 100644 providers/kilo/models/anthropic/claude-opus-4.toml
 create mode 100644 providers/kilo/models/x-ai/grok-4.7.toml

diff --git a/providers/kilo/models/anthropic/claude-opus-4.toml b/providers/kilo/models/anthropic/claude-opus-4.toml
deleted file mode 100644
index 6e622f068ab..00000000000
--- a/providers/kilo/models/anthropic/claude-opus-4.toml
+++ /dev/null
@@ -1,29 +0,0 @@
-name = "Anthropic: Claude Opus 4 ($$$$)"
-description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
-family = "claude-opus"
-release_date = "2025-05-22"
-last_updated = "2025-05-22"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 15
-output = 75
-cache_read = 1.5
-cache_write = 18.75
-
-[limit]
-context = 200_000
-output = 32_000
-
-[modalities]
-input = ["image", "text", "pdf"]
-output = ["text"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index 2bdfa3cbd7b..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/x-ai/grok-4.7.toml b/providers/kilo/models/x-ai/grok-4.7.toml
new file mode 100644
index 00000000000..bdb4be1fce7
--- /dev/null
+++ b/providers/kilo/models/x-ai/grok-4.7.toml
@@ -0,0 +1,17 @@
+base_model = "xai/grok-4.7"
+description = "Grok 4.7 is SpaceXAI's smartest model with frontier performance on coding, knowledge work, and STEM."
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.6
+output = 4.8
+cache_read = 0.4
+
+[limit]
+output = 450_000
+
+[modalities]
+input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index a92bf5149a1..d1321db247a 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57816
-output = 1.73448
-cache_read = 0.019272
+input = 0.57288
+output = 1.71864
+cache_read = 0.019096
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~x-ai/grok-latest.toml b/providers/kilo/models/~x-ai/grok-latest.toml
index 5f182ec62c7..85f00745faa 100644
--- a/providers/kilo/models/~x-ai/grok-latest.toml
+++ b/providers/kilo/models/~x-ai/grok-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 2
-output = 6
-cache_read = 0.5
+input = 1.6
+output = 4.8
+cache_read = 0.4
 
 [limit]
 context = 500_000

From b23f4e5f2da6b78079466db4fd331823097f46d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:44 +0000
Subject: [PATCH 1802/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7646)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/step-5-preview.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/empiriolabs/models/step-5-preview.toml

diff --git a/providers/empiriolabs/models/step-5-preview.toml b/providers/empiriolabs/models/step-5-preview.toml
new file mode 100644
index 00000000000..c4ca4d28659
--- /dev/null
+++ b/providers/empiriolabs/models/step-5-preview.toml
@@ -0,0 +1,16 @@
+base_model = "stepfun/step-5-preview"
+base_model_omit = ["limit.input"]
+temperature = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 1
+output = 2.7
+cache_read = 0.05
+
+[limit]
+context = 1_024_000
+output = 131_072

From 9e4e5ce97e4a014184f6bb76db7c2a8ed35a22aa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 16:29:53 +0000
Subject: [PATCH 1803/2012] chore(sync): update NanoGPT model catalog (#7642)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/gemma-4-12b-it.toml | 11 +---------
 .../models/stepfun/step-5-preview.toml        | 20 +------------------
 2 files changed, 2 insertions(+), 29 deletions(-)

diff --git a/providers/nano-gpt/models/gemma-4-12b-it.toml b/providers/nano-gpt/models/gemma-4-12b-it.toml
index 6959fb985f1..3d3dd5e5581 100644
--- a/providers/nano-gpt/models/gemma-4-12b-it.toml
+++ b/providers/nano-gpt/models/gemma-4-12b-it.toml
@@ -1,13 +1,6 @@
+base_model = "google/gemma-4-12b-it"
 name = "Gemma 4 12B Instruct"
-description = "Google's Gemma 4 12B Instruct is an open-weight multimodal model for text, image, audio, and video understanding, with tool calling and structured output support."
-family = "gemma"
-release_date = "2026-08-01"
-last_updated = "2026-08-01"
-attachment = true
 reasoning = false
-tool_call = true
-structured_output = true
-open_weights = true
 
 [cost]
 input = 0.05
@@ -17,8 +10,6 @@ cache_read = 0.025
 [limit]
 context = 131_072
 input = 131_072
-output = 32_768
 
 [modalities]
 input = ["text", "image", "video", "audio"]
-output = ["text"]
diff --git a/providers/nano-gpt/models/stepfun/step-5-preview.toml b/providers/nano-gpt/models/stepfun/step-5-preview.toml
index e6b122e57eb..20ebdd2b522 100644
--- a/providers/nano-gpt/models/stepfun/step-5-preview.toml
+++ b/providers/nano-gpt/models/stepfun/step-5-preview.toml
@@ -1,13 +1,4 @@
-name = "Step 5 Preview"
-description = "Step 5 Preview is StepFun's 600B sparse MoE frontier model for production-scale agents, activating 27B parameters per token. It is built for software engineering, long-horizon tool use, research, professional knowledge work, and finance, with native text, image, and video understanding and a 1M-token context window. ⚠️ Note: This model routes through StepFun, so privacy and logging guarantees may be limited."
-family = "step"
-release_date = "2026-09-20"
-last_updated = "2026-09-20"
-attachment = true
-reasoning = true
-tool_call = true
-structured_output = true
-open_weights = false
+base_model = "stepfun/step-5-preview"
 
 [[reasoning_options]]
 type = "effort"
@@ -17,12 +8,3 @@ values = ["low", "medium", "high"]
 input = 1
 output = 2.7
 cache_read = 0.05
-
-[limit]
-context = 1_000_000
-input = 1_000_000
-output = 1_000_000
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]

From ce2cb0547484fb62bee8813352123d5a6a5f550b Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 11:40:51 -0500
Subject: [PATCH 1804/2012] fix: [missing-model] xai: grok-4.7 (#7648)

* fix: [missing-model] xai: grok-4.7

* Add 'pdf' to input modalities in grok-4.7.toml

* Delete modalities input from grok-4.7.toml

Removed modalities section from grok-4.7.toml

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
---
 models/xai/grok-4.7.toml           |  5 +++--
 providers/xai/models/grok-4.7.toml | 18 ++++++++++++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)
 create mode 100644 providers/xai/models/grok-4.7.toml

diff --git a/models/xai/grok-4.7.toml b/models/xai/grok-4.7.toml
index 1142f86cd13..7872d4c7219 100644
--- a/models/xai/grok-4.7.toml
+++ b/models/xai/grok-4.7.toml
@@ -1,7 +1,8 @@
+# Sources: https://docs.x.ai/developers/models/grok-4.7, https://docs.x.ai/developers/models
 name = "Grok 4.7"
 description = "xAI's frontier model for long-running agents, coding, knowledge work, and visual projects"
 family = "grok"
-knowledge = "2026-02-01"
+knowledge = "2026-05"
 release_date = "2026-09-21"
 last_updated = "2026-09-21"
 attachment = true
@@ -16,5 +17,5 @@ context = 500_000
 output = 500_000
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "pdf"]
 output = ["text"]
diff --git a/providers/xai/models/grok-4.7.toml b/providers/xai/models/grok-4.7.toml
new file mode 100644
index 00000000000..c205cfd8209
--- /dev/null
+++ b/providers/xai/models/grok-4.7.toml
@@ -0,0 +1,18 @@
+# Sources: https://docs.x.ai/developers/models/grok-4.7, https://docs.x.ai/developers/pricing, and https://docs.x.ai/developers/model-capabilities/text/reasoning
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+

From 1c042a3aaf0a037ba4fdf7332af31481db0c2974 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 11:50:28 -0500
Subject: [PATCH 1805/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7641)

* chore(sync): update Vercel AI Gateway model catalog

* fix(sync): preserve Vercel tiered pricing

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/vercel.ts    | 69 +++++++++++++++++--
 packages/core/test/sync.test.ts               | 28 +++++++-
 .../models/alibaba/qwen-3.6-max-preview.toml  |  7 ++
 .../models/alibaba/qwen3-coder-plus.toml      | 18 +++++
 .../vercel/models/alibaba/qwen3-coder.toml    | 12 ++++
 .../models/alibaba/qwen3-max-preview.toml     | 12 ++++
 .../models/alibaba/qwen3-max-thinking.toml    | 18 ++++-
 .../vercel/models/alibaba/qwen3-max.toml      | 12 ++++
 .../vercel/models/alibaba/qwen3.5-plus.toml   |  7 ++
 .../vercel/models/alibaba/qwen3.6-plus.toml   |  7 ++
 .../vercel/models/alibaba/qwen3.7-flash.toml  | 14 ++++
 .../vercel/models/alibaba/qwen3.7-plus.toml   |  7 ++
 .../models/anthropic/claude-sonnet-4.toml     |  7 ++
 .../vercel/models/bytedance/seed-1.6.toml     |  6 ++
 .../vercel/models/bytedance/seed-1.8.toml     | 14 +++-
 .../models/google/gemini-3.1-pro-preview.toml |  6 ++
 .../vercel/models/openai/gpt-5.4-pro.toml     |  5 ++
 providers/vercel/models/openai/gpt-5.4.toml   | 12 +++-
 .../vercel/models/openai/gpt-5.5-pro.toml     |  5 ++
 providers/vercel/models/openai/gpt-5.5.toml   |  6 ++
 .../models/openai/gpt-5.6-luna-fast.toml      |  7 ++
 .../vercel/models/openai/gpt-5.6-luna.toml    |  7 ++
 .../models/openai/gpt-5.6-sol-fast.toml       |  7 ++
 .../vercel/models/openai/gpt-5.6-sol.toml     |  7 ++
 .../models/openai/gpt-5.6-terra-fast.toml     |  7 ++
 .../vercel/models/openai/gpt-5.6-terra.toml   |  7 ++
 .../models/openai/gpt-6-astra-fast.toml       |  2 +-
 .../vercel/models/sakana/fugu-ultra-v2.toml   | 11 ++-
 .../vercel/models/sakana/fugu-ultra.toml      |  6 ++
 .../spacexai/grok-4.20-multi-agent-beta.toml  |  6 ++
 .../spacexai/grok-4.20-multi-agent.toml       |  6 ++
 .../grok-4.20-non-reasoning-beta.toml         |  6 ++
 .../spacexai/grok-4.20-non-reasoning.toml     |  6 ++
 .../spacexai/grok-4.20-reasoning-beta.toml    |  6 ++
 .../models/spacexai/grok-4.20-reasoning.toml  |  6 ++
 .../vercel/models/spacexai/grok-4.3.toml      |  7 ++
 .../vercel/models/spacexai/grok-4.5.toml      |  7 ++
 .../vercel/models/spacexai/grok-4.6.toml      |  7 ++
 .../vercel/models/spacexai/grok-4.7.toml      | 13 ++++
 .../models/spacexai/grok-build-0.1.toml       |  6 ++
 .../vercel/models/stepfun/step-5-preview.toml | 12 +---
 41 files changed, 396 insertions(+), 25 deletions(-)
 create mode 100644 providers/vercel/models/spacexai/grok-4.7.toml

diff --git a/packages/core/src/sync/providers/vercel.ts b/packages/core/src/sync/providers/vercel.ts
index fdf8be631dd..e7800f77e32 100644
--- a/packages/core/src/sync/providers/vercel.ts
+++ b/packages/core/src/sync/providers/vercel.ts
@@ -237,17 +237,75 @@ function price(value: string | undefined) {
     : undefined;
 }
 
+function tieredPrice(value: string | undefined, tiers: z.infer<typeof PricingTier>[] | undefined) {
+  const base = price(value);
+  const normalized = (tiers ?? [])
+    .map((tier, index, values) => ({
+      start: tier.min ?? (index === 0 ? 0 : values[index - 1]?.max ?? 0),
+      cost: price(tier.cost),
+    }))
+    .filter((tier): tier is { start: number; cost: number } => tier.cost !== undefined)
+    .sort((a, b) => a.start - b.start);
+
+  return {
+    base: normalized[0]?.cost ?? base,
+    thresholds: normalized.map((tier) => tier.start).filter((start) => start > 0),
+    at(threshold: number) {
+      return normalized.findLast((tier) => tier.start <= threshold)?.cost ?? base;
+    },
+  };
+}
+
 function buildCost(pricing: VercelModel["pricing"], existing?: ExistingModel["cost"]) {
-  const input = price(pricing?.input_tiers?.[0]?.cost ?? pricing?.input);
-  const output = price(pricing?.output_tiers?.[0]?.cost ?? pricing?.output);
+  const hasPricingTiers = [
+    pricing?.input_tiers,
+    pricing?.output_tiers,
+    pricing?.input_cache_read_tiers,
+    pricing?.input_cache_write_tiers,
+  ].some((tiers) => (tiers?.length ?? 0) > 0);
+  const inputPrice = tieredPrice(pricing?.input, pricing?.input_tiers);
+  const outputPrice = tieredPrice(pricing?.output, pricing?.output_tiers);
+  const cacheReadPrice = tieredPrice(pricing?.input_cache_read, pricing?.input_cache_read_tiers);
+  const cacheWritePrice = tieredPrice(pricing?.input_cache_write, pricing?.input_cache_write_tiers);
+  const input = inputPrice.base;
+  const output = outputPrice.base;
   if (input === undefined || output === undefined) return undefined;
+
+  const thresholds = new Set([
+    ...inputPrice.thresholds,
+    ...outputPrice.thresholds,
+    ...cacheReadPrice.thresholds,
+    ...cacheWritePrice.thresholds,
+  ]);
+  const tiers: NonNullable<NonNullable<ExistingModel["cost"]>["tiers"]> = [];
+  let previous = {
+    input,
+    output,
+    cache_read: cacheReadPrice.base,
+    cache_write: cacheWritePrice.base,
+  };
+  for (const size of [...thresholds].sort((a, b) => a - b)) {
+    const tierInput = inputPrice.at(size);
+    const tierOutput = outputPrice.at(size);
+    if (tierInput === undefined || tierOutput === undefined) continue;
+    const current = {
+      input: tierInput,
+      output: tierOutput,
+      cache_read: cacheReadPrice.at(size),
+      cache_write: cacheWritePrice.at(size),
+    };
+    if (JSON.stringify(current) === JSON.stringify(previous)) continue;
+    tiers.push({ tier: { type: "context", size }, ...current });
+    previous = current;
+  }
+
   return {
     input,
     output,
     reasoning: existing?.reasoning,
-    cache_read: price(pricing?.input_cache_read_tiers?.[0]?.cost ?? pricing?.input_cache_read),
-    cache_write: price(pricing?.input_cache_write_tiers?.[0]?.cost ?? pricing?.input_cache_write),
-    tiers: existing?.tiers,
+    cache_read: cacheReadPrice.base,
+    cache_write: cacheWritePrice.base,
+    tiers: hasPricingTiers ? (tiers.length > 0 ? tiers : undefined) : existing?.tiers,
   };
 }
 
@@ -291,6 +349,7 @@ function sameVercelModel(current: ExistingModel, desired: SyncedModel) {
     [current.cost?.output, desiredModel.cost?.output, true],
     [current.cost?.cache_read, desiredModel.cost?.cache_read, true],
     [current.cost?.cache_write, desiredModel.cost?.cache_write, true],
+    [current.cost?.tiers, desiredModel.cost?.tiers],
     [current.limit?.context, desiredModel.limit?.context],
     [current.limit?.input, desiredModel.limit?.input],
     [current.limit?.output, desiredModel.limit?.output],
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index d0609631969..0281aac4808 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -4423,7 +4423,7 @@ test("retains Merge Gateway models missing from an API-key-scoped response", ()
   expect(mergeGateway.deleteMissing).toBe(false);
 });
 
-test("parses Vercel pricing tiers with an implicit zero minimum", () => {
+test("translates Vercel pricing tiers with an implicit zero minimum", () => {
   const [model] = vercel.parseModels({
     data: [{
       id: "openai/gpt-5.6-luna",
@@ -4440,14 +4440,36 @@ test("parses Vercel pricing tiers with an implicit zero minimum", () => {
           { cost: "0.0000001", max: 272_000 },
           { cost: "0.0000002", min: 272_000 },
         ],
+        input_tiers: [
+          { cost: "0.000001", max: 272_000 },
+          { cost: "0.000002", min: 272_000 },
+        ],
+        output_tiers: [
+          { cost: "0.000006", max: 272_000 },
+          { cost: "0.000009", min: 272_000 },
+        ],
       },
     }],
   });
 
   expect(model).toBeDefined();
-  expect(buildVercelModel(model!, undefined)).toMatchObject({
-    cost: { input: 1, output: 6, cache_read: 0.1 },
+  const synced = buildVercelModel(model!, undefined);
+  expect(synced).toMatchObject({
+    cost: {
+      input: 1,
+      output: 6,
+      cache_read: 0.1,
+      tiers: [{
+        tier: { type: "context", size: 272_000 },
+        input: 2,
+        output: 9,
+        cache_read: 0.2,
+      }],
+    },
   });
+  expect(vercel.sameModel?.({
+    cost: { input: 1, output: 6, cache_read: 0.1 },
+  }, synced)).toBe(false);
 });
 
 test("Vercel factored models inherit temperature from base metadata", () => {
diff --git a/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml b/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
index 2978e3e117c..df7e9852286 100644
--- a/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
+++ b/providers/vercel/models/alibaba/qwen-3.6-max-preview.toml
@@ -23,6 +23,13 @@ output = 7.8
 cache_read = 0.13
 cache_write = 1.625
 
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
+
 [limit]
 context = 240_000
 output = 64_000
diff --git a/providers/vercel/models/alibaba/qwen3-coder-plus.toml b/providers/vercel/models/alibaba/qwen3-coder-plus.toml
index 06f877de2eb..656712e8607 100644
--- a/providers/vercel/models/alibaba/qwen3-coder-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3-coder-plus.toml
@@ -5,5 +5,23 @@ input = 1
 output = 5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_001 }
+input = 1.8
+output = 9
+cache_read = 0.36
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 3
+output = 15
+cache_read = 0.6
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_001 }
+input = 6
+output = 60
+cache_read = 1.2
+
 [limit]
 context = 1_000_000
diff --git a/providers/vercel/models/alibaba/qwen3-coder.toml b/providers/vercel/models/alibaba/qwen3-coder.toml
index e61dd0d67dd..f5fe46e649f 100644
--- a/providers/vercel/models/alibaba/qwen3-coder.toml
+++ b/providers/vercel/models/alibaba/qwen3-coder.toml
@@ -16,6 +16,18 @@ input = 1.5
 output = 7.5
 cache_read = 0.3
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_001 }
+input = 2.7
+output = 13.5
+cache_read = 0.54
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 4.5
+output = 22.5
+cache_read = 0.9
+
 [limit]
 context = 262_144
 output = 65_536
diff --git a/providers/vercel/models/alibaba/qwen3-max-preview.toml b/providers/vercel/models/alibaba/qwen3-max-preview.toml
index 94b902c8984..69a704070b5 100644
--- a/providers/vercel/models/alibaba/qwen3-max-preview.toml
+++ b/providers/vercel/models/alibaba/qwen3-max-preview.toml
@@ -15,6 +15,18 @@ input = 1.2
 output = 6
 cache_read = 0.24
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_001 }
+input = 2.4
+output = 12
+cache_read = 0.48
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 3
+output = 15
+cache_read = 0.6
+
 [limit]
 context = 262_144
 output = 32_768
diff --git a/providers/vercel/models/alibaba/qwen3-max-thinking.toml b/providers/vercel/models/alibaba/qwen3-max-thinking.toml
index db97c26cd56..b00cd1597a9 100644
--- a/providers/vercel/models/alibaba/qwen3-max-thinking.toml
+++ b/providers/vercel/models/alibaba/qwen3-max-thinking.toml
@@ -5,17 +5,33 @@ release_date = "2026-01-23"
 last_updated = "2025-01"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "budget_tokens", min = 1, max = 81_920 }]
 temperature = true
 tool_call = true
 knowledge = "2025-01"
 open_weights = true
 
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 81_920
+
 [cost]
 input = 1.2
 output = 6
 cache_read = 0.24
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_001 }
+input = 2.4
+output = 12
+cache_read = 0.48
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 3
+output = 15
+cache_read = 0.6
+
 [limit]
 context = 256_000
 output = 65_536
diff --git a/providers/vercel/models/alibaba/qwen3-max.toml b/providers/vercel/models/alibaba/qwen3-max.toml
index 74f80d21720..556b71930db 100644
--- a/providers/vercel/models/alibaba/qwen3-max.toml
+++ b/providers/vercel/models/alibaba/qwen3-max.toml
@@ -5,5 +5,17 @@ input = 1.2
 output = 6
 cache_read = 0.24
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_001 }
+input = 2.4
+output = 12
+cache_read = 0.48
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 3
+output = 15
+cache_read = 0.6
+
 [limit]
 output = 32_768
diff --git a/providers/vercel/models/alibaba/qwen3.5-plus.toml b/providers/vercel/models/alibaba/qwen3.5-plus.toml
index fab6054e8ca..3f4c69a844a 100644
--- a/providers/vercel/models/alibaba/qwen3.5-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3.5-plus.toml
@@ -16,6 +16,13 @@ output = 2.5
 cache_read = 0.04
 cache_write = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 256_001 }
+input = 0.5
+output = 3
+cache_read = 0.05
+cache_write = 0.625
+
 [limit]
 output = 64_000
 
diff --git a/providers/vercel/models/alibaba/qwen3.6-plus.toml b/providers/vercel/models/alibaba/qwen3.6-plus.toml
index 1b29656202b..e87a6735aae 100644
--- a/providers/vercel/models/alibaba/qwen3.6-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3.6-plus.toml
@@ -15,6 +15,13 @@ output = 3
 cache_read = 0.05
 cache_write = 0.625
 
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 2
+output = 6
+cache_read = 0.2
+cache_write = 2.5
+
 [limit]
 output = 64_000
 
diff --git a/providers/vercel/models/alibaba/qwen3.7-flash.toml b/providers/vercel/models/alibaba/qwen3.7-flash.toml
index 774d5fe725d..8aaf3abadf1 100644
--- a/providers/vercel/models/alibaba/qwen3.7-flash.toml
+++ b/providers/vercel/models/alibaba/qwen3.7-flash.toml
@@ -10,6 +10,20 @@ output = 0.13
 cache_read = 0.006
 cache_write = 0.038
 
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.1
+output = 0.4
+cache_read = 0.02
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.2
+output = 0.8
+cache_read = 0.04
+cache_write = 0.25
+
 [limit]
 context = 991_000
 output = 64_000
diff --git a/providers/vercel/models/alibaba/qwen3.7-plus.toml b/providers/vercel/models/alibaba/qwen3.7-plus.toml
index 819ab65c9e3..4511b24804c 100644
--- a/providers/vercel/models/alibaba/qwen3.7-plus.toml
+++ b/providers/vercel/models/alibaba/qwen3.7-plus.toml
@@ -24,5 +24,12 @@ output = 1.6
 cache_read = 0.08
 cache_write = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 1.2
+output = 4.8
+cache_read = 0.24
+cache_write = 1.5
+
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/anthropic/claude-sonnet-4.toml b/providers/vercel/models/anthropic/claude-sonnet-4.toml
index df6a061b4b1..61e522dfd97 100644
--- a/providers/vercel/models/anthropic/claude-sonnet-4.toml
+++ b/providers/vercel/models/anthropic/claude-sonnet-4.toml
@@ -7,6 +7,13 @@ output = 15
 cache_read = 0.3
 cache_write = 3.75
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 6
+output = 22.5
+cache_read = 0.6
+cache_write = 7.5
+
 [limit]
 context = 1_000_000
 output = 8_192
diff --git a/providers/vercel/models/bytedance/seed-1.6.toml b/providers/vercel/models/bytedance/seed-1.6.toml
index c0346917f50..bc03000e407 100644
--- a/providers/vercel/models/bytedance/seed-1.6.toml
+++ b/providers/vercel/models/bytedance/seed-1.6.toml
@@ -18,6 +18,12 @@ input = 0.25
 output = 2
 cache_read = 0.05
 
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 0.5
+output = 4
+cache_read = 0.05
+
 [limit]
 context = 256_000
 output = 32_000
diff --git a/providers/vercel/models/bytedance/seed-1.8.toml b/providers/vercel/models/bytedance/seed-1.8.toml
index 1fa077191ce..7e02aaccd52 100644
--- a/providers/vercel/models/bytedance/seed-1.8.toml
+++ b/providers/vercel/models/bytedance/seed-1.8.toml
@@ -5,17 +5,29 @@ release_date = "2025-09-01"
 last_updated = "2025-10"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["minimal", "low", "medium", "high"] }]
 temperature = true
 tool_call = true
 knowledge = "2024-10"
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
 [cost]
 input = 0.25
 output = 2
 cache_read = 0.05
 
+[[cost.tiers]]
+tier = { type = "context", size = 128_001 }
+input = 0.5
+output = 4
+cache_read = 0.05
+
 [limit]
 context = 256_000
 output = 64_000
diff --git a/providers/vercel/models/google/gemini-3.1-pro-preview.toml b/providers/vercel/models/google/gemini-3.1-pro-preview.toml
index 31fde20cebe..742f9779a46 100644
--- a/providers/vercel/models/google/gemini-3.1-pro-preview.toml
+++ b/providers/vercel/models/google/gemini-3.1-pro-preview.toml
@@ -9,6 +9,12 @@ input = 2
 output = 12
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 4
+output = 18
+cache_read = 0.4
+
 [limit]
 context = 1_000_000
 output = 64_000
diff --git a/providers/vercel/models/openai/gpt-5.4-pro.toml b/providers/vercel/models/openai/gpt-5.4-pro.toml
index 802e2b01e7d..8f49caeaef4 100644
--- a/providers/vercel/models/openai/gpt-5.4-pro.toml
+++ b/providers/vercel/models/openai/gpt-5.4-pro.toml
@@ -10,5 +10,10 @@ values = ["medium", "high", "xhigh"]
 input = 30
 output = 180
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 60
+output = 270
+
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/openai/gpt-5.4.toml b/providers/vercel/models/openai/gpt-5.4.toml
index 117744507a5..e80519bdd16 100644
--- a/providers/vercel/models/openai/gpt-5.4.toml
+++ b/providers/vercel/models/openai/gpt-5.4.toml
@@ -1,9 +1,17 @@
 base_model = "openai/gpt-5.4"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 name = "GPT 5.4"
-temperature = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
 
 [cost]
 input = 2.5
 output = 15
 cache_read = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 5
+output = 22.5
+cache_read = 0.5
diff --git a/providers/vercel/models/openai/gpt-5.5-pro.toml b/providers/vercel/models/openai/gpt-5.5-pro.toml
index effd8f55a5a..b124e744fda 100644
--- a/providers/vercel/models/openai/gpt-5.5-pro.toml
+++ b/providers/vercel/models/openai/gpt-5.5-pro.toml
@@ -10,6 +10,11 @@ values = ["medium", "high", "xhigh"]
 input = 30
 output = 180
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 60
+output = 270
+
 [limit]
 context = 1_000_000
 input = 872_000
diff --git a/providers/vercel/models/openai/gpt-5.5.toml b/providers/vercel/models/openai/gpt-5.5.toml
index 4d37b9c3632..6fa3472701a 100644
--- a/providers/vercel/models/openai/gpt-5.5.toml
+++ b/providers/vercel/models/openai/gpt-5.5.toml
@@ -11,6 +11,12 @@ input = 5
 output = 30
 cache_read = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 45
+cache_read = 1
+
 [limit]
 context = 1_000_000
 input = 872_000
diff --git a/providers/vercel/models/openai/gpt-5.6-luna-fast.toml b/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
index 0677db18d76..74dd96c0e31 100644
--- a/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-luna-fast.toml
@@ -10,3 +10,10 @@ input = 0.4
 output = 2.4
 cache_read = 0.04
 cache_write = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.8
+output = 3.6
+cache_read = 0.08
+cache_write = 1
diff --git a/providers/vercel/models/openai/gpt-5.6-luna.toml b/providers/vercel/models/openai/gpt-5.6-luna.toml
index f33d0f6f9b5..e48b3985f6c 100644
--- a/providers/vercel/models/openai/gpt-5.6-luna.toml
+++ b/providers/vercel/models/openai/gpt-5.6-luna.toml
@@ -12,3 +12,10 @@ input = 0.2
 output = 1.2
 cache_read = 0.02
 cache_write = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.4
+output = 1.8
+cache_read = 0.04
+cache_write = 0.5
diff --git a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
index 007699dc75e..d989b000e02 100644
--- a/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-sol-fast.toml
@@ -10,3 +10,10 @@ input = 8
 output = 40
 cache_read = 0.8
 cache_write = 10
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 16
+output = 60
+cache_read = 1.6
+cache_write = 20
diff --git a/providers/vercel/models/openai/gpt-5.6-sol.toml b/providers/vercel/models/openai/gpt-5.6-sol.toml
index b07d7577a9a..ed5b73a9699 100644
--- a/providers/vercel/models/openai/gpt-5.6-sol.toml
+++ b/providers/vercel/models/openai/gpt-5.6-sol.toml
@@ -12,3 +12,10 @@ input = 4
 output = 20
 cache_read = 0.4
 cache_write = 5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10
diff --git a/providers/vercel/models/openai/gpt-5.6-terra-fast.toml b/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
index c5e03cad536..e2caf6406e9 100644
--- a/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
+++ b/providers/vercel/models/openai/gpt-5.6-terra-fast.toml
@@ -10,3 +10,10 @@ input = 4
 output = 24
 cache_read = 0.4
 cache_write = 5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 8
+output = 36
+cache_read = 0.8
+cache_write = 10
diff --git a/providers/vercel/models/openai/gpt-5.6-terra.toml b/providers/vercel/models/openai/gpt-5.6-terra.toml
index 7c036e01d1c..cc47098412f 100644
--- a/providers/vercel/models/openai/gpt-5.6-terra.toml
+++ b/providers/vercel/models/openai/gpt-5.6-terra.toml
@@ -12,3 +12,10 @@ input = 2
 output = 12
 cache_read = 0.2
 cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 18
+cache_read = 0.4
+cache_write = 5
diff --git a/providers/vercel/models/openai/gpt-6-astra-fast.toml b/providers/vercel/models/openai/gpt-6-astra-fast.toml
index ad67cd2908a..bbb8660be6c 100644
--- a/providers/vercel/models/openai/gpt-6-astra-fast.toml
+++ b/providers/vercel/models/openai/gpt-6-astra-fast.toml
@@ -16,4 +16,4 @@ tier = { type = "context", size = 272_001 }
 input = 40
 output = 150
 cache_read = 4
-cache_write = 25
+cache_write = 50
diff --git a/providers/vercel/models/sakana/fugu-ultra-v2.toml b/providers/vercel/models/sakana/fugu-ultra-v2.toml
index f77055d8e6e..3ada0b1935d 100644
--- a/providers/vercel/models/sakana/fugu-ultra-v2.toml
+++ b/providers/vercel/models/sakana/fugu-ultra-v2.toml
@@ -9,13 +9,22 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = [{ type = "effort", values = ["high", "xhigh", "max"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "xhigh", "max"]
 
 [cost]
 input = 5
 output = 30
 cache_read = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 10
+output = 45
+cache_read = 1
+
 [limit]
 context = 1_000_000
 output = 1_000_000
diff --git a/providers/vercel/models/sakana/fugu-ultra.toml b/providers/vercel/models/sakana/fugu-ultra.toml
index 099d441fbe6..f36e9177ff8 100644
--- a/providers/vercel/models/sakana/fugu-ultra.toml
+++ b/providers/vercel/models/sakana/fugu-ultra.toml
@@ -8,5 +8,11 @@ input = 5
 output = 30
 cache_read = 0.5
 
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 10
+output = 45
+cache_read = 1
+
 [limit]
 output = 1_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
index ed37aa69ee5..d45eab3701e 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent-beta.toml
@@ -14,6 +14,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
index 37a06048cc0..a90774a3c3f 100644
--- a/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-multi-agent.toml
@@ -14,6 +14,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-non-reasoning-beta.toml b/providers/vercel/models/spacexai/grok-4.20-non-reasoning-beta.toml
index acd189124d0..ac3eafaaecd 100644
--- a/providers/vercel/models/spacexai/grok-4.20-non-reasoning-beta.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-non-reasoning-beta.toml
@@ -13,6 +13,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.4
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-non-reasoning.toml b/providers/vercel/models/spacexai/grok-4.20-non-reasoning.toml
index 468d77b85b2..0e9a2baf86f 100644
--- a/providers/vercel/models/spacexai/grok-4.20-non-reasoning.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-non-reasoning.toml
@@ -13,6 +13,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-reasoning-beta.toml b/providers/vercel/models/spacexai/grok-4.20-reasoning-beta.toml
index 19522d5f419..691eded5322 100644
--- a/providers/vercel/models/spacexai/grok-4.20-reasoning-beta.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-reasoning-beta.toml
@@ -14,6 +14,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.20-reasoning.toml b/providers/vercel/models/spacexai/grok-4.20-reasoning.toml
index 8fb2843f07d..e6c7e7ae17b 100644
--- a/providers/vercel/models/spacexai/grok-4.20-reasoning.toml
+++ b/providers/vercel/models/spacexai/grok-4.20-reasoning.toml
@@ -14,6 +14,12 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 context = 2_000_000
 output = 2_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.3.toml b/providers/vercel/models/spacexai/grok-4.3.toml
index d379c871d7f..1ce53c49412 100644
--- a/providers/vercel/models/spacexai/grok-4.3.toml
+++ b/providers/vercel/models/spacexai/grok-4.3.toml
@@ -1,5 +1,6 @@
 base_model = "xai/grok-4.3"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
@@ -9,5 +10,11 @@ input = 1.25
 output = 2.5
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.5
+output = 5
+cache_read = 0.4
+
 [limit]
 output = 1_000_000
diff --git a/providers/vercel/models/spacexai/grok-4.5.toml b/providers/vercel/models/spacexai/grok-4.5.toml
index 2eb4ebf0d6e..ec9457541d2 100644
--- a/providers/vercel/models/spacexai/grok-4.5.toml
+++ b/providers/vercel/models/spacexai/grok-4.5.toml
@@ -1,5 +1,6 @@
 base_model = "xai/grok-4.5"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
@@ -9,5 +10,11 @@ input = 2
 output = 6
 cache_read = 0.3
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 4
+output = 12
+cache_read = 0.6
+
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/spacexai/grok-4.6.toml b/providers/vercel/models/spacexai/grok-4.6.toml
index 713ac05e4b3..58cede40b86 100644
--- a/providers/vercel/models/spacexai/grok-4.6.toml
+++ b/providers/vercel/models/spacexai/grok-4.6.toml
@@ -1,5 +1,6 @@
 base_model = "xai/grok-4.6"
 description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high"]
@@ -8,3 +9,9 @@ values = ["low", "medium", "high"]
 input = 2
 output = 6
 cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 4
+output = 12
+cache_read = 1
diff --git a/providers/vercel/models/spacexai/grok-4.7.toml b/providers/vercel/models/spacexai/grok-4.7.toml
new file mode 100644
index 00000000000..defc4659f57
--- /dev/null
+++ b/providers/vercel/models/spacexai/grok-4.7.toml
@@ -0,0 +1,13 @@
+base_model = "xai/grok-4.7"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 1.2
+output = 3.6
+cache_read = 0.3
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2.4
+output = 7.2
+cache_read = 0.6
diff --git a/providers/vercel/models/spacexai/grok-build-0.1.toml b/providers/vercel/models/spacexai/grok-build-0.1.toml
index dfcdb207ca7..ec232622716 100644
--- a/providers/vercel/models/spacexai/grok-build-0.1.toml
+++ b/providers/vercel/models/spacexai/grok-build-0.1.toml
@@ -7,5 +7,11 @@ input = 1
 output = 2
 cache_read = 0.2
 
+[[cost.tiers]]
+tier = { type = "context", size = 200_001 }
+input = 2
+output = 4
+cache_read = 0.4
+
 [modalities]
 input = ["text", "image"]
diff --git a/providers/vercel/models/stepfun/step-5-preview.toml b/providers/vercel/models/stepfun/step-5-preview.toml
index 8db9a1d1622..ce090c9a924 100644
--- a/providers/vercel/models/stepfun/step-5-preview.toml
+++ b/providers/vercel/models/stepfun/step-5-preview.toml
@@ -1,22 +1,12 @@
-name = "Step 5 Preview"
+base_model = "stepfun/step-5-preview"
 description = "StepFun flash model for efficient multimodal reasoning, coding, and tool use"
-family = "step"
-release_date = "2026-09-20"
-last_updated = "2026-09-20"
-attachment = true
 reasoning = false
 tool_call = false
-open_weights = false
 
 [cost]
 input = 1
 output = 2.7
 cache_read = 0.05
 
-[limit]
-context = 1_000_000
-output = 1_000_000
-
 [modalities]
 input = ["text", "image"]
-output = ["text"]

From a2217f3a6b4a30e33e74ba10169f83fda5af76fb Mon Sep 17 00:00:00 2001
From: Johnny Chadda <johnny@chadda.se>
Date: Mon, 21 Sep 2026 18:53:49 +0200
Subject: [PATCH 1806/2012] Use pooled model IDs for Opper and refresh catalog
 (#6542)

* Use pooled model IDs for Opper and refresh catalog

* Add verified Muse max effort and document pool overrides
---
 .../models/anthropic/claude-haiku-4-5.toml    | 13 ---------
 .../models/anthropic/claude-opus-4-7.toml     | 15 ----------
 .../opper/models/anthropic/claude-opus-5.toml | 15 ----------
 .../models/anthropic/claude-sonnet-4-5.toml   | 13 ---------
 .../models/anthropic/claude-sonnet-4-6.toml   | 15 ----------
 .../models/anthropic/claude-sonnet-5.toml     | 15 ----------
 providers/opper/models/claude-fable-5-1.toml  | 15 ++++++++++
 .../{anthropic => }/claude-fable-5.toml       |  6 ++--
 providers/opper/models/claude-haiku-4-5.toml  | 18 ++++++++++++
 .../{anthropic => }/claude-opus-4-5.toml      |  6 ++--
 .../{anthropic => }/claude-opus-4-6.toml      |  6 ++--
 providers/opper/models/claude-opus-4-7.toml   | 18 ++++++++++++
 .../{anthropic => }/claude-opus-4-8.toml      |  6 ++--
 providers/opper/models/claude-opus-5.toml     | 15 ++++++++++
 providers/opper/models/claude-sonnet-4-5.toml | 23 +++++++++++++++
 providers/opper/models/claude-sonnet-4-6.toml | 15 ++++++++++
 providers/opper/models/claude-sonnet-5.toml   | 18 ++++++++++++
 providers/opper/models/deepseek-v4-flash.toml | 12 ++++++++
 providers/opper/models/deepseek-v4-pro.toml   | 15 ++++++++++
 providers/opper/models/devstral-2512.toml     | 11 ++++++++
 .../opper/models/gemini-3-flash-preview.toml  | 13 +++++++++
 .../opper/models/gemini-3.1-pro-preview.toml  | 20 +++++++++++++
 .../opper/models/gemini-3.5-flash-lite.toml   | 13 +++++++++
 providers/opper/models/gemini-3.5-flash.toml  | 13 +++++++++
 providers/opper/models/gemini-3.7-flash.toml  | 13 +++++++++
 providers/opper/models/gemini-3.8-flash.toml  | 16 +++++++++++
 .../models/gemini/gemini-3-flash-preview.toml | 10 -------
 .../models/gemini/gemini-3.1-pro-preview.toml | 16 -----------
 .../models/gemini/gemini-3.5-flash-lite.toml  | 10 -------
 .../opper/models/gemini/gemini-3.5-flash.toml | 10 -------
 providers/opper/models/gemma-4-31b-it.toml    | 16 +++++++++++
 providers/opper/models/glm-5.2.toml           | 12 ++++++++
 providers/opper/models/glm-5.3-flash.toml     | 21 ++++++++++++++
 providers/opper/models/glm-5.3.toml           | 14 ++++++++++
 .../opper/models/gpt-5.3-chat-latest.toml     |  9 ++++++
 providers/opper/models/gpt-5.3-codex.toml     | 18 ++++++++++++
 providers/opper/models/gpt-5.4-mini.toml      | 14 ++++++++++
 providers/opper/models/gpt-5.4-nano.toml      | 14 ++++++++++
 providers/opper/models/gpt-5.4-pro.toml       | 18 ++++++++++++
 providers/opper/models/gpt-5.4.toml           | 20 +++++++++++++
 providers/opper/models/gpt-5.5-pro.toml       | 18 ++++++++++++
 providers/opper/models/gpt-5.5.toml           | 20 +++++++++++++
 providers/opper/models/gpt-5.6-luna.toml      | 25 +++++++++++++++++
 providers/opper/models/gpt-5.6-sol.toml       | 25 +++++++++++++++++
 providers/opper/models/gpt-5.6-terra.toml     | 25 +++++++++++++++++
 providers/opper/models/gpt-6-astra.toml       | 25 +++++++++++++++++
 providers/opper/models/gpt-oss-120b.toml      | 16 +++++++++++
 providers/opper/models/gpt-oss-20b.toml       | 16 +++++++++++
 providers/opper/models/grok-4.3.toml          | 18 ++++++++++++
 providers/opper/models/grok-4.5.toml          | 15 ++++++++++
 providers/opper/models/grok-4.6.toml          | 20 +++++++++++++
 providers/opper/models/grok-build-0.1.toml    | 15 ++++++++++
 providers/opper/models/kimi-k3.toml           | 16 +++++++++++
 providers/opper/models/minimax-m2.7.toml      | 16 +++++++++++
 providers/opper/models/minimax-m3.toml        | 28 +++++++++++++++++++
 providers/opper/models/minimax/m3.toml        | 24 ----------------
 .../opper/models/mistral-large-2512.toml      | 11 ++++++++
 .../opper/models/mistral-small-2603.toml      | 16 +++++++++++
 .../opper/models/mistral/devstral-2512.toml   |  5 ----
 .../models/mistral/mistral-large-2512.toml    |  5 ----
 .../models/mistral/mistral-small-2603.toml    |  9 ------
 providers/opper/models/moonshot/kimi-k3.toml  | 13 ---------
 .../models/{meta => }/muse-spark-1.2.toml     |  5 ++++
 providers/opper/models/muse-spark-1.3.toml    | 15 ++++++++++
 .../models/openai/gpt-5.3-chat-latest.toml    |  6 ----
 .../opper/models/openai/gpt-5.3-codex.toml    |  8 ------
 .../opper/models/openai/gpt-5.4-mini.toml     |  8 ------
 .../opper/models/openai/gpt-5.4-nano.toml     |  8 ------
 .../opper/models/openai/gpt-5.4-pro.toml      | 12 --------
 providers/opper/models/openai/gpt-5.4.toml    | 14 ----------
 .../opper/models/openai/gpt-5.5-pro.toml      | 12 --------
 providers/opper/models/openai/gpt-5.5.toml    | 14 ----------
 .../opper/models/openai/gpt-5.6-luna.toml     | 16 -----------
 .../opper/models/openai/gpt-5.6-sol.toml      | 16 -----------
 .../opper/models/openai/gpt-5.6-terra.toml    | 16 -----------
 .../opper/models/perplexity/sonar-pro.toml    |  5 ----
 .../perplexity/sonar-reasoning-pro.toml       |  7 -----
 providers/opper/models/perplexity/sonar.toml  |  5 ----
 providers/opper/models/qwen3-coder-next.toml  |  6 ++++
 providers/opper/models/qwen3.6-35b-a3b.toml   | 15 ++++++++++
 providers/opper/models/qwen3.8-2.4t-a95b.toml | 13 +++++++++
 providers/opper/models/qwen3.8-27b.toml       | 17 +++++++++++
 providers/opper/models/qwen3.8-max.toml       | 20 +++++++++++++
 providers/opper/models/sonar-pro.toml         | 14 ++++++++++
 .../opper/models/sonar-reasoning-pro.toml     | 17 +++++++++++
 providers/opper/models/sonar.toml             |  7 +++++
 .../models/vertexai/gemini-3.7-flash-eu.toml  | 11 --------
 .../models/vertexai/gemini-3.7-flash.toml     | 10 -------
 providers/opper/models/xai/grok-4.3.toml      | 19 -------------
 providers/opper/models/xai/grok-4.5.toml      | 19 -------------
 providers/opper/models/xai/grok-4.6.toml      | 19 -------------
 .../opper/models/xai/grok-build-0.1.toml      | 17 -----------
 providers/opper/provider.toml                 | 23 +++++++++------
 93 files changed, 885 insertions(+), 450 deletions(-)
 delete mode 100644 providers/opper/models/anthropic/claude-haiku-4-5.toml
 delete mode 100644 providers/opper/models/anthropic/claude-opus-4-7.toml
 delete mode 100644 providers/opper/models/anthropic/claude-opus-5.toml
 delete mode 100644 providers/opper/models/anthropic/claude-sonnet-4-5.toml
 delete mode 100644 providers/opper/models/anthropic/claude-sonnet-4-6.toml
 delete mode 100644 providers/opper/models/anthropic/claude-sonnet-5.toml
 create mode 100644 providers/opper/models/claude-fable-5-1.toml
 rename providers/opper/models/{anthropic => }/claude-fable-5.toml (50%)
 create mode 100644 providers/opper/models/claude-haiku-4-5.toml
 rename providers/opper/models/{anthropic => }/claude-opus-4-5.toml (50%)
 rename providers/opper/models/{anthropic => }/claude-opus-4-6.toml (50%)
 create mode 100644 providers/opper/models/claude-opus-4-7.toml
 rename providers/opper/models/{anthropic => }/claude-opus-4-8.toml (50%)
 create mode 100644 providers/opper/models/claude-opus-5.toml
 create mode 100644 providers/opper/models/claude-sonnet-4-5.toml
 create mode 100644 providers/opper/models/claude-sonnet-4-6.toml
 create mode 100644 providers/opper/models/claude-sonnet-5.toml
 create mode 100644 providers/opper/models/deepseek-v4-flash.toml
 create mode 100644 providers/opper/models/deepseek-v4-pro.toml
 create mode 100644 providers/opper/models/devstral-2512.toml
 create mode 100644 providers/opper/models/gemini-3-flash-preview.toml
 create mode 100644 providers/opper/models/gemini-3.1-pro-preview.toml
 create mode 100644 providers/opper/models/gemini-3.5-flash-lite.toml
 create mode 100644 providers/opper/models/gemini-3.5-flash.toml
 create mode 100644 providers/opper/models/gemini-3.7-flash.toml
 create mode 100644 providers/opper/models/gemini-3.8-flash.toml
 delete mode 100644 providers/opper/models/gemini/gemini-3-flash-preview.toml
 delete mode 100644 providers/opper/models/gemini/gemini-3.1-pro-preview.toml
 delete mode 100644 providers/opper/models/gemini/gemini-3.5-flash-lite.toml
 delete mode 100644 providers/opper/models/gemini/gemini-3.5-flash.toml
 create mode 100644 providers/opper/models/gemma-4-31b-it.toml
 create mode 100644 providers/opper/models/glm-5.2.toml
 create mode 100644 providers/opper/models/glm-5.3-flash.toml
 create mode 100644 providers/opper/models/glm-5.3.toml
 create mode 100644 providers/opper/models/gpt-5.3-chat-latest.toml
 create mode 100644 providers/opper/models/gpt-5.3-codex.toml
 create mode 100644 providers/opper/models/gpt-5.4-mini.toml
 create mode 100644 providers/opper/models/gpt-5.4-nano.toml
 create mode 100644 providers/opper/models/gpt-5.4-pro.toml
 create mode 100644 providers/opper/models/gpt-5.4.toml
 create mode 100644 providers/opper/models/gpt-5.5-pro.toml
 create mode 100644 providers/opper/models/gpt-5.5.toml
 create mode 100644 providers/opper/models/gpt-5.6-luna.toml
 create mode 100644 providers/opper/models/gpt-5.6-sol.toml
 create mode 100644 providers/opper/models/gpt-5.6-terra.toml
 create mode 100644 providers/opper/models/gpt-6-astra.toml
 create mode 100644 providers/opper/models/gpt-oss-120b.toml
 create mode 100644 providers/opper/models/gpt-oss-20b.toml
 create mode 100644 providers/opper/models/grok-4.3.toml
 create mode 100644 providers/opper/models/grok-4.5.toml
 create mode 100644 providers/opper/models/grok-4.6.toml
 create mode 100644 providers/opper/models/grok-build-0.1.toml
 create mode 100644 providers/opper/models/kimi-k3.toml
 create mode 100644 providers/opper/models/minimax-m2.7.toml
 create mode 100644 providers/opper/models/minimax-m3.toml
 delete mode 100644 providers/opper/models/minimax/m3.toml
 create mode 100644 providers/opper/models/mistral-large-2512.toml
 create mode 100644 providers/opper/models/mistral-small-2603.toml
 delete mode 100644 providers/opper/models/mistral/devstral-2512.toml
 delete mode 100644 providers/opper/models/mistral/mistral-large-2512.toml
 delete mode 100644 providers/opper/models/mistral/mistral-small-2603.toml
 delete mode 100644 providers/opper/models/moonshot/kimi-k3.toml
 rename providers/opper/models/{meta => }/muse-spark-1.2.toml (51%)
 create mode 100644 providers/opper/models/muse-spark-1.3.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.3-chat-latest.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.3-codex.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.4-mini.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.4-nano.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.4-pro.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.4.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.5-pro.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.5.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.6-luna.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.6-sol.toml
 delete mode 100644 providers/opper/models/openai/gpt-5.6-terra.toml
 delete mode 100644 providers/opper/models/perplexity/sonar-pro.toml
 delete mode 100644 providers/opper/models/perplexity/sonar-reasoning-pro.toml
 delete mode 100644 providers/opper/models/perplexity/sonar.toml
 create mode 100644 providers/opper/models/qwen3-coder-next.toml
 create mode 100644 providers/opper/models/qwen3.6-35b-a3b.toml
 create mode 100644 providers/opper/models/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/opper/models/qwen3.8-27b.toml
 create mode 100644 providers/opper/models/qwen3.8-max.toml
 create mode 100644 providers/opper/models/sonar-pro.toml
 create mode 100644 providers/opper/models/sonar-reasoning-pro.toml
 create mode 100644 providers/opper/models/sonar.toml
 delete mode 100644 providers/opper/models/vertexai/gemini-3.7-flash-eu.toml
 delete mode 100644 providers/opper/models/vertexai/gemini-3.7-flash.toml
 delete mode 100644 providers/opper/models/xai/grok-4.3.toml
 delete mode 100644 providers/opper/models/xai/grok-4.5.toml
 delete mode 100644 providers/opper/models/xai/grok-4.6.toml
 delete mode 100644 providers/opper/models/xai/grok-build-0.1.toml

diff --git a/providers/opper/models/anthropic/claude-haiku-4-5.toml b/providers/opper/models/anthropic/claude-haiku-4-5.toml
deleted file mode 100644
index 63e9f09896d..00000000000
--- a/providers/opper/models/anthropic/claude-haiku-4-5.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-haiku-4-5"
-structured_output = true
-
-reasoning_options = []
-
-[cost]
-input = 1
-output = 5
-cache_read = 0.1
-cache_write = 1.25
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-opus-4-7.toml b/providers/opper/models/anthropic/claude-opus-4-7.toml
deleted file mode 100644
index d4f86815775..00000000000
--- a/providers/opper/models/anthropic/claude-opus-4-7.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "anthropic/claude-opus-4-7"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-opus-5.toml b/providers/opper/models/anthropic/claude-opus-5.toml
deleted file mode 100644
index 634c2f987e6..00000000000
--- a/providers/opper/models/anthropic/claude-opus-5.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "anthropic/claude-opus-5"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-sonnet-4-5.toml b/providers/opper/models/anthropic/claude-sonnet-4-5.toml
deleted file mode 100644
index fdc69c21166..00000000000
--- a/providers/opper/models/anthropic/claude-sonnet-4-5.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-5"
-structured_output = true
-
-reasoning_options = []
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-sonnet-4-6.toml b/providers/opper/models/anthropic/claude-sonnet-4-6.toml
deleted file mode 100644
index aeed8195b8f..00000000000
--- a/providers/opper/models/anthropic/claude-sonnet-4-6.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "anthropic/claude-sonnet-4-6"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "max"]
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-cache_write = 3.75
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-sonnet-5.toml b/providers/opper/models/anthropic/claude-sonnet-5.toml
deleted file mode 100644
index c49fbcbdf8a..00000000000
--- a/providers/opper/models/anthropic/claude-sonnet-5.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "anthropic/claude-sonnet-5"
-structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
-
-[cost]
-input = 2
-output = 10
-cache_read = 0.2
-cache_write = 2.5
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/claude-fable-5-1.toml b/providers/opper/models/claude-fable-5-1.toml
new file mode 100644
index 00000000000..a17d2fb8c29
--- /dev/null
+++ b/providers/opper/models/claude-fable-5-1.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-fable-5.toml b/providers/opper/models/claude-fable-5.toml
similarity index 50%
rename from providers/opper/models/anthropic/claude-fable-5.toml
rename to providers/opper/models/claude-fable-5.toml
index 78ec64e8c7f..2bd15839fc0 100644
--- a/providers/opper/models/anthropic/claude-fable-5.toml
+++ b/providers/opper/models/claude-fable-5.toml
@@ -1,9 +1,9 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
 base_model = "anthropic/claude-fable-5"
 structured_output = true
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+reasoning_options = []
 
 [cost]
 input = 10
diff --git a/providers/opper/models/claude-haiku-4-5.toml b/providers/opper/models/claude-haiku-4-5.toml
new file mode 100644
index 00000000000..79bd6828653
--- /dev/null
+++ b/providers/opper/models/claude-haiku-4-5.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-haiku-4-5"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 1.1
+output = 5.5
+cache_read = 0.11
+cache_write = 1.375
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-opus-4-5.toml b/providers/opper/models/claude-opus-4-5.toml
similarity index 50%
rename from providers/opper/models/anthropic/claude-opus-4-5.toml
rename to providers/opper/models/claude-opus-4-5.toml
index 45ac31901cf..fe7f973c895 100644
--- a/providers/opper/models/anthropic/claude-opus-4-5.toml
+++ b/providers/opper/models/claude-opus-4-5.toml
@@ -1,9 +1,9 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
 base_model = "anthropic/claude-opus-4-5"
 structured_output = true
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+reasoning_options = []
 
 [cost]
 input = 5
diff --git a/providers/opper/models/anthropic/claude-opus-4-6.toml b/providers/opper/models/claude-opus-4-6.toml
similarity index 50%
rename from providers/opper/models/anthropic/claude-opus-4-6.toml
rename to providers/opper/models/claude-opus-4-6.toml
index dd56adf14b9..548c9efc22f 100644
--- a/providers/opper/models/anthropic/claude-opus-4-6.toml
+++ b/providers/opper/models/claude-opus-4-6.toml
@@ -1,9 +1,9 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
 base_model = "anthropic/claude-opus-4-6"
 structured_output = true
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "max"]
+reasoning_options = []
 
 [cost]
 input = 5
diff --git a/providers/opper/models/claude-opus-4-7.toml b/providers/opper/models/claude-opus-4-7.toml
new file mode 100644
index 00000000000..8b222c7b3c4
--- /dev/null
+++ b/providers/opper/models/claude-opus-4-7.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-opus-4-7"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
+
+[limit]
+output = 64000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/anthropic/claude-opus-4-8.toml b/providers/opper/models/claude-opus-4-8.toml
similarity index 50%
rename from providers/opper/models/anthropic/claude-opus-4-8.toml
rename to providers/opper/models/claude-opus-4-8.toml
index 83e1e7059c1..0d7904267ce 100644
--- a/providers/opper/models/anthropic/claude-opus-4-8.toml
+++ b/providers/opper/models/claude-opus-4-8.toml
@@ -1,9 +1,9 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
 base_model = "anthropic/claude-opus-4-8"
 structured_output = true
 
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+reasoning_options = []
 
 [cost]
 input = 5
diff --git a/providers/opper/models/claude-opus-5.toml b/providers/opper/models/claude-opus-5.toml
new file mode 100644
index 00000000000..052e3ac7cd9
--- /dev/null
+++ b/providers/opper/models/claude-opus-5.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-opus-5"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/claude-sonnet-4-5.toml b/providers/opper/models/claude-sonnet-4-5.toml
new file mode 100644
index 00000000000..953685007ad
--- /dev/null
+++ b/providers/opper/models/claude-sonnet-4-5.toml
@@ -0,0 +1,23 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "anthropic/claude-sonnet-4-5"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 3.3
+output = 16.5
+cache_read = 0.33
+cache_write = 4.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 200000 }
+input = 6.6
+output = 24.75
+cache_read = 0.66
+cache_write = 8.25
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/claude-sonnet-4-6.toml b/providers/opper/models/claude-sonnet-4-6.toml
new file mode 100644
index 00000000000..8efa496556f
--- /dev/null
+++ b/providers/opper/models/claude-sonnet-4-6.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-sonnet-4-6"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 3.3
+output = 16.5
+cache_read = 0.33
+cache_write = 4.125
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/claude-sonnet-5.toml b/providers/opper/models/claude-sonnet-5.toml
new file mode 100644
index 00000000000..3b91951e85f
--- /dev/null
+++ b/providers/opper/models/claude-sonnet-5.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "anthropic/claude-sonnet-5"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 2.2
+output = 11
+cache_read = 0.22
+cache_write = 2.75
+
+[limit]
+output = 64000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/deepseek-v4-flash.toml b/providers/opper/models/deepseek-v4-flash.toml
new file mode 100644
index 00000000000..b332afbbd8d
--- /dev/null
+++ b/providers/opper/models/deepseek-v4-flash.toml
@@ -0,0 +1,12 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "deepseek/deepseek-v4-flash"
+
+reasoning_options = []
+
+[cost]
+input = 0.25
+output = 0.66
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/deepseek-v4-pro.toml b/providers/opper/models/deepseek-v4-pro.toml
new file mode 100644
index 00000000000..fb6c5faa0c1
--- /dev/null
+++ b/providers/opper/models/deepseek-v4-pro.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "deepseek/deepseek-v4-pro"
+
+reasoning_options = []
+
+[cost]
+input = 1.78812
+output = 3.57624
+
+[limit]
+output = 65536
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/devstral-2512.toml b/providers/opper/models/devstral-2512.toml
new file mode 100644
index 00000000000..d3595c76c26
--- /dev/null
+++ b/providers/opper/models/devstral-2512.toml
@@ -0,0 +1,11 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+base_model = "mistral/devstral-2512"
+structured_output = true
+
+[cost]
+input = 0.4
+output = 2
+
+[limit]
+context = 256000
+output = 8192
diff --git a/providers/opper/models/gemini-3-flash-preview.toml b/providers/opper/models/gemini-3-flash-preview.toml
new file mode 100644
index 00000000000..0e6941f334a
--- /dev/null
+++ b/providers/opper/models/gemini-3-flash-preview.toml
@@ -0,0 +1,13 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemini-3-flash-preview"
+
+reasoning_options = []
+
+[cost]
+input = 0.5
+output = 3
+cache_read = 0.05
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini-3.1-pro-preview.toml b/providers/opper/models/gemini-3.1-pro-preview.toml
new file mode 100644
index 00000000000..e25eda06eb1
--- /dev/null
+++ b/providers/opper/models/gemini-3.1-pro-preview.toml
@@ -0,0 +1,20 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "google/gemini-3.1-pro-preview"
+
+reasoning_options = []
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200000 }
+input = 4
+output = 18
+cache_read = 0.4
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini-3.5-flash-lite.toml b/providers/opper/models/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..0f10955993d
--- /dev/null
+++ b/providers/opper/models/gemini-3.5-flash-lite.toml
@@ -0,0 +1,13 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemini-3.5-flash-lite"
+
+reasoning_options = []
+
+[cost]
+input = 0.3
+output = 2.5
+cache_read = 0.03
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini-3.5-flash.toml b/providers/opper/models/gemini-3.5-flash.toml
new file mode 100644
index 00000000000..b597560116a
--- /dev/null
+++ b/providers/opper/models/gemini-3.5-flash.toml
@@ -0,0 +1,13 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemini-3.5-flash"
+
+reasoning_options = []
+
+[cost]
+input = 1.5
+output = 9
+cache_read = 0.15
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini-3.7-flash.toml b/providers/opper/models/gemini-3.7-flash.toml
new file mode 100644
index 00000000000..514a0c8cd9f
--- /dev/null
+++ b/providers/opper/models/gemini-3.7-flash.toml
@@ -0,0 +1,13 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemini-3.7-flash"
+
+reasoning_options = []
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini-3.8-flash.toml b/providers/opper/models/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..b7dfba7907c
--- /dev/null
+++ b/providers/opper/models/gemini-3.8-flash.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# All five members have params but omit params.temperature; Opper drops sampling controls
+# for such cards before dispatch, so temperature is accepted but not respected.
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemini-3.8-flash"
+temperature = false
+
+reasoning_options = []
+
+[cost]
+input = 0.825
+output = 4.125
+cache_read = 0.0825
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gemini/gemini-3-flash-preview.toml b/providers/opper/models/gemini/gemini-3-flash-preview.toml
deleted file mode 100644
index b5b4efd521b..00000000000
--- a/providers/opper/models/gemini/gemini-3-flash-preview.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "google/gemini-3-flash-preview"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 0.5
-output = 3
-cache_read = 0.05
diff --git a/providers/opper/models/gemini/gemini-3.1-pro-preview.toml b/providers/opper/models/gemini/gemini-3.1-pro-preview.toml
deleted file mode 100644
index 8cb3844011e..00000000000
--- a/providers/opper/models/gemini/gemini-3.1-pro-preview.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "google/gemini-3.1-pro-preview"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 2
-output = 12
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { size = 200_000 }
-input = 4.00
-output = 18.00
-cache_read = 0.40
diff --git a/providers/opper/models/gemini/gemini-3.5-flash-lite.toml b/providers/opper/models/gemini/gemini-3.5-flash-lite.toml
deleted file mode 100644
index e5452932870..00000000000
--- a/providers/opper/models/gemini/gemini-3.5-flash-lite.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "google/gemini-3.5-flash-lite"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 0.3
-output = 2.5
-cache_read = 0.03
diff --git a/providers/opper/models/gemini/gemini-3.5-flash.toml b/providers/opper/models/gemini/gemini-3.5-flash.toml
deleted file mode 100644
index aa9842db8a7..00000000000
--- a/providers/opper/models/gemini/gemini-3.5-flash.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "google/gemini-3.5-flash"
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high"]
-
-[cost]
-input = 1.5
-output = 9
-cache_read = 0.15
diff --git a/providers/opper/models/gemma-4-31b-it.toml b/providers/opper/models/gemma-4-31b-it.toml
new file mode 100644
index 00000000000..8ee52039c78
--- /dev/null
+++ b/providers/opper/models/gemma-4-31b-it.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "google/gemma-4-31b-it"
+
+reasoning_options = []
+
+[cost]
+input = 0.46488
+output = 2.44062
+
+[limit]
+context = 256000
+output = 8192
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/glm-5.2.toml b/providers/opper/models/glm-5.2.toml
new file mode 100644
index 00000000000..bc1044a1917
--- /dev/null
+++ b/providers/opper/models/glm-5.2.toml
@@ -0,0 +1,12 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "zhipuai/glm-5.2"
+
+reasoning_options = []
+
+[cost]
+input = 1.62708
+output = 5.811
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/glm-5.3-flash.toml b/providers/opper/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..c7db468e1b7
--- /dev/null
+++ b/providers/opper/models/glm-5.3-flash.toml
@@ -0,0 +1,21 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "zhipuai/glm-5.3-flash"
+attachment = false
+structured_output = false
+
+reasoning_options = []
+
+[cost]
+input = 0.2
+output = 0.5
+cache_read = 0.07
+
+[limit]
+output = 128000
+
+[modalities]
+input = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/glm-5.3.toml b/providers/opper/models/glm-5.3.toml
new file mode 100644
index 00000000000..b6c8534060c
--- /dev/null
+++ b/providers/opper/models/glm-5.3.toml
@@ -0,0 +1,14 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: members do not share a supported effort level.
+base_model = "zhipuai/glm-5.3"
+structured_output = false
+
+reasoning_options = []
+
+[cost]
+input = 1.75
+output = 4.6488
+cache_read = 0.44
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.3-chat-latest.toml b/providers/opper/models/gpt-5.3-chat-latest.toml
new file mode 100644
index 00000000000..7568b754041
--- /dev/null
+++ b/providers/opper/models/gpt-5.3-chat-latest.toml
@@ -0,0 +1,9 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Live pool call on 2026-09-08 reports this upstream model has been deprecated.
+base_model = "openai/gpt-5.3-chat-latest"
+status = "deprecated"
+
+[cost]
+input = 1.75
+output = 14
+cache_read = 0.175
diff --git a/providers/opper/models/gpt-5.3-codex.toml b/providers/opper/models/gpt-5.3-codex.toml
new file mode 100644
index 00000000000..ea0de06a8c5
--- /dev/null
+++ b/providers/opper/models/gpt-5.3-codex.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; OpenAI native levels on all pool members.
+# Catalog context equals the lab input cap; retain the lab total context window.
+base_model = "openai/gpt-5.3-codex"
+temperature = false
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 1.75
+output = 14
+cache_read = 0.175
+
+[modalities]
+input = ["text", "pdf"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.4-mini.toml b/providers/opper/models/gpt-5.4-mini.toml
new file mode 100644
index 00000000000..5daca15ba01
--- /dev/null
+++ b/providers/opper/models/gpt-5.4-mini.toml
@@ -0,0 +1,14 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Native OpenAI effort levels verified through this single-member pool on 2026-09-08.
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+base_model = "openai/gpt-5.4-mini"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.75
+output = 4.5
+cache_read = 0.075
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.4-nano.toml b/providers/opper/models/gpt-5.4-nano.toml
new file mode 100644
index 00000000000..dbbf8bf940a
--- /dev/null
+++ b/providers/opper/models/gpt-5.4-nano.toml
@@ -0,0 +1,14 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; OpenAI native levels on all pool members.
+base_model = "openai/gpt-5.4-nano"
+temperature = false
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.2
+output = 1.25
+cache_read = 0.02
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.4-pro.toml b/providers/opper/models/gpt-5.4-pro.toml
new file mode 100644
index 00000000000..03bb664a101
--- /dev/null
+++ b/providers/opper/models/gpt-5.4-pro.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.4-pro"
+
+reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
+
+[cost]
+input = 30
+output = 180
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 60
+output = 270
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.4.toml b/providers/opper/models/gpt-5.4.toml
new file mode 100644
index 00000000000..b83170741b4
--- /dev/null
+++ b/providers/opper/models/gpt-5.4.toml
@@ -0,0 +1,20 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.4"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2.5
+output = 15
+cache_read = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 5
+output = 22.5
+cache_read = 0.5
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.5-pro.toml b/providers/opper/models/gpt-5.5-pro.toml
new file mode 100644
index 00000000000..254bf5b66c9
--- /dev/null
+++ b/providers/opper/models/gpt-5.5-pro.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Native OpenAI effort levels verified through this single-member pool on 2026-09-08.
+# All pool members publish flat rates: no pricing.thresholds or input_surcharge_threshold_tokens.
+# Opper billing uses this schedule; native context surcharges do not apply to these pools.
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+base_model = "openai/gpt-5.5-pro"
+
+reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
+
+[cost]
+input = 30
+output = 180
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.5.toml b/providers/opper/models/gpt-5.5.toml
new file mode 100644
index 00000000000..b05f12abb7f
--- /dev/null
+++ b/providers/opper/models/gpt-5.5.toml
@@ -0,0 +1,20 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.5"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 10
+output = 45
+cache_read = 1
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.6-luna.toml b/providers/opper/models/gpt-5.6-luna.toml
new file mode 100644
index 00000000000..254ff2a8555
--- /dev/null
+++ b/providers/opper/models/gpt-5.6-luna.toml
@@ -0,0 +1,25 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.6-luna"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.2
+output = 1.2
+cache_read = 0.02
+cache_write = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 0.4
+output = 1.8
+cache_read = 0.04
+cache_write = 0.5
+
+[limit]
+context = 1000000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.6-sol.toml b/providers/opper/models/gpt-5.6-sol.toml
new file mode 100644
index 00000000000..e7b4f94f5f8
--- /dev/null
+++ b/providers/opper/models/gpt-5.6-sol.toml
@@ -0,0 +1,25 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.6-sol"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 5
+output = 30
+cache_read = 0.5
+cache_write = 6.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 10
+output = 45
+cache_read = 1
+cache_write = 12.5
+
+[limit]
+context = 1000000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-5.6-terra.toml b/providers/opper/models/gpt-5.6-terra.toml
new file mode 100644
index 00000000000..209fe14cd29
--- /dev/null
+++ b/providers/opper/models/gpt-5.6-terra.toml
@@ -0,0 +1,25 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-5.6-terra"
+
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 4
+output = 18
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+context = 1000000
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-6-astra.toml b/providers/opper/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..62f6292c6cb
--- /dev/null
+++ b/providers/opper/models/gpt-6-astra.toml
@@ -0,0 +1,25 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "openai/gpt-6-astra"
+
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272000 }
+input = 20
+output = 75
+cache_read = 2
+cache_write = 25
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-oss-120b.toml b/providers/opper/models/gpt-oss-120b.toml
new file mode 100644
index 00000000000..1a576848f91
--- /dev/null
+++ b/providers/opper/models/gpt-oss-120b.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "openai/gpt-oss-120b"
+
+reasoning_options = []
+
+[cost]
+input = 1.1622
+output = 4.88124
+
+[limit]
+context = 128000
+output = 8192
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/gpt-oss-20b.toml b/providers/opper/models/gpt-oss-20b.toml
new file mode 100644
index 00000000000..ce3715afc7b
--- /dev/null
+++ b/providers/opper/models/gpt-oss-20b.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "openai/gpt-oss-20b"
+
+reasoning_options = []
+
+[cost]
+input = 0.11622
+output = 0.488124
+
+[limit]
+context = 128000
+output = 8192
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/grok-4.3.toml b/providers/opper/models/grok-4.3.toml
new file mode 100644
index 00000000000..e40b4396000
--- /dev/null
+++ b/providers/opper/models/grok-4.3.toml
@@ -0,0 +1,18 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# All pool members publish flat rates: no pricing.thresholds or input_surcharge_threshold_tokens.
+# Opper billing uses this schedule; native context surcharges do not apply to these pools.
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "xai/grok-4.3"
+
+reasoning_options = []
+
+[cost]
+input = 1.25
+output = 2.5
+cache_read = 0.2
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/grok-4.5.toml b/providers/opper/models/grok-4.5.toml
new file mode 100644
index 00000000000..8cb4220459f
--- /dev/null
+++ b/providers/opper/models/grok-4.5.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# All pool members publish flat rates: no pricing.thresholds or input_surcharge_threshold_tokens.
+# Opper billing uses this schedule; native context surcharges do not apply to these pools.
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "xai/grok-4.5"
+
+reasoning_options = []
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/grok-4.6.toml b/providers/opper/models/grok-4.6.toml
new file mode 100644
index 00000000000..36c860ca4c1
--- /dev/null
+++ b/providers/opper/models/grok-4.6.toml
@@ -0,0 +1,20 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "xai/grok-4.6"
+
+reasoning_options = []
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200000 }
+input = 4
+output = 12
+cache_read = 1
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/grok-build-0.1.toml b/providers/opper/models/grok-build-0.1.toml
new file mode 100644
index 00000000000..13eef952286
--- /dev/null
+++ b/providers/opper/models/grok-build-0.1.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# All pool members publish flat rates: no pricing.thresholds or input_surcharge_threshold_tokens.
+# Opper billing uses this schedule; native context surcharges do not apply to these pools.
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "xai/grok-build-0.1"
+
+reasoning_options = []
+
+[cost]
+input = 1
+output = 2
+cache_read = 0.2
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/kimi-k3.toml b/providers/opper/models/kimi-k3.toml
new file mode 100644
index 00000000000..b4519ce43f6
--- /dev/null
+++ b/providers/opper/models/kimi-k3.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: members do not share a supported effort level.
+base_model = "moonshotai/kimi-k3"
+attachment = false
+
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+
+[modalities]
+input = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/minimax-m2.7.toml b/providers/opper/models/minimax-m2.7.toml
new file mode 100644
index 00000000000..ed41ef98618
--- /dev/null
+++ b/providers/opper/models/minimax-m2.7.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "minimax/MiniMax-M2.7"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 0.69732
+output = 2.78928
+
+[limit]
+context = 196608
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/minimax-m3.toml b/providers/opper/models/minimax-m3.toml
new file mode 100644
index 00000000000..fae074d323c
--- /dev/null
+++ b/providers/opper/models/minimax-m3.toml
@@ -0,0 +1,28 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+# Context tiers are conservative pool price ceilings, including marginal-rate bands.
+base_model = "minimax/MiniMax-M3"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 0.6
+output = 2.4
+cache_read = 0.12
+
+[[cost.tiers]]
+tier = { type = "context", size = 524288 }
+input = 1.2
+output = 4.8
+cache_read = 0.24
+
+[limit]
+context = 1000000
+output = 131072
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/minimax/m3.toml b/providers/opper/models/minimax/m3.toml
deleted file mode 100644
index d53af696874..00000000000
--- a/providers/opper/models/minimax/m3.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-# Opper billed schedule, verified against GET /v3/compat/models and /v3/models (2026-08-20):
-# base 0.6/2.4 (cache 0.12) up to 524_288, then 2x. This differs from MiniMax first-party
-# list (0.30/1.20 base, upper band 0.60/2.40); the delta is flagged to Opper catalog
-# maintainers and this entry will be updated if the billed schedule changes.
-base_model = "minimax/MiniMax-M3"
-
-reasoning_options = []
-
-[cost]
-input = 0.6
-output = 2.4
-cache_read = 0.12
-
-[[cost.tiers]]
-tier = { size = 524_288 }
-input = 1.2
-output = 4.8
-cache_read = 0.24
-
-[limit]
-context = 1_048_576
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/mistral-large-2512.toml b/providers/opper/models/mistral-large-2512.toml
new file mode 100644
index 00000000000..1f67818b4a1
--- /dev/null
+++ b/providers/opper/models/mistral-large-2512.toml
@@ -0,0 +1,11 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+base_model = "mistral/mistral-large-2512"
+structured_output = true
+
+[cost]
+input = 0.5
+output = 1.5
+
+[limit]
+context = 256000
+output = 8192
diff --git a/providers/opper/models/mistral-small-2603.toml b/providers/opper/models/mistral-small-2603.toml
new file mode 100644
index 00000000000..cff65202966
--- /dev/null
+++ b/providers/opper/models/mistral-small-2603.toml
@@ -0,0 +1,16 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "mistral/mistral-small-2603"
+structured_output = false
+
+reasoning_options = []
+
+[cost]
+input = 0.5811
+output = 2.44062
+
+[limit]
+output = 8192
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/mistral/devstral-2512.toml b/providers/opper/models/mistral/devstral-2512.toml
deleted file mode 100644
index 62661ddc309..00000000000
--- a/providers/opper/models/mistral/devstral-2512.toml
+++ /dev/null
@@ -1,5 +0,0 @@
-base_model = "mistral/devstral-2512"
-
-[cost]
-input = 0.4
-output = 2
diff --git a/providers/opper/models/mistral/mistral-large-2512.toml b/providers/opper/models/mistral/mistral-large-2512.toml
deleted file mode 100644
index 49df75a6342..00000000000
--- a/providers/opper/models/mistral/mistral-large-2512.toml
+++ /dev/null
@@ -1,5 +0,0 @@
-base_model = "mistral/mistral-large-2512"
-
-[cost]
-input = 0.5
-output = 1.5
diff --git a/providers/opper/models/mistral/mistral-small-2603.toml b/providers/opper/models/mistral/mistral-small-2603.toml
deleted file mode 100644
index c82dfc68357..00000000000
--- a/providers/opper/models/mistral/mistral-small-2603.toml
+++ /dev/null
@@ -1,9 +0,0 @@
-base_model = "mistral/mistral-small-2603"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0.15
-output = 0.6
diff --git a/providers/opper/models/moonshot/kimi-k3.toml b/providers/opper/models/moonshot/kimi-k3.toml
deleted file mode 100644
index 0862dd21e4c..00000000000
--- a/providers/opper/models/moonshot/kimi-k3.toml
+++ /dev/null
@@ -1,13 +0,0 @@
-base_model = "moonshotai/kimi-k3"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
-
-[cost]
-input = 3
-output = 15
-cache_read = 0.3
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/meta/muse-spark-1.2.toml b/providers/opper/models/muse-spark-1.2.toml
similarity index 51%
rename from providers/opper/models/meta/muse-spark-1.2.toml
rename to providers/opper/models/muse-spark-1.2.toml
index aea877da54a..200e6da0293 100644
--- a/providers/opper/models/meta/muse-spark-1.2.toml
+++ b/providers/opper/models/muse-spark-1.2.toml
@@ -1,3 +1,5 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
 base_model = "meta/muse-spark-1.2"
 
 reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]
@@ -6,3 +8,6 @@ reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "h
 input = 1.25
 output = 4.25
 cache_read = 0.15
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/muse-spark-1.3.toml b/providers/opper/models/muse-spark-1.3.toml
new file mode 100644
index 00000000000..bba0703ecde
--- /dev/null
+++ b/providers/opper/models/muse-spark-1.3.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Standard-tier pool forwards native effort unchanged; max verified live on 2026-09-08.
+# The catalog reasoning.supported list omits max, but the serving path accepts it.
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+base_model = "meta/muse-spark-1.3"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 1.25
+output = 4.25
+cache_read = 0.15
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/openai/gpt-5.3-chat-latest.toml b/providers/opper/models/openai/gpt-5.3-chat-latest.toml
deleted file mode 100644
index 197284ef5f0..00000000000
--- a/providers/opper/models/openai/gpt-5.3-chat-latest.toml
+++ /dev/null
@@ -1,6 +0,0 @@
-base_model = "openai/gpt-5.3-chat-latest"
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175
diff --git a/providers/opper/models/openai/gpt-5.3-codex.toml b/providers/opper/models/openai/gpt-5.3-codex.toml
deleted file mode 100644
index e85f5642a44..00000000000
--- a/providers/opper/models/openai/gpt-5.3-codex.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "openai/gpt-5.3-codex"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-
-[cost]
-input = 1.75
-output = 14
-cache_read = 0.175
diff --git a/providers/opper/models/openai/gpt-5.4-mini.toml b/providers/opper/models/openai/gpt-5.4-mini.toml
deleted file mode 100644
index a21229d609e..00000000000
--- a/providers/opper/models/openai/gpt-5.4-mini.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "openai/gpt-5.4-mini"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-
-[cost]
-input = 0.75
-output = 4.5
-cache_read = 0.075
diff --git a/providers/opper/models/openai/gpt-5.4-nano.toml b/providers/opper/models/openai/gpt-5.4-nano.toml
deleted file mode 100644
index 0db45e05b9e..00000000000
--- a/providers/opper/models/openai/gpt-5.4-nano.toml
+++ /dev/null
@@ -1,8 +0,0 @@
-base_model = "openai/gpt-5.4-nano"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-
-[cost]
-input = 0.2
-output = 1.25
-cache_read = 0.02
diff --git a/providers/opper/models/openai/gpt-5.4-pro.toml b/providers/opper/models/openai/gpt-5.4-pro.toml
deleted file mode 100644
index 6da8f6aa095..00000000000
--- a/providers/opper/models/openai/gpt-5.4-pro.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "openai/gpt-5.4-pro"
-
-reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
-
-[cost]
-input = 30
-output = 180
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 60.00
-output = 270.00
diff --git a/providers/opper/models/openai/gpt-5.4.toml b/providers/opper/models/openai/gpt-5.4.toml
deleted file mode 100644
index 7398a595b17..00000000000
--- a/providers/opper/models/openai/gpt-5.4.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "openai/gpt-5.4"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-
-[cost]
-input = 2.5
-output = 15
-cache_read = 0.25
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 5.00
-output = 22.50
-cache_read = 0.50
diff --git a/providers/opper/models/openai/gpt-5.5-pro.toml b/providers/opper/models/openai/gpt-5.5-pro.toml
deleted file mode 100644
index 77d5371cb8e..00000000000
--- a/providers/opper/models/openai/gpt-5.5-pro.toml
+++ /dev/null
@@ -1,12 +0,0 @@
-base_model = "openai/gpt-5.5-pro"
-
-reasoning_options = [{ type = "effort", values = ["medium", "high", "xhigh"] }]
-
-[cost]
-input = 30
-output = 180
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 60.00
-output = 270.00
diff --git a/providers/opper/models/openai/gpt-5.5.toml b/providers/opper/models/openai/gpt-5.5.toml
deleted file mode 100644
index a58539b3e09..00000000000
--- a/providers/opper/models/openai/gpt-5.5.toml
+++ /dev/null
@@ -1,14 +0,0 @@
-base_model = "openai/gpt-5.5"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
-
-[cost]
-input = 5
-output = 30
-cache_read = 0.5
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 10.00
-output = 45.00
-cache_read = 1.00
diff --git a/providers/opper/models/openai/gpt-5.6-luna.toml b/providers/opper/models/openai/gpt-5.6-luna.toml
deleted file mode 100644
index 07d577afa36..00000000000
--- a/providers/opper/models/openai/gpt-5.6-luna.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "openai/gpt-5.6-luna"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 0.2
-output = 1.2
-cache_read = 0.02
-cache_write = 0.25
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 0.40
-output = 1.80
-cache_read = 0.04
-cache_write = 0.50
diff --git a/providers/opper/models/openai/gpt-5.6-sol.toml b/providers/opper/models/openai/gpt-5.6-sol.toml
deleted file mode 100644
index 31470933d23..00000000000
--- a/providers/opper/models/openai/gpt-5.6-sol.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "openai/gpt-5.6-sol"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 5
-output = 30
-cache_read = 0.5
-cache_write = 6.25
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 10.00
-output = 45.00
-cache_read = 1.00
-cache_write = 12.50
diff --git a/providers/opper/models/openai/gpt-5.6-terra.toml b/providers/opper/models/openai/gpt-5.6-terra.toml
deleted file mode 100644
index 9af64197094..00000000000
--- a/providers/opper/models/openai/gpt-5.6-terra.toml
+++ /dev/null
@@ -1,16 +0,0 @@
-base_model = "openai/gpt-5.6-terra"
-
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 2
-output = 12
-cache_read = 0.2
-cache_write = 2.5
-
-[[cost.tiers]]
-tier = { size = 272_000 }
-input = 4.00
-output = 18.00
-cache_read = 0.40
-cache_write = 5.00
diff --git a/providers/opper/models/perplexity/sonar-pro.toml b/providers/opper/models/perplexity/sonar-pro.toml
deleted file mode 100644
index f03972e3868..00000000000
--- a/providers/opper/models/perplexity/sonar-pro.toml
+++ /dev/null
@@ -1,5 +0,0 @@
-base_model = "perplexity/sonar-pro"
-
-[cost]
-input = 3
-output = 15
diff --git a/providers/opper/models/perplexity/sonar-reasoning-pro.toml b/providers/opper/models/perplexity/sonar-reasoning-pro.toml
deleted file mode 100644
index 6189d9d9ff8..00000000000
--- a/providers/opper/models/perplexity/sonar-reasoning-pro.toml
+++ /dev/null
@@ -1,7 +0,0 @@
-base_model = "perplexity/sonar-reasoning-pro"
-
-reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high"] }]
-
-[cost]
-input = 2
-output = 8
diff --git a/providers/opper/models/perplexity/sonar.toml b/providers/opper/models/perplexity/sonar.toml
deleted file mode 100644
index a6c921f515e..00000000000
--- a/providers/opper/models/perplexity/sonar.toml
+++ /dev/null
@@ -1,5 +0,0 @@
-base_model = "perplexity/sonar"
-
-[cost]
-input = 1
-output = 1
diff --git a/providers/opper/models/qwen3-coder-next.toml b/providers/opper/models/qwen3-coder-next.toml
new file mode 100644
index 00000000000..2816f8c2b7e
--- /dev/null
+++ b/providers/opper/models/qwen3-coder-next.toml
@@ -0,0 +1,6 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+base_model = "alibaba/qwen3-coder-next"
+
+[cost]
+input = 0.5811
+output = 2.3244
diff --git a/providers/opper/models/qwen3.6-35b-a3b.toml b/providers/opper/models/qwen3.6-35b-a3b.toml
new file mode 100644
index 00000000000..f7e98234caf
--- /dev/null
+++ b/providers/opper/models/qwen3.6-35b-a3b.toml
@@ -0,0 +1,15 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "alibaba/qwen3.6-35b-a3b"
+
+reasoning_options = []
+
+[cost]
+input = 0.248
+output = 1.485
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/qwen3.8-2.4t-a95b.toml b/providers/opper/models/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..db70485f41e
--- /dev/null
+++ b/providers/opper/models/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,13 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+
+reasoning_options = []
+
+[cost]
+input = 2.5
+output = 6
+cache_read = 0.63
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/qwen3.8-27b.toml b/providers/opper/models/qwen3.8-27b.toml
new file mode 100644
index 00000000000..8cd1f5f6882
--- /dev/null
+++ b/providers/opper/models/qwen3.8-27b.toml
@@ -0,0 +1,17 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# Effort: top-level reasoning_effort; levels shared by the pool members.
+base_model = "alibaba/qwen3.8-27b"
+attachment = false
+structured_output = false
+
+reasoning_options = [{ type = "effort", values = ["low", "medium", "xhigh"] }]
+
+[cost]
+input = 0.5811
+output = 3
+
+[modalities]
+input = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/qwen3.8-max.toml b/providers/opper/models/qwen3.8-max.toml
new file mode 100644
index 00000000000..65905a41b73
--- /dev/null
+++ b/providers/opper/models/qwen3.8-max.toml
@@ -0,0 +1,20 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide graded effort or toggle: members share only reasoning_effort=none.
+base_model = "alibaba/qwen3.8-max"
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 983616
+
+[modalities]
+input = ["text", "image"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/sonar-pro.toml b/providers/opper/models/sonar-pro.toml
new file mode 100644
index 00000000000..de476cca476
--- /dev/null
+++ b/providers/opper/models/sonar-pro.toml
@@ -0,0 +1,14 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+base_model = "perplexity/sonar-pro"
+attachment = false
+structured_output = true
+
+[cost]
+input = 3
+output = 15
+
+[limit]
+output = 8000
+
+[modalities]
+input = ["text"]
diff --git a/providers/opper/models/sonar-reasoning-pro.toml b/providers/opper/models/sonar-reasoning-pro.toml
new file mode 100644
index 00000000000..98d5cd92398
--- /dev/null
+++ b/providers/opper/models/sonar-reasoning-pro.toml
@@ -0,0 +1,17 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+# No pool-wide effort control: at least one member does not forward reasoning_effort.
+base_model = "perplexity/sonar-reasoning-pro"
+attachment = false
+structured_output = true
+
+reasoning_options = []
+
+[cost]
+input = 2
+output = 8
+
+[modalities]
+input = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/opper/models/sonar.toml b/providers/opper/models/sonar.toml
new file mode 100644
index 00000000000..7242f639443
--- /dev/null
+++ b/providers/opper/models/sonar.toml
@@ -0,0 +1,7 @@
+# Pool metadata: https://api.opper.ai/v3/models (2026-09-08).
+base_model = "perplexity/sonar"
+structured_output = true
+
+[cost]
+input = 1
+output = 1
diff --git a/providers/opper/models/vertexai/gemini-3.7-flash-eu.toml b/providers/opper/models/vertexai/gemini-3.7-flash-eu.toml
deleted file mode 100644
index 03e8b29cab7..00000000000
--- a/providers/opper/models/vertexai/gemini-3.7-flash-eu.toml
+++ /dev/null
@@ -1,11 +0,0 @@
-base_model = "google/gemini-3.7-flash"
-name = "Gemini 3.7 Flash (EU)"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.75
-output = 3.75
-cache_read = 0.075
diff --git a/providers/opper/models/vertexai/gemini-3.7-flash.toml b/providers/opper/models/vertexai/gemini-3.7-flash.toml
deleted file mode 100644
index e9bacad26c9..00000000000
--- a/providers/opper/models/vertexai/gemini-3.7-flash.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "google/gemini-3.7-flash"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 0.75
-output = 3.75
-cache_read = 0.075
diff --git a/providers/opper/models/xai/grok-4.3.toml b/providers/opper/models/xai/grok-4.3.toml
deleted file mode 100644
index 329b82bb4ac..00000000000
--- a/providers/opper/models/xai/grok-4.3.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "xai/grok-4.3"
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "low", "medium", "high"]
-
-[cost]
-input = 1.25
-output = 2.5
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { size = 200_000 }
-input = 2.5
-output = 5
-cache_read = 0.4
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/xai/grok-4.5.toml b/providers/opper/models/xai/grok-4.5.toml
deleted file mode 100644
index fd10a1caace..00000000000
--- a/providers/opper/models/xai/grok-4.5.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "xai/grok-4.5"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
-
-[cost]
-input = 2
-output = 6
-cache_read = 0.3
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 12
-cache_read = 0.6
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/xai/grok-4.6.toml b/providers/opper/models/xai/grok-4.6.toml
deleted file mode 100644
index 1f8386fd483..00000000000
--- a/providers/opper/models/xai/grok-4.6.toml
+++ /dev/null
@@ -1,19 +0,0 @@
-base_model = "xai/grok-4.6"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high", "xhigh"]
-
-[cost]
-input = 2
-output = 6
-cache_read = 0.5
-
-[[cost.tiers]]
-tier = { type = "context", size = 200_000 }
-input = 4
-output = 12
-cache_read = 1
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/models/xai/grok-build-0.1.toml b/providers/opper/models/xai/grok-build-0.1.toml
deleted file mode 100644
index b2c4aeeafad..00000000000
--- a/providers/opper/models/xai/grok-build-0.1.toml
+++ /dev/null
@@ -1,17 +0,0 @@
-base_model = "xai/grok-build-0.1"
-
-reasoning_options = []
-
-[cost]
-input = 1
-output = 2
-cache_read = 0.2
-
-[[cost.tiers]]
-tier = { size = 200_000 }
-input = 2.00
-output = 4.00
-cache_read = 0.40
-
-[interleaved]
-field = "reasoning_content"
diff --git a/providers/opper/provider.toml b/providers/opper/provider.toml
index 7d14eff2de3..7ccd9102a07 100644
--- a/providers/opper/provider.toml
+++ b/providers/opper/provider.toml
@@ -1,13 +1,20 @@
+# Sources (2026-09-08): https://api.opper.ai/v3/models?limit=0&type=llm
+# and authenticated https://api.opper.ai/v3/compat/models?type=model,pool.
+# Model IDs are bare Opper pool names; Opper selects the underlying provider.
+# Prices are USD/MTok ceilings across pool members, not a fixed rate for every
+# request. Cache-read discounts are listed only when all priced members offer
+# them. Context tiers include the highest marginal rates and whole-request
+# surcharges, so estimates can exceed the actual bill. No currency conversion
+# is needed: the catalog already converts non-USD provider prices to USD.
+# Limits do not exceed either the lab model or the pool's smallest known limit;
+# input modalities are restricted to those shared by the lab and pool members.
+# POST /v3/compat/chat/completions accepts top-level reasoning_effort. Native
+# thinking toggles and reasoning-token budgets are not exposed on this surface.
+# Effort options describe common pool controls; [] means no pool-wide control,
+# including pools where some adapters ignore effort. Reasoning output, when
+# present, uses message.reasoning_content or streaming delta.reasoning_content.
 name = "Opper"
 env = ["OPPER_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
 api = "https://api.opper.ai/v3/compat"
-# Reasoning HTTP format (measured against POST /v3/compat/chat/completions, 2026-08-19):
-# top-level reasoning_effort takes an effort string that is passed through to the
-# upstream model API unchanged, so each model accepts its native set (Anthropic:
-# low|medium|high|max; OpenAI GPT-5.x: none|low|medium|high|xhigh and max where the
-# model supports it; Gemini: minimal|low|medium|high). Budget-token values are not
-# accepted on this surface, and no separate reasoning on/off toggle is exposed. Models with no upstream effort control accept the
-# parameter without effect, so those entries use reasoning_options = [].
-# Pricing is mirrored from GET /v3/compat/models (USD per token, no gateway markup).
 doc = "https://opper.ai/models"

From 4c98cacfe9bdeec580eba60827bf696c97c77f0c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 17:23:22 +0000
Subject: [PATCH 1807/2012] chore(sync): update OpenRouter model catalog
 (#7654)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 ++++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/ibm-granite/granite-4.2-8b.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-27b.toml         | 6 +++---
 providers/openrouter/models/x-ai/grok-4.7.toml            | 3 ---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 6 files changed, 17 insertions(+), 19 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 13dfedbfc61..745a5e03d0f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57288
-output = 1.71864
-cache_read = 0.019096
+input = 0.57156
+output = 1.71468
+cache_read = 0.018186
 
 [limit]
 context = 1_048_576
+output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a705f18ca62..9c2770326f2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.936294
-output = 1.872588
-cache_read = 0.078025
+input = 0.931074
+output = 1.862148
+cache_read = 0.07759
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
index 1b683dd536e..62943cfaadb 100644
--- a/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
+++ b/providers/openrouter/models/ibm-granite/granite-4.2-8b.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.1
-output = 0.15
-cache_read = 0.05
+input = 0.06
+output = 0.25
+cache_read = 0.015
 
 [limit]
 context = 131_072
diff --git a/providers/openrouter/models/qwen/qwen3.8-27b.toml b/providers/openrouter/models/qwen/qwen3.8-27b.toml
index 64a9a2f6128..d1ddcffa2ed 100644
--- a/providers/openrouter/models/qwen/qwen3.8-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-27b.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.2
-output = 2.5
-cache_read = 0.05
+input = 0.42
+output = 3
+cache_read = 0.085
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/x-ai/grok-4.7.toml b/providers/openrouter/models/x-ai/grok-4.7.toml
index 49d2235d6f9..3ffa2ff7ae3 100644
--- a/providers/openrouter/models/x-ai/grok-4.7.toml
+++ b/providers/openrouter/models/x-ai/grok-4.7.toml
@@ -18,6 +18,3 @@ cache_read = 0.8
 
 [limit]
 output = 450_000
-
-[modalities]
-input = ["text", "image", "pdf"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 3a21dc3b34a..24cf56881f7 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57288
-output = 1.71864
-cache_read = 0.019096
+input = 0.57156
+output = 1.71468
+cache_read = 0.018186
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 85d8b02bfd7ea081f462e62c7a875545d12be4d9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 17:23:25 +0000
Subject: [PATCH 1808/2012] chore(sync): update Venice model catalog (#7653)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/grok-4-7.toml | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/venice/models/grok-4-7.toml

diff --git a/providers/venice/models/grok-4-7.toml b/providers/venice/models/grok-4-7.toml
new file mode 100644
index 00000000000..e405114dc5c
--- /dev/null
+++ b/providers/venice/models/grok-4-7.toml
@@ -0,0 +1,24 @@
+base_model = "xai/grok-4.7"
+description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+release_date = "2026-09-16"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2.27
+output = 6.8
+cache_read = 0.57
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4.53
+output = 13.6
+cache_read = 1.13
+
+[limit]
+output = 200_000
+
+[modalities]
+input = ["text", "image"]

From ddc239ff7d95cc769723f2ee5be6a4139e3120de Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 17:23:38 +0000
Subject: [PATCH 1809/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7655)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/spacexai/grok-4.7.toml | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/providers/vercel/models/spacexai/grok-4.7.toml b/providers/vercel/models/spacexai/grok-4.7.toml
index defc4659f57..95c1c2ce193 100644
--- a/providers/vercel/models/spacexai/grok-4.7.toml
+++ b/providers/vercel/models/spacexai/grok-4.7.toml
@@ -1,5 +1,8 @@
 base_model = "xai/grok-4.7"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 1.2
@@ -11,3 +14,6 @@ tier = { type = "context", size = 200_001 }
 input = 2.4
 output = 7.2
 cache_read = 0.6
+
+[modalities]
+input = ["text", "image"]

From c8bc0ea7c90ae825c251476236d1403e4d64292b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 17:24:06 +0000
Subject: [PATCH 1810/2012] chore(sync): update Kilo model catalog (#7656)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 ++-
 providers/kilo/models/qwen/qwen3.8-27b.toml            |  1 +
 providers/kilo/models/x-ai/grok-4.7.toml               |  3 ---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 4 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..d4d63d57a55 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,5 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
+output = 393_216
diff --git a/providers/kilo/models/qwen/qwen3.8-27b.toml b/providers/kilo/models/qwen/qwen3.8-27b.toml
index 9b18fb98b33..820f69fb194 100644
--- a/providers/kilo/models/qwen/qwen3.8-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-27b.toml
@@ -12,4 +12,5 @@ cache_read = 0.085
 cache_write = 0.53125
 
 [limit]
+context = 1_000_000
 output = 131_072
diff --git a/providers/kilo/models/x-ai/grok-4.7.toml b/providers/kilo/models/x-ai/grok-4.7.toml
index bdb4be1fce7..7e904bb941d 100644
--- a/providers/kilo/models/x-ai/grok-4.7.toml
+++ b/providers/kilo/models/x-ai/grok-4.7.toml
@@ -12,6 +12,3 @@ cache_read = 0.4
 
 [limit]
 output = 450_000
-
-[modalities]
-input = ["text", "image", "pdf"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index d1321db247a..c10a096a2ae 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57288
-output = 1.71864
-cache_read = 0.019096
+input = 0.57156
+output = 1.71468
+cache_read = 0.018186
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]

From c4eefc3346ada0c358f9c5f7878cdf0cda22d9cc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:30 +0000
Subject: [PATCH 1811/2012] chore(sync): update OpenRouter model catalog
 (#7663)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash-0731.toml | 2 +-
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 providers/openrouter/models/moonshotai/kimi-k3.toml        | 6 +++---
 .../openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml  | 5 +++--
 providers/openrouter/models/z-ai/glm-5.3-flash.toml        | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml   | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml         | 4 ++--
 providers/openrouter/models/~z-ai/glm-flash-latest.toml    | 4 ++--
 9 files changed, 24 insertions(+), 22 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index be2f907d844..48200f680e3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -11,7 +11,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.16
+output = 0.32
 cache_read = 0.016
 
 [limit]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 745a5e03d0f..c11c30e3519 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57156
-output = 1.71468
-cache_read = 0.018186
+input = 0.56628
+output = 1.69884
+cache_read = 0.018018
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 9c2770326f2..399e63186f7 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.931074
-output = 1.862148
-cache_read = 0.07759
+input = 0.924462
+output = 1.848924
+cache_read = 0.077039
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k3.toml b/providers/openrouter/models/moonshotai/kimi-k3.toml
index 22b1407542a..c11eb6c3dae 100644
--- a/providers/openrouter/models/moonshotai/kimi-k3.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k3.toml
@@ -12,9 +12,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 3
+output = 15
+cache_read = 0.3
 
 [limit]
 output = 943_718
diff --git a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
index 297dfb56987..bf732029012 100644
--- a/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3-nano-30b-a3b.toml
@@ -7,8 +7,9 @@ structured_output = true
 type = "toggle"
 
 [cost]
-input = 0.06
-output = 0.24
+input = 0.05
+output = 0.2
+cache_read = 0.03
 
 [limit]
 output = 235_929
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index f82829f96ea..6fd3f066cfd 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,12 +6,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.09
-output = 0.3
-cache_read = 0.018
+input = 0.075
+output = 0.25
+cache_read = 0.02
 
 [limit]
 context = 1_310_720
+output = 102_400
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 24cf56881f7..427614e0228 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57156
-output = 1.71468
-cache_read = 0.018186
+input = 0.56628
+output = 1.69884
+cache_read = 0.018018
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index c98a00af617..867340e2671 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,8 +21,8 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.16
-cache_read = 0.016
+output = 0.2
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..ac623f94e52 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -17,11 +17,11 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.075
 output = 0.25
-cache_read = 0.015
+cache_read = 0.02
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 102_400
 
 [modalities]
 input = ["text", "image", "video"]

From d626e17c832129c7b6f8f1f5005234f167b9867c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:37 +0000
Subject: [PATCH 1812/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7661)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../llmgateway/models/glm-5.3-flash.toml      |  6 ++--
 providers/llmgateway/models/grok-4-7.toml     | 28 +++++++++++++++++++
 2 files changed, 31 insertions(+), 3 deletions(-)
 create mode 100644 providers/llmgateway/models/grok-4-7.toml

diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index d46987e8acc..5aecfa64069 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.088
-output = 0.25
-cache_read = 0.025
+input = 0.07
+output = 0.19
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
diff --git a/providers/llmgateway/models/grok-4-7.toml b/providers/llmgateway/models/grok-4-7.toml
new file mode 100644
index 00000000000..4e89d64b78f
--- /dev/null
+++ b/providers/llmgateway/models/grok-4-7.toml
@@ -0,0 +1,28 @@
+name = "Grok 4.7"
+description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+family = "grok"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[limit]
+context = 500_000
+output = 500_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From d7ce2f138822ee40b5b2b51c8ab1ba74b2d56c20 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:45 +0000
Subject: [PATCH 1813/2012] chore(sync): update Kilo model catalog (#7659)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/moonshotai/kimi-k3.toml               | 6 +++---
 providers/kilo/models/z-ai/glm-5.3-flash.toml               | 1 +
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 4 ++--
 providers/kilo/models/~z-ai/glm-flash-latest.toml           | 4 ++--
 5 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/moonshotai/kimi-k3.toml b/providers/kilo/models/moonshotai/kimi-k3.toml
index fd81e087ca6..1ebab0a52b7 100644
--- a/providers/kilo/models/moonshotai/kimi-k3.toml
+++ b/providers/kilo/models/moonshotai/kimi-k3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 3
+output = 15
+cache_read = 0.3
 
 [limit]
 output = 943_718
diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 0104ea56ade..1bc5d8b7a69 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -12,6 +12,7 @@ cache_read = 0.03
 
 [limit]
 context = 1_048_576
+output = 102_400
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index c10a096a2ae..191b436787a 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57156
-output = 1.71468
-cache_read = 0.018186
+input = 0.56628
+output = 1.69884
+cache_read = 0.018018
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 7ead984259a..49d602e67e6 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,8 +16,8 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.16
-cache_read = 0.016
+output = 0.2
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..5035574d3d5 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -17,11 +17,11 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.075
 output = 0.25
-cache_read = 0.015
+cache_read = 0.02
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 102_400
 
 [modalities]
 input = ["text", "image", "video"]

From e30e2fc6e88d1e5ad06a4780a9516f1de179cdc8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:47 +0000
Subject: [PATCH 1814/2012] chore(sync): update Merge Gateway model catalog
 (#7662)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../merge-gateway/models/qwen/qwen3.5-27b.toml      |  7 ++++---
 providers/merge-gateway/models/xai/grok-4.7.toml    | 13 +++++++++++++
 2 files changed, 17 insertions(+), 3 deletions(-)
 create mode 100644 providers/merge-gateway/models/xai/grok-4.7.toml

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
index f0885b3822b..d7d8d42decc 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-27b.toml
@@ -1,5 +1,6 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-27b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-27b"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -10,8 +11,8 @@ output = 0.688
 cache_read = 0.0172
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/xai/grok-4.7.toml b/providers/merge-gateway/models/xai/grok-4.7.toml
new file mode 100644
index 00000000000..19465981ea7
--- /dev/null
+++ b/providers/merge-gateway/models/xai/grok-4.7.toml
@@ -0,0 +1,13 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[modalities]
+input = ["text", "image"]

From c28d1e58ad585c41613d7762b8a8ab438b6bf737 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:49 +0000
Subject: [PATCH 1815/2012] chore(sync): update Charm Hyper model catalog
 (#7660)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/hyper/models/deepseek-v4.1-flash.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/providers/hyper/models/deepseek-v4.1-flash.toml b/providers/hyper/models/deepseek-v4.1-flash.toml
index c703c4fcf0c..7fd6fdfb3b1 100644
--- a/providers/hyper/models/deepseek-v4.1-flash.toml
+++ b/providers/hyper/models/deepseek-v4.1-flash.toml
@@ -2,7 +2,7 @@ base_model = "deepseek/deepseek-v4.1-flash"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "max"]
+values = ["low", "high", "xhigh"]
 
 [cost]
 input = 0.3
@@ -10,4 +10,5 @@ output = 1.2
 cache_read = 0.03
 
 [limit]
+context = 1_048_576
 output = 32_768

From d7d8109e69c5bf10cf7fc6172e04d64234ea180b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 18:31:52 +0000
Subject: [PATCH 1816/2012] chore(sync): update LLM Gateway model catalog
 (#7664)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gonka24/glm-5.3-flash.toml         | 20 +++++++++++++
 .../models/xai/grok-4-7.toml                  | 28 +++++++++++++++++++
 2 files changed, 48 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
 create mode 100644 providers/llmgateway-providers/models/xai/grok-4-7.toml

diff --git a/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml b/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
new file mode 100644
index 00000000000..921d7dc252d
--- /dev/null
+++ b/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
@@ -0,0 +1,20 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (Gonka24)"
+attachment = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.07
+output = 0.19
+cache_read = 0.015
+
+[limit]
+context = 200_000
+output = 16_384
+
+[modalities]
+input = ["text"]
diff --git a/providers/llmgateway-providers/models/xai/grok-4-7.toml b/providers/llmgateway-providers/models/xai/grok-4-7.toml
new file mode 100644
index 00000000000..03bd9afa268
--- /dev/null
+++ b/providers/llmgateway-providers/models/xai/grok-4-7.toml
@@ -0,0 +1,28 @@
+name = "Grok 4.7 (xAI)"
+description = "Grok model for agentic tool use, reasoning, coding, and live assistance"
+family = "grok"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[limit]
+context = 500_000
+output = 500_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 422b34ab53646b470469b1bc9900a252f288b079 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:35:19 -0500
Subject: [PATCH 1817/2012] fix(sync): ingest aiand pricing tiers (#7650)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/aiand.ts | 44 ++++++++++++-----
 packages/core/test/aiand.test.ts          | 58 ++++++++++++++++++++++-
 2 files changed, 87 insertions(+), 15 deletions(-)

diff --git a/packages/core/src/sync/providers/aiand.ts b/packages/core/src/sync/providers/aiand.ts
index 9545bd91e0f..860bbc50944 100644
--- a/packages/core/src/sync/providers/aiand.ts
+++ b/packages/core/src/sync/providers/aiand.ts
@@ -31,6 +31,28 @@ const FeedReasoningOption = z.union([
   CatalogReasoningOption,
 ]);
 
+const PRICE = z.number().nonnegative();
+const FeedCostFields = {
+  input: PRICE,
+  output: PRICE,
+  reasoning: PRICE.optional(),
+  cache_read: PRICE.optional(),
+  cache_write: PRICE.optional(),
+  input_audio: PRICE.optional(),
+  output_audio: PRICE.optional(),
+};
+const FeedCostTier = z
+  .object({
+    ...FeedCostFields,
+    tier: z
+      .object({
+        type: z.literal("context").default("context"),
+        size: z.number().int().nonnegative(),
+      })
+      .passthrough(),
+  })
+  .passthrough();
+
 export const AiandModel = z
   .object({
     id: z.string().min(1),
@@ -45,13 +67,7 @@ export const AiandModel = z
     temperature: z.boolean(),
     tool_call: z.boolean(),
     structured_output: z.boolean().optional(),
-    cost: z
-      .object({
-        input: z.number().nonnegative(),
-        output: z.number().nonnegative(),
-        cache_read: z.number().nonnegative().optional(),
-      })
-      .passthrough(),
+    cost: z.object({ ...FeedCostFields, tiers: z.array(FeedCostTier).optional() }).passthrough(),
     limit: z
       .object({
         context: z.number().int().positive(),
@@ -182,12 +198,14 @@ export function buildAiandModel(
     cost: {
       input: model.cost.input,
       output: model.cost.output,
-      reasoning: authored?.cost?.reasoning,
-      cache_read: model.cost.cache_read,
-      cache_write: authored?.cost?.cache_write,
-      input_audio: authored?.cost?.input_audio,
-      output_audio: authored?.cost?.output_audio,
-      tiers: authored?.cost?.tiers,
+      reasoning: model.cost.reasoning ?? authored?.cost?.reasoning,
+      cache_read: model.cost.cache_read ?? authored?.cost?.cache_read,
+      cache_write: model.cost.cache_write ?? authored?.cost?.cache_write,
+      input_audio: model.cost.input_audio ?? authored?.cost?.input_audio,
+      output_audio: model.cost.output_audio ?? authored?.cost?.output_audio,
+      // Unlike auxiliary prices, tiers are a complete source assertion: an
+      // omitted list means flat pricing and clears stale authored tiers.
+      tiers: model.cost.tiers,
     },
     // Absence means active on the feed, and the feed owns deprecation: a
     // curated alpha/beta survives omission, a curated deprecated does not,
diff --git a/packages/core/test/aiand.test.ts b/packages/core/test/aiand.test.ts
index d4e0c74c1fb..9e531806203 100644
--- a/packages/core/test/aiand.test.ts
+++ b/packages/core/test/aiand.test.ts
@@ -173,13 +173,67 @@ test("skippedNotice stays silent on a clean sync", () => {
 test("authored-only cost fields ride along; feed prices are authoritative", () => {
   const built = buildAiandModel(
     aiandModel(),
-    { cost: { input: 9, output: 9, cache_write: 0.5 }, limit: { context: 1, input: 128_000, output: 1 } },
+    {
+      cost: { input: 9, output: 9, reasoning: 0.75, cache_write: 0.5, input_audio: 1.25 },
+      limit: { context: 1, input: 128_000, output: 1 },
+    },
     null,
   );
-  expect(built.cost).toMatchObject({ input: 0.15, output: 0.25, cache_write: 0.5 });
+  expect(built.cost).toMatchObject({
+    input: 0.15,
+    output: 0.25,
+    reasoning: 0.75,
+    cache_read: 0.08,
+    cache_write: 0.5,
+    input_audio: 1.25,
+  });
   expect(built.limit).toEqual({ context: 1_048_576, input: 128_000, output: 384_000 });
 });
 
+test("feed cost tiers and auxiliary prices replace authored values", () => {
+  const tiers = [{
+    tier: { type: "context" as const, size: 200_000 },
+    input: 0.3,
+    output: 0.5,
+    cache_read: 0.16,
+  }];
+  const model = AiandModel.parse({
+    ...aiandModel(),
+    cost: {
+      ...aiandModel().cost,
+      reasoning: 0.4,
+      cache_write: 0.12,
+      output_audio: 2,
+      tiers,
+    },
+  });
+  const built = buildAiandModel(model, {
+    cost: {
+      input: 9,
+      output: 9,
+      reasoning: 9,
+      cache_write: 9,
+      output_audio: 9,
+      tiers: [{ tier: { type: "context", size: 100_000 }, input: 9, output: 9 }],
+    },
+  }, null);
+
+  expect(built.cost).toMatchObject({ reasoning: 0.4, cache_write: 0.12, output_audio: 2 });
+  expect(built.cost?.tiers).toEqual(tiers);
+});
+
+test("omitted feed cost tiers clear authored tiers", () => {
+  const built = buildAiandModel(aiandModel(), {
+    cost: {
+      input: 9,
+      output: 9,
+      tiers: [{ tier: { type: "context", size: 200_000 }, input: 9, output: 9 }],
+    },
+  }, null);
+
+  expect(built.cost?.tiers).toBeUndefined();
+});
+
 test("a non-reasoning feed model omits reasoning_options entirely", () => {
   const built = buildAiandModel(
     aiandModel({ reasoning: false, reasoning_options: undefined }),

From 68fbad48e4bdba5168048d80c8c3962ae4583f91 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:36:00 -0500
Subject: [PATCH 1818/2012] fix(sync): use Requesty pricing bands (#7652)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 packages/core/src/sync/providers/requesty.ts |  9 ++-
 packages/core/test/requesty.test.ts          | 78 ++++++++++++++++++++
 2 files changed, 83 insertions(+), 4 deletions(-)

diff --git a/packages/core/src/sync/providers/requesty.ts b/packages/core/src/sync/providers/requesty.ts
index 00f6dbcc990..d8b0bdf4de5 100644
--- a/packages/core/src/sync/providers/requesty.ts
+++ b/packages/core/src/sync/providers/requesty.ts
@@ -184,8 +184,9 @@ function reasoningOptions(
 }
 
 function buildCost(model: RequestyModel): SyncedFullModel["cost"] {
-    const input = model.input_price;
-    const output = model.output_price;
+    const base = model.pricing?.[0] ?? model;
+    const input = base.input_price;
+    const output = base.output_price;
     if (input == null || output == null) return undefined;
 
     const tiers = (model.pricing ?? []).slice(1).map((band) => ({
@@ -198,8 +199,8 @@ function buildCost(model: RequestyModel): SyncedFullModel["cost"] {
     return {
         input: pricePerMillion(input),
         output: pricePerMillion(output),
-        cache_read: chargedPricePerMillion(model.cached_price),
-        cache_write: chargedPricePerMillion(model.caching_price),
+        cache_read: chargedPricePerMillion(base.cached_price),
+        cache_write: chargedPricePerMillion(base.caching_price),
         tiers: tiers.length > 0 ? tiers : undefined,
     };
 }
diff --git a/packages/core/test/requesty.test.ts b/packages/core/test/requesty.test.ts
index 597b6de65c8..52eaec2ccde 100644
--- a/packages/core/test/requesty.test.ts
+++ b/packages/core/test/requesty.test.ts
@@ -49,3 +49,81 @@ test.each(["claude-fable-5.1", "claude-fable-5.1@eu"])(
     });
   },
 );
+
+test("uses the first Requesty pricing band as the base cost", () => {
+  const model = buildRequestyModel(RequestyModel.parse({
+    id: "requesty-priced-model",
+    created: Date.parse("2026-09-01") / 1_000,
+    description: "Requesty description",
+    context_window: 1_000_000,
+    max_output_tokens: 128_000,
+    input_price: 0.000001,
+    output_price: 0.000002,
+    cached_price: 0.000003,
+    caching_price: 0.000004,
+    pricing: [
+      {
+        prompt_tokens_threshold: 0,
+        input_price: 0.00001,
+        output_price: 0.00005,
+        cached_price: 0.00000025,
+        caching_price: 0.0000125,
+      },
+      {
+        prompt_tokens_threshold: 200_000,
+        input_price: 0.00002,
+        output_price: 0.000075,
+        cached_price: 0.0000005,
+        caching_price: 0.000025,
+      },
+    ],
+  }));
+
+  expect(model.cost).toEqual({
+    input: 10,
+    output: 50,
+    cache_read: 0.25,
+    cache_write: 12.5,
+    tiers: [{
+      tier: { type: "context", size: 200_000 },
+      input: 20,
+      output: 75,
+      cache_read: 0.5,
+      cache_write: 25,
+    }],
+  });
+});
+
+test("uses Requesty pricing bands when top-level prices are null", () => {
+  const model = buildRequestyModel(RequestyModel.parse({
+    id: "requesty-priced-model",
+    created: Date.parse("2026-09-01") / 1_000,
+    description: "Requesty description",
+    context_window: 1_000_000,
+    max_output_tokens: 128_000,
+    input_price: null,
+    output_price: null,
+    pricing: [
+      {
+        prompt_tokens_threshold: 0,
+        input_price: 0.00001,
+        output_price: 0.00005,
+      },
+      {
+        prompt_tokens_threshold: 200_000,
+        input_price: null,
+        output_price: null,
+      },
+    ],
+  }));
+
+  expect(model.cost).toEqual({
+    input: 10,
+    output: 50,
+    tiers: [{
+      tier: { type: "context", size: 200_000 },
+      input: 10,
+      output: 50,
+    }],
+  });
+});

From b449ad12403ef4903ed837d6c9fb8ea3b10b25f8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:38:26 -0500
Subject: [PATCH 1819/2012] fix(sync): clear stale CrossModel pricing tiers
 (#7649)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../core/src/sync/providers/crossmodel.ts     |  7 ++--
 packages/core/test/sync.test.ts               | 32 +++++++++++++++++++
 2 files changed, 36 insertions(+), 3 deletions(-)

diff --git a/packages/core/src/sync/providers/crossmodel.ts b/packages/core/src/sync/providers/crossmodel.ts
index e405a811a8f..0a1f3d067fe 100644
--- a/packages/core/src/sync/providers/crossmodel.ts
+++ b/packages/core/src/sync/providers/crossmodel.ts
@@ -193,8 +193,9 @@ export function buildCrossModel(
   // CrossModel serves threshold-tiered pricing. The lowest-threshold tier is the
   // headline [cost]; every higher tier maps to a [[cost.tiers]] context band
   // (threshold -> tier size), so tier pricing stays fresh on each sync instead of
-  // being frozen at hand-authored values. Fall back to the existing tiers only
-  // when the API reports none.
+  // being frozen at hand-authored values. When the API reports usable pricing,
+  // its tier list is authoritative; fall back to the existing cost only when the
+  // source pricing is absent or unusable.
   const tiers = [...(model.pricing?.tiers ?? [])].sort(
     (a, b) => (a.threshold ?? 0) - (b.threshold ?? 0),
   );
@@ -210,7 +211,7 @@ export function buildCrossModel(
     .filter((entry): entry is NonNullable<typeof entry> => entry !== undefined);
   const cost =
     base !== undefined
-      ? { ...base, tiers: contextTiers.length > 0 ? contextTiers : existing?.cost?.tiers }
+      ? { ...base, tiers: contextTiers.length > 0 ? contextTiers : undefined }
       : existing?.cost;
 
   // Every served model reports a context window; without one (and no existing
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 0281aac4808..3da3fe65f8f 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -449,6 +449,38 @@ test("syncs CrossModel's structured-output capability", () => {
   });
 });
 
+test("clears stale CrossModel context tiers only when source pricing is usable", () => {
+  const existing: ExistingModel = {
+    base_model: "alibaba/qwen3.8-max",
+    cost: {
+      input: 9,
+      output: 27,
+      tiers: [
+        {
+          tier: { type: "context", size: 200_000 },
+          input: 18,
+          output: 54,
+        },
+      ],
+    },
+  };
+
+  const authoritative = buildCrossModel(crossModelModel(), existing);
+  const absent = buildCrossModel(crossModelModel({ pricing: undefined }), existing);
+  const unusable = buildCrossModel(
+    crossModelModel({
+      pricing: {
+        tiers: [{ threshold: 0, input_micro_per_1m: 1_880_000 }],
+      },
+    }),
+    existing,
+  );
+
+  expect(authoritative?.cost).toEqual({ input: 1.88, output: 5.63 });
+  expect(absent?.cost).toEqual(existing.cost);
+  expect(unusable?.cost).toEqual(existing.cost);
+});
+
 test("parses CrossModel's nullable reasoning controls", () => {
   const parsed = CrossModelResponse.parse({
     data: [

From a7332d5569a1398c8177ec7e4db3b7c5d3dc0677 Mon Sep 17 00:00:00 2001
From: Arpit Mishra <63418443+ArpitMishra17@users.noreply.github.com>
Date: Tue, 22 Sep 2026 00:08:49 +0530
Subject: [PATCH 1820/2012] feat(stepfun-ai): add step-5-preview (#7657)

StepFun's flagship base model, verified live on the Global PAYG gateway
(api.stepfun.ai/v1). Reuses existing lab metadata
(models/stepfun/step-5-preview.toml): 1M-token context/input, text/image/video
input, reasoning + tool calling + structured output, closed weights.
Provider override: reasoning_effort accepts low/medium/high; reasoning side
channel is reasoning_content; per-token cost input 1.00, output 2.70,
cache_read 0.05 USD/MTok.

Standalone file (not symlinked to providers/stepfun) because China
(api.stepfun.com) is unverified with a Global key (401). Global Step Plan
(step_plan/v1) verified absent for this model, so not added there.

Sources:
- https://platform.stepfun.ai/docs/en/guides/models/step-5-preview
- https://platform.stepfun.ai/docs/en/guides/pricing/details
- GET https://api.stepfun.ai/v1/models (accessed 2026-09-21)

bun validate passes.
---
 .../stepfun-ai/models/step-5-preview.toml     | 21 +++++++++++++++++++
 providers/stepfun-ai/provider.toml            |  4 ++--
 2 files changed, 23 insertions(+), 2 deletions(-)
 create mode 100644 providers/stepfun-ai/models/step-5-preview.toml

diff --git a/providers/stepfun-ai/models/step-5-preview.toml b/providers/stepfun-ai/models/step-5-preview.toml
new file mode 100644
index 00000000000..c7a645ba40a
--- /dev/null
+++ b/providers/stepfun-ai/models/step-5-preview.toml
@@ -0,0 +1,21 @@
+# Chat `reasoning_effort`, Messages `output_config.effort`, and Responses
+# `reasoning.effort` accept low/medium/high (accessed 2026-09-21).
+# https://platform.stepfun.ai/docs/en/guides/models/step-5-preview
+# Verified live: GET https://api.stepfun.ai/v1/models lists step-5-preview
+# with reasoning_effort_support_list ["low", "medium", "high"] and
+# max_input_tokens 1024000; POST /v1/chat/completions succeeds.
+# Pricing (USD/MTok): input 1.00, output 2.70, cache_read 0.05 (95% cache
+# discount), matching existing mirrors and vendor reporting.
+base_model = "stepfun/step-5-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1
+output = 2.7
+cache_read = 0.05
diff --git a/providers/stepfun-ai/provider.toml b/providers/stepfun-ai/provider.toml
index 0c84a7aaa5d..84b890902aa 100644
--- a/providers/stepfun-ai/provider.toml
+++ b/providers/stepfun-ai/provider.toml
@@ -4,8 +4,8 @@ npm = "@ai-sdk/openai-compatible"
 # Reasoning HTTP format (accessed 2026-06-29):
 # POST /v1/chat/completions uses top-level `reasoning_effort`; POST /v1/messages
 # uses `output_config.effort`; POST /v1/responses uses `reasoning.effort`.
-# Values are low/medium/high for step-3.7-flash; step-3.5-flash-2603 accepts
-# low/high. Responses supports only step-3.7-flash. Chat returns reasoning at
+# Values are low/medium/high for step-3.7-flash and step-5-preview; step-3.5-flash-2603 accepts
+# low/high. Responses supports step-3.7-flash and step-5-preview. Chat returns reasoning at
 # `choices[].message.reasoning` or streamed `choices[].delta.reasoning`;
 # `reasoning_format` is general (default), the latter using `reasoning_content`.
 # Responses streams `response.reasoning_text.delta`.

From 36a09cd2cdb684821d660f5b9062ca941b6a36f7 Mon Sep 17 00:00:00 2001
From: thatdevguy <46411187+chrissalomon@users.noreply.github.com>
Date: Mon, 21 Sep 2026 14:39:05 -0400
Subject: [PATCH 1821/2012] Add Tempr as a provider, with its Anthropic models
 (#7638)

Tempr's AI Gateway (api.temprhq.io/v1) is a BYOK, OpenAI-compatible
relay: requests go to the caller's own key on the named provider,
with no markup on tokens. These 14 Anthropic models base_model the
lab entries with Anthropic's own pricing; reasoning_options are what
Tempr's GET /v1/models reports, with a leading comment giving the
fields a caller sets.
---
 providers/tempr/logo.svg                      |  4 ++++
 .../models/anthropic/claude-fable-5-1.toml    | 16 +++++++++++++++
 .../models/anthropic/claude-fable-5.toml      | 16 +++++++++++++++
 .../anthropic/claude-haiku-4-5-20251001.toml  | 18 +++++++++++++++++
 .../models/anthropic/claude-haiku-4-5.toml    | 18 +++++++++++++++++
 .../anthropic/claude-opus-4-5-20251101.toml   | 20 +++++++++++++++++++
 .../models/anthropic/claude-opus-4-5.toml     | 20 +++++++++++++++++++
 .../models/anthropic/claude-opus-4-6.toml     | 20 +++++++++++++++++++
 .../models/anthropic/claude-opus-4-7.toml     | 18 +++++++++++++++++
 .../models/anthropic/claude-opus-4-8.toml     | 18 +++++++++++++++++
 .../tempr/models/anthropic/claude-opus-5.toml | 16 +++++++++++++++
 .../anthropic/claude-sonnet-4-5-20250929.toml | 18 +++++++++++++++++
 .../models/anthropic/claude-sonnet-4-5.toml   | 18 +++++++++++++++++
 .../models/anthropic/claude-sonnet-4-6.toml   | 20 +++++++++++++++++++
 .../models/anthropic/claude-sonnet-5.toml     | 18 +++++++++++++++++
 providers/tempr/provider.toml                 |  5 +++++
 16 files changed, 263 insertions(+)
 create mode 100644 providers/tempr/logo.svg
 create mode 100644 providers/tempr/models/anthropic/claude-fable-5-1.toml
 create mode 100644 providers/tempr/models/anthropic/claude-fable-5.toml
 create mode 100644 providers/tempr/models/anthropic/claude-haiku-4-5-20251001.toml
 create mode 100644 providers/tempr/models/anthropic/claude-haiku-4-5.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-4-5-20251101.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-4-5.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-4-6.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-4-7.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-4-8.toml
 create mode 100644 providers/tempr/models/anthropic/claude-opus-5.toml
 create mode 100644 providers/tempr/models/anthropic/claude-sonnet-4-5-20250929.toml
 create mode 100644 providers/tempr/models/anthropic/claude-sonnet-4-5.toml
 create mode 100644 providers/tempr/models/anthropic/claude-sonnet-4-6.toml
 create mode 100644 providers/tempr/models/anthropic/claude-sonnet-5.toml
 create mode 100644 providers/tempr/provider.toml

diff --git a/providers/tempr/logo.svg b/providers/tempr/logo.svg
new file mode 100644
index 00000000000..a9fb18e6892
--- /dev/null
+++ b/providers/tempr/logo.svg
@@ -0,0 +1,4 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
+  <rect x="10" y="3" width="4" height="12" rx="2"/>
+  <circle cx="12" cy="18" r="4"/>
+</svg>
diff --git a/providers/tempr/models/anthropic/claude-fable-5-1.toml b/providers/tempr/models/anthropic/claude-fable-5-1.toml
new file mode 100644
index 00000000000..78d27e34265
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-fable-5-1.toml
@@ -0,0 +1,16 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = low|medium|high|xhigh|max
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 10
+output = 50
+cache_read = 0.25
+cache_write = 12.5
diff --git a/providers/tempr/models/anthropic/claude-fable-5.toml b/providers/tempr/models/anthropic/claude-fable-5.toml
new file mode 100644
index 00000000000..03847400952
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-fable-5.toml
@@ -0,0 +1,16 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = low|medium|high|xhigh|max
+base_model = "anthropic/claude-fable-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
diff --git a/providers/tempr/models/anthropic/claude-haiku-4-5-20251001.toml b/providers/tempr/models/anthropic/claude-haiku-4-5-20251001.toml
new file mode 100644
index 00000000000..c2502f29a34
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-haiku-4-5-20251001.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high (sized to a thinking budget)
+base_model = "anthropic/claude-haiku-4-5-20251001"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/tempr/models/anthropic/claude-haiku-4-5.toml b/providers/tempr/models/anthropic/claude-haiku-4-5.toml
new file mode 100644
index 00000000000..102da98f92e
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-haiku-4-5.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high (sized to a thinking budget)
+base_model = "anthropic/claude-haiku-4-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/tempr/models/anthropic/claude-opus-4-5-20251101.toml b/providers/tempr/models/anthropic/claude-opus-4-5-20251101.toml
new file mode 100644
index 00000000000..1ff845fa823
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-4-5-20251101.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high (alias: top-level reasoning_effort)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high
+base_model = "anthropic/claude-opus-4-5-20251101"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high"] },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-opus-4-5.toml b/providers/tempr/models/anthropic/claude-opus-4-5.toml
new file mode 100644
index 00000000000..35b1bae9829
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-4-5.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high (alias: top-level reasoning_effort)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high
+base_model = "anthropic/claude-opus-4-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high"] },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-opus-4-6.toml b/providers/tempr/models/anthropic/claude-opus-4-6.toml
new file mode 100644
index 00000000000..bd4664a3a5f
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-4-6.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high|max (alias: top-level reasoning_effort)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high|max; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high|max
+base_model = "anthropic/claude-opus-4-6"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high", "max"] },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-opus-4-7.toml b/providers/tempr/models/anthropic/claude-opus-4-7.toml
new file mode 100644
index 00000000000..5796ecdbe51
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-4-7.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = none|low|medium|high|xhigh|max
+base_model = "anthropic/claude-opus-4-7"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-opus-4-8.toml b/providers/tempr/models/anthropic/claude-opus-4-8.toml
new file mode 100644
index 00000000000..23f9c6ab18e
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-4-8.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = none|low|medium|high|xhigh|max
+base_model = "anthropic/claude-opus-4-8"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-opus-5.toml b/providers/tempr/models/anthropic/claude-opus-5.toml
new file mode 100644
index 00000000000..ca74b5d1101
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-opus-5.toml
@@ -0,0 +1,16 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = low|medium|high|xhigh|max
+base_model = "anthropic/claude-opus-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/tempr/models/anthropic/claude-sonnet-4-5-20250929.toml b/providers/tempr/models/anthropic/claude-sonnet-4-5-20250929.toml
new file mode 100644
index 00000000000..8cda1b84187
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-sonnet-4-5-20250929.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high (sized to a thinking budget)
+base_model = "anthropic/claude-sonnet-4-5-20250929"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/tempr/models/anthropic/claude-sonnet-4-5.toml b/providers/tempr/models/anthropic/claude-sonnet-4-5.toml
new file mode 100644
index 00000000000..aba63782362
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-sonnet-4-5.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high (sized to a thinking budget)
+base_model = "anthropic/claude-sonnet-4-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/tempr/models/anthropic/claude-sonnet-4-6.toml b/providers/tempr/models/anthropic/claude-sonnet-4-6.toml
new file mode 100644
index 00000000000..396e2f09c38
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-sonnet-4-6.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high|max (alias: top-level reasoning_effort)
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high|max; thinking.budget_tokens
+# /v1/responses: reasoning.effort = none|low|medium|high|max
+base_model = "anthropic/claude-sonnet-4-6"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high", "max"] },
+  { type = "budget_tokens", min = 1024 },
+]
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/tempr/models/anthropic/claude-sonnet-5.toml b/providers/tempr/models/anthropic/claude-sonnet-5.toml
new file mode 100644
index 00000000000..a7908c1fc7c
--- /dev/null
+++ b/providers/tempr/models/anthropic/claude-sonnet-5.toml
@@ -0,0 +1,18 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# Effort: reasoning.effort = none|low|medium|high|xhigh|max (alias: top-level reasoning_effort)
+# /v1/messages: thinking.type = enabled|disabled; output_config.effort = low|medium|high|xhigh|max
+# /v1/responses: reasoning.effort = none|low|medium|high|xhigh|max
+base_model = "anthropic/claude-sonnet-5"
+structured_output = true
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/tempr/provider.toml b/providers/tempr/provider.toml
new file mode 100644
index 00000000000..70ceee60f53
--- /dev/null
+++ b/providers/tempr/provider.toml
@@ -0,0 +1,5 @@
+name = "Tempr"
+npm = "@ai-sdk/openai-compatible"
+env = ["TEMPR_API_KEY"]
+api = "https://api.temprhq.io/v1"
+doc = "https://temprhq.io/docs/gateway-reference.html"

From fa5f6eadf9b90935f601e06ab46a806fd04c969c Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 13:40:54 -0500
Subject: [PATCH 1822/2012] fix: [missing-model] github-copilot: grok-4.7
 (#7665)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/github-copilot/models/grok-4.7.toml | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)
 create mode 100644 providers/github-copilot/models/grok-4.7.toml

diff --git a/providers/github-copilot/models/grok-4.7.toml b/providers/github-copilot/models/grok-4.7.toml
new file mode 100644
index 00000000000..25d6bf546e2
--- /dev/null
+++ b/providers/github-copilot/models/grok-4.7.toml
@@ -0,0 +1,24 @@
+# Sources:
+# - https://github.blog/changelog/2026-09-21-grok-4-7-is-now-available-in-github-copilot/
+# - https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# - https://docs.github.com/en/copilot/reference/ai-models/supported-models
+# - https://docs.x.ai/developers/models/grok-4.7
+# - https://docs.x.ai/developers/pricing
+base_model = "xai/grok-4.7"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+
+[limit]
+context = 500_000
+input = 372_000
+output = 128_000

From 9312e146098c165f7ac3e418be02d2a13cf0fff8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 19:24:31 +0000
Subject: [PATCH 1823/2012] chore(sync): update Kilo model catalog (#7667)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 -
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index d4d63d57a55..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 191b436787a..6caf71dbf01 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.56628
-output = 1.69884
-cache_read = 0.018018
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From c8cf3b64cae40e537cfae939318c95eebb6785c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 19:25:24 +0000
Subject: [PATCH 1824/2012] chore(sync): update OpenRouter model catalog
 (#7668)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c11c30e3519..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56628
-output = 1.69884
-cache_read = 0.018018
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 393_216
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 399e63186f7..5af84d1accb 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.924462
-output = 1.848924
-cache_read = 0.077039
+input = 0.922722
+output = 1.845444
+cache_read = 0.076894
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 427614e0228..1db14b9b08b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56628
-output = 1.69884
-cache_read = 0.018018
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From cdd70092cb739402465d19122438bd656473752e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 19:26:14 +0000
Subject: [PATCH 1825/2012] chore(sync): update Merge Gateway model catalog
 (#7666)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml | 7 ++++---
 providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml | 7 ++++---
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
index b86a0e589cd..5ef98e7a3a0 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-122b-a10b.toml
@@ -1,6 +1,7 @@
 # Source: https://api-gateway.merge.dev/v1/models?model=qwen%2Fqwen3.5-122b-a10b (accessed 2026-07-21)
 base_model = "alibaba/qwen3.5-122b-a10b"
 name = "Qwen3.5 122B A10B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -11,8 +12,8 @@ output = 0.917
 cache_read = 0.023
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]
diff --git a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
index 0264c6bc733..84592d1b8dd 100644
--- a/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
+++ b/providers/merge-gateway/models/qwen/qwen3.5-397b-a17b.toml
@@ -2,6 +2,7 @@
 # Selected route reasoning.controls = ["thinking"], disable_supported = true.
 base_model = "alibaba/qwen3.5-397b-a17b"
 name = "Qwen3.5 397B A17B"
+attachment = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -12,8 +13,8 @@ output = 1.032
 cache_read = 0.0344
 
 [limit]
-context = 256_000
-output = 64_000
+context = 131_072
+output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text"]

From 3feb2f1df5f738390e575795c5c0e1b8134f2bf4 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Tue, 22 Sep 2026 03:45:13 +0800
Subject: [PATCH 1826/2012] feat: add MiMo V2.6 models

---
 models/xiaomi/mimo-v2.6-flash.toml            | 19 +++++++++++++++++++
 models/xiaomi/mimo-v2.6-pro.toml              | 19 +++++++++++++++++++
 .../opencode-go/models/mimo-v2.6-flash.toml   | 14 ++++++++++++++
 .../opencode-go/models/mimo-v2.6-pro.toml     | 14 ++++++++++++++
 .../opencode/models/mimo-v2.6-flash-free.toml | 15 +++++++++++++++
 5 files changed, 81 insertions(+)
 create mode 100644 models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 models/xiaomi/mimo-v2.6-pro.toml
 create mode 100644 providers/opencode-go/models/mimo-v2.6-flash.toml
 create mode 100644 providers/opencode-go/models/mimo-v2.6-pro.toml
 create mode 100644 providers/opencode/models/mimo-v2.6-flash-free.toml

diff --git a/models/xiaomi/mimo-v2.6-flash.toml b/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..3c23f154d83
--- /dev/null
+++ b/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,19 @@
+name = "MiMo-V2.6-Flash"
+description = "MiMo Flash model for multimodal coding agents and long-context automation"
+family = "mimo"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-12"
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "audio", "video"]
+output = ["text"]
diff --git a/models/xiaomi/mimo-v2.6-pro.toml b/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..43fd862667d
--- /dev/null
+++ b/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,19 @@
+name = "MiMo-V2.6-Pro"
+description = "Stronger MiMo Pro tier for multimodal reasoning and coding-agent execution"
+family = "mimo"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-12"
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/opencode-go/models/mimo-v2.6-flash.toml b/providers/opencode-go/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..018c58c0b68
--- /dev/null
+++ b/providers/opencode-go/models/mimo-v2.6-flash.toml
@@ -0,0 +1,14 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_000_000
+output = 128_000
diff --git a/providers/opencode-go/models/mimo-v2.6-pro.toml b/providers/opencode-go/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..40f32a90e8b
--- /dev/null
+++ b/providers/opencode-go/models/mimo-v2.6-pro.toml
@@ -0,0 +1,14 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+attachment = true
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.003625
+
+[limit]
+output = 128_000
diff --git a/providers/opencode/models/mimo-v2.6-flash-free.toml b/providers/opencode/models/mimo-v2.6-flash-free.toml
new file mode 100644
index 00000000000..753e0075aa2
--- /dev/null
+++ b/providers/opencode/models/mimo-v2.6-flash-free.toml
@@ -0,0 +1,15 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+name = "MiMo-V2.6-Flash Free"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+
+[limit]
+context = 200_000
+output = 32_000

From aa6265846839066b9eb7019b406924ad517ee242 Mon Sep 17 00:00:00 2001
From: Jack <jack@anoma.ly>
Date: Tue, 22 Sep 2026 03:48:04 +0800
Subject: [PATCH 1827/2012] Update knowledge date to 2026-09-22

---
 models/xiaomi/mimo-v2.6-pro.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/models/xiaomi/mimo-v2.6-pro.toml b/models/xiaomi/mimo-v2.6-pro.toml
index 43fd862667d..85d587e9f63 100644
--- a/models/xiaomi/mimo-v2.6-pro.toml
+++ b/models/xiaomi/mimo-v2.6-pro.toml
@@ -7,7 +7,7 @@ attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2024-12"
+knowledge = "2026-09-22"
 open_weights = false
 
 [limit]

From 546269d6e0b90b8b825528f1b1a66c76af4eae8c Mon Sep 17 00:00:00 2001
From: Jack <jack@anoma.ly>
Date: Tue, 22 Sep 2026 03:49:20 +0800
Subject: [PATCH 1828/2012] Update knowledge date and input modalities

---
 models/xiaomi/mimo-v2.6-flash.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/models/xiaomi/mimo-v2.6-flash.toml b/models/xiaomi/mimo-v2.6-flash.toml
index 3c23f154d83..f065c84b627 100644
--- a/models/xiaomi/mimo-v2.6-flash.toml
+++ b/models/xiaomi/mimo-v2.6-flash.toml
@@ -7,7 +7,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2024-12"
+knowledge = "2026-09-22"
 open_weights = false
 
 [limit]
@@ -15,5 +15,5 @@ context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "audio", "video"]
+input = ["text", "image", "audio", "video", "pdf"]
 output = ["text"]

From 98646c1bf4091865e478e2bcd4f809ba57ca4340 Mon Sep 17 00:00:00 2001
From: Jack <jack@anoma.ly>
Date: Tue, 22 Sep 2026 03:50:14 +0800
Subject: [PATCH 1829/2012] Enable attachment and expand input modalities

Updated attachment status and input modalities.
---
 models/xiaomi/mimo-v2.6-pro.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/models/xiaomi/mimo-v2.6-pro.toml b/models/xiaomi/mimo-v2.6-pro.toml
index 85d587e9f63..fea580df360 100644
--- a/models/xiaomi/mimo-v2.6-pro.toml
+++ b/models/xiaomi/mimo-v2.6-pro.toml
@@ -3,7 +3,7 @@ description = "Stronger MiMo Pro tier for multimodal reasoning and coding-agent
 family = "mimo"
 release_date = "2026-09-22"
 last_updated = "2026-09-22"
-attachment = false
+attachment = true
 reasoning = true
 temperature = true
 tool_call = true
@@ -15,5 +15,5 @@ context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text"]
+input = ["text", "image", "audio", "video", "pdf"]
 output = ["text"]

From 093bb9e02a61b81a3cd5e95f0bfcc71d3add7b76 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Tue, 22 Sep 2026 04:02:39 +0800
Subject: [PATCH 1830/2012] fix: remove redundant MiMo override

---
 providers/opencode-go/models/mimo-v2.6-pro.toml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/providers/opencode-go/models/mimo-v2.6-pro.toml b/providers/opencode-go/models/mimo-v2.6-pro.toml
index 40f32a90e8b..82f9688f50f 100644
--- a/providers/opencode-go/models/mimo-v2.6-pro.toml
+++ b/providers/opencode-go/models/mimo-v2.6-pro.toml
@@ -1,5 +1,4 @@
 base_model = "xiaomi/mimo-v2.6-pro"
-attachment = true
 reasoning_options = []
 
 [interleaved]

From 544b829052f24ce5dbef80d498a6e9823aa6426d Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Tue, 22 Sep 2026 04:04:52 +0800
Subject: [PATCH 1831/2012] fix: inherit MiMo Go limits

---
 providers/opencode-go/models/mimo-v2.6-flash.toml | 4 ----
 providers/opencode-go/models/mimo-v2.6-pro.toml   | 3 ---
 2 files changed, 7 deletions(-)

diff --git a/providers/opencode-go/models/mimo-v2.6-flash.toml b/providers/opencode-go/models/mimo-v2.6-flash.toml
index 018c58c0b68..06b0dcf1faa 100644
--- a/providers/opencode-go/models/mimo-v2.6-flash.toml
+++ b/providers/opencode-go/models/mimo-v2.6-flash.toml
@@ -8,7 +8,3 @@ field = "reasoning_content"
 input = 0.14
 output = 0.28
 cache_read = 0.0028
-
-[limit]
-context = 1_000_000
-output = 128_000
diff --git a/providers/opencode-go/models/mimo-v2.6-pro.toml b/providers/opencode-go/models/mimo-v2.6-pro.toml
index 82f9688f50f..7513e538ec8 100644
--- a/providers/opencode-go/models/mimo-v2.6-pro.toml
+++ b/providers/opencode-go/models/mimo-v2.6-pro.toml
@@ -8,6 +8,3 @@ field = "reasoning_content"
 input = 0.435
 output = 0.87
 cache_read = 0.003625
-
-[limit]
-output = 128_000

From 702c1c4c6d93f358871143b8e5c94bb49a474160 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 20:26:21 +0000
Subject: [PATCH 1832/2012] chore(sync): update OpenRouter model catalog
 (#7673)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/xiaomi/mimo-v2.6-flash.toml        | 16 ++++++++++
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml      | 29 +++++++++++++++++++
 .../models/xiaomi/mimo-v2.6-pro.toml          | 16 ++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 5 files changed, 67 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
 create mode 100644 providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 5af84d1accb..81e8b644948 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.922722
-output = 1.845444
-cache_read = 0.076894
+input = 0.91263
+output = 1.82526
+cache_read = 0.076053
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..3e52e783fa7
--- /dev/null
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,16 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "xiaomi/mimo-v2.6-flash"
+description = "MiMo flash model for fast multimodal assistance and agent workflows"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..40cfae78edb
--- /dev/null
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,29 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "MiMo-V2.6-Pro-UltraSpeed"
+description = "MiMo pro model for strong multimodal reasoning and agent execution"
+family = "mimo"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 4.35
+output = 8.7
+cache_read = 0.036
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "audio"]
+output = ["text"]
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..2f40e81f221
--- /dev/null
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,16 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "xiaomi/mimo-v2.6-pro"
+description = "MiMo pro model for strong multimodal reasoning and agent execution"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1db14b9b08b..1c92257a1ff 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.5544
+output = 1.6632
+cache_read = 0.01848
 
 [limit]
 context = 1_048_576

From 256a8e05011870931c8ed2bf092b0691bb01928d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 20:26:34 +0000
Subject: [PATCH 1833/2012] chore(sync): update Kilo model catalog (#7672)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/xiaomi/mimo-v2.6-flash.toml | 15 +++++++++++++++
 providers/kilo/models/xiaomi/mimo-v2.6-pro.toml   | 15 +++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml     |  6 +++---
 3 files changed, 33 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/kilo/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml b/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..b4165fb24ad
--- /dev/null
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,15 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+description = "MiMo-V2.6-Flash is an open-source foundation model developed by Xiaomi. Built on a Mixture-of-Experts architecture with 309B total parameters and 15B activated per token, it employs a hybrid attention mechanism for..."
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml b/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..e3e0a47689b
--- /dev/null
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,15 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+description = "MiMo-V2.6-Pro is the flagship foundation model developed by Xiaomi. Built at a scale of over 1T parameters, it is designed to push the ceiling of capability for the most demanding..."
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6caf71dbf01..b5bfc21fba2 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.5544
+output = 1.6632
+cache_read = 0.01848
 
 [limit]
 context = 1_024_000

From ba6fdc8e65d2b62122b9f888e0c79eb8380b52a1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 20:26:43 +0000
Subject: [PATCH 1834/2012] chore(sync): update Deep Infra model catalog
 (#7671)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/deepinfra/models/tencent/Hy3.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/deepinfra/models/tencent/Hy3.toml b/providers/deepinfra/models/tencent/Hy3.toml
index 141b04fc694..743a6959ec1 100644
--- a/providers/deepinfra/models/tencent/Hy3.toml
+++ b/providers/deepinfra/models/tencent/Hy3.toml
@@ -3,9 +3,9 @@ structured_output = true
 reasoning_options = []
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.13
+output = 0.53
+cache_read = 0.033
 
 [limit]
 context = 262_144

From 38cb3a87413159c919b6616ebeebb9280c7ee0d6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:25:20 +0000
Subject: [PATCH 1835/2012] chore(sync): update Kilo model catalog (#7676)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml      | 28 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 2 files changed, 31 insertions(+), 3 deletions(-)
 create mode 100644 providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml

diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..08e541fac0d
--- /dev/null
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,28 @@
+name = "Xiaomi: MiMo-V2.6-Pro-UltraSpeed"
+description = "MiMo-V2.6-Pro-UltraSpeed is the fast speed edition of Xiaomi's flagship foundation model, MiMo-V2.6-Pro. Built from the same 1T MiMo-V2.6-Pro checkpoint, it matches the original model in quality while delivering roughly 10x..."
+family = "mimo"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 4.35
+output = 8.7
+cache_read = 0.036
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "audio"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index b5bfc21fba2..3563cca9cda 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.5544
-output = 1.6632
-cache_read = 0.01848
+input = 0.54648
+output = 1.63944
+cache_read = 0.018216
 
 [limit]
 context = 1_024_000

From 276327558e857b8f3340572d2cd94a089031e1c5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:25:24 +0000
Subject: [PATCH 1836/2012] chore(sync): update NanoGPT model catalog (#7677)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/x-ai/grok-4.7.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/nano-gpt/models/x-ai/grok-4.7.toml

diff --git a/providers/nano-gpt/models/x-ai/grok-4.7.toml b/providers/nano-gpt/models/x-ai/grok-4.7.toml
new file mode 100644
index 00000000000..81b4067b343
--- /dev/null
+++ b/providers/nano-gpt/models/x-ai/grok-4.7.toml
@@ -0,0 +1,16 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.6
+output = 4.8
+cache_read = 0.4
+
+[limit]
+input = 500_000
+
+[modalities]
+input = ["text", "image"]

From 5d374d40b34fd48c28957a94cd6b397419a3cfd8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:25:27 +0000
Subject: [PATCH 1837/2012] chore(sync): update OpenRouter model catalog
 (#7679)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 81e8b644948..4a55e5fac13 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.91263
-output = 1.82526
-cache_read = 0.076053
+input = 0.900798
+output = 1.801596
+cache_read = 0.075067
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1c92257a1ff..37d1b27ede8 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5544
-output = 1.6632
-cache_read = 0.01848
+input = 0.54648
+output = 1.63944
+cache_read = 0.018216
 
 [limit]
 context = 1_048_576

From deef71c3c18f440764ba19d58a2dc462579cd90e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 22:25:50 +0000
Subject: [PATCH 1838/2012] chore(sync): update Kilo model catalog (#7683)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/z-ai/glm-5.3-flash.toml               | 2 +-
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 2 +-
 providers/kilo/models/~moonshotai/kimi-latest.toml          | 6 +++---
 4 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/z-ai/glm-5.3-flash.toml b/providers/kilo/models/z-ai/glm-5.3-flash.toml
index 1bc5d8b7a69..c464db465b8 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flash.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flash.toml
@@ -12,7 +12,7 @@ cache_read = 0.03
 
 [limit]
 context = 1_048_576
-output = 102_400
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 3563cca9cda..85d8d21cda2 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.54648
-output = 1.63944
-cache_read = 0.018216
+input = 0.53856
+output = 1.61568
+cache_read = 0.017952
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 49d602e67e6..8aac86d2dac 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.2
+output = 0.4
 cache_read = 0.01
 
 [limit]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 631c1758bcd..afff7105d92 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.5
+output = 7.5
+cache_read = 0.15
 
 [limit]
 context = 1_048_576

From 59955f4459ca03e3b7c824021ff4d502be6cba17 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 22:25:53 +0000
Subject: [PATCH 1839/2012] chore(sync): update OpenRouter model catalog
 (#7682)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-0731.toml           | 2 +-
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3-flash.toml       | 8 ++++----
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 2 +-
 providers/openrouter/models/~moonshotai/kimi-latest.toml  | 6 +++---
 6 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
index 48200f680e3..3de935d9215 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash-0731.toml
@@ -11,7 +11,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.32
+output = 0.64
 cache_read = 0.016
 
 [limit]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 4a55e5fac13..43ba15830d0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.900798
-output = 1.801596
-cache_read = 0.075067
+input = 0.892272
+output = 1.784544
+cache_read = 0.074356
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3-flash.toml b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
index 6fd3f066cfd..188a84f8fb5 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flash.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flash.toml
@@ -6,13 +6,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.075
-output = 0.25
-cache_read = 0.02
+input = 0.15
+output = 0.5
+cache_read = 0.05
 
 [limit]
 context = 1_310_720
-output = 102_400
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 37d1b27ede8..3783e77eceb 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.54648
-output = 1.63944
-cache_read = 0.018216
+input = 0.53856
+output = 1.61568
+cache_read = 0.017952
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 867340e2671..6d98cbc2838 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.04
-output = 0.2
+output = 0.4
 cache_read = 0.01
 
 [limit]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 7031b748f98..0bb8650abd5 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.7
-output = 8.5
-cache_read = 0.17
+input = 1.5
+output = 7.5
+cache_read = 0.15
 
 [limit]
 context = 1_048_576

From 64964841a65dbd1f262ede752a702ec018ef3084 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 22:26:20 +0000
Subject: [PATCH 1840/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7681)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../empiriolabs/models/mimo-v2-6-flash.toml     | 17 +++++++++++++++++
 providers/empiriolabs/models/mimo-v2-6-pro.toml | 17 +++++++++++++++++
 2 files changed, 34 insertions(+)
 create mode 100644 providers/empiriolabs/models/mimo-v2-6-flash.toml
 create mode 100644 providers/empiriolabs/models/mimo-v2-6-pro.toml

diff --git a/providers/empiriolabs/models/mimo-v2-6-flash.toml b/providers/empiriolabs/models/mimo-v2-6-flash.toml
new file mode 100644
index 00000000000..ab796f4d7ef
--- /dev/null
+++ b/providers/empiriolabs/models/mimo-v2-6-flash.toml
@@ -0,0 +1,17 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+name = "MiMo V2.6 Flash"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.7
+output = 1.4
+cache_read = 0.7
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/empiriolabs/models/mimo-v2-6-pro.toml b/providers/empiriolabs/models/mimo-v2-6-pro.toml
new file mode 100644
index 00000000000..04dcc7ceb0e
--- /dev/null
+++ b/providers/empiriolabs/models/mimo-v2-6-pro.toml
@@ -0,0 +1,17 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+name = "MiMo V2.6 Pro"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 2.175
+output = 4.35
+cache_read = 2.175
+
+[limit]
+context = 1_000_000
+
+[modalities]
+input = ["text", "image", "video", "audio"]

From d51cf7c4cc7bb39fe5e82247625d161f9f120376 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:25:04 +0000
Subject: [PATCH 1841/2012] chore(sync): update Kilo model catalog (#7685)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/meta-llama/llama-4-maverick.toml |  2 +-
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/meta-llama/llama-4-maverick.toml b/providers/kilo/models/meta-llama/llama-4-maverick.toml
index 537ab8cd5fd..fab1dd0fc03 100644
--- a/providers/kilo/models/meta-llama/llama-4-maverick.toml
+++ b/providers/kilo/models/meta-llama/llama-4-maverick.toml
@@ -15,7 +15,7 @@ input = 0.1875
 output = 0.6525
 
 [limit]
-context = 1_048_576
+context = 128_000
 output = 16_384
 
 [modalities]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 85d8d21cda2..022394b5a93 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.53856
-output = 1.61568
-cache_read = 0.017952
+input = 0.53196
+output = 1.59588
+cache_read = 0.016926
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]

From 32107e3c9042bc6ee3c1d9ed21a2b92e738621cf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:25:07 +0000
Subject: [PATCH 1842/2012] chore(sync): update OpenRouter model catalog
 (#7684)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/meta-llama/llama-4-maverick.toml    | 4 ++--
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 43ba15830d0..3cb613ddd30 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.892272
-output = 1.784544
-cache_read = 0.074356
+input = 0.883746
+output = 1.767492
+cache_read = 0.073646
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/meta-llama/llama-4-maverick.toml b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
index de6e5240ff9..b9b42749c83 100644
--- a/providers/openrouter/models/meta-llama/llama-4-maverick.toml
+++ b/providers/openrouter/models/meta-llama/llama-4-maverick.toml
@@ -12,8 +12,8 @@ knowledge = "2024-08-31"
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.8
+input = 0.1875
+output = 0.6525
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 3783e77eceb..85762922f96 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53856
-output = 1.61568
-cache_read = 0.017952
+input = 0.53196
+output = 1.59588
+cache_read = 0.016926
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From f71cbc84cad92ffdd2bc7b09445c5bdf24c7dc5b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 00:59:36 +0000
Subject: [PATCH 1843/2012] chore(sync): update OpenRouter model catalog
 (#7686)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 7 files changed, 23 insertions(+), 23 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index f9ab26fd006..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.05544
-output = 0.11088
-cache_read = 0.011088
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 3cb613ddd30..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.883746
-output = 1.767492
-cache_read = 0.073646
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 6ca3ad0c4f6..b68509e1e67 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.91
-output = 2.86
-cache_read = 0.169
+input = 0.784
+output = 2.464
+cache_read = 0.1456
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 85762922f96..1db14b9b08b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.53196
-output = 1.59588
-cache_read = 0.016926
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 6d98cbc2838..85c55eed894 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.4
-cache_read = 0.01
+input = 0.03
+output = 1
+cache_read = 0.012
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 709b8881ffa..12ade84d329 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7728
-output = 2.4288
-cache_read = 0.14352
+input = 0.7839
+output = 2.6532
+cache_read = 0.15678
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 235_929
 
 [modalities]
 input = ["text"]

From 1a23b290742855a2960a4422da5c8116b9711618 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 00:59:39 +0000
Subject: [PATCH 1844/2012] chore(sync): update DigitalOcean model catalog
 (#7688)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/digitalocean/models/deepseek-3.2.toml           | 6 +++---
 providers/digitalocean/models/deepseek-4-flash.toml       | 6 +++---
 providers/digitalocean/models/deepseek-v4-flash-0731.toml | 6 +++---
 providers/digitalocean/models/deepseek-v4-pro.toml        | 6 +++---
 providers/digitalocean/models/glm-5.2.toml                | 6 +++---
 providers/digitalocean/models/glm-5.3.toml                | 6 +++---
 providers/digitalocean/models/kimi-k3.toml                | 6 +++---
 providers/digitalocean/models/mimo-v2.5-pro.toml          | 6 +++---
 providers/digitalocean/models/openai-gpt-oss-120b.toml    | 4 ++--
 9 files changed, 26 insertions(+), 26 deletions(-)

diff --git a/providers/digitalocean/models/deepseek-3.2.toml b/providers/digitalocean/models/deepseek-3.2.toml
index 86a0601e001..d8d92fa42fa 100644
--- a/providers/digitalocean/models/deepseek-3.2.toml
+++ b/providers/digitalocean/models/deepseek-3.2.toml
@@ -18,9 +18,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.25
-output = 0.8
-cache_read = 0.075
+input = 0.5
+output = 1.6
+cache_read = 0.15
 
 [limit]
 context = 163_840
diff --git a/providers/digitalocean/models/deepseek-4-flash.toml b/providers/digitalocean/models/deepseek-4-flash.toml
index 2e033308f91..143346a359b 100644
--- a/providers/digitalocean/models/deepseek-4-flash.toml
+++ b/providers/digitalocean/models/deepseek-4-flash.toml
@@ -10,9 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0.0679
-output = 0.168
-cache_read = 0.0168
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-flash-0731.toml b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
index 74ce3b0ae91..2e9e64c5632 100644
--- a/providers/digitalocean/models/deepseek-v4-flash-0731.toml
+++ b/providers/digitalocean/models/deepseek-v4-flash-0731.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.08
-output = 0.252
-cache_read = 0.0252
+input = 0.14
+output = 0.28
+cache_read = 0.028
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/deepseek-v4-pro.toml b/providers/digitalocean/models/deepseek-v4-pro.toml
index 8a4edb98729..b4ebe0207c1 100644
--- a/providers/digitalocean/models/deepseek-v4-pro.toml
+++ b/providers/digitalocean/models/deepseek-v4-pro.toml
@@ -19,9 +19,9 @@ type = "effort"
 values = ["low", "medium", "high", "xhigh"]
 
 [cost]
-input = 0.87
-output = 1.74
-cache_read = 0.174
+input = 1.74
+output = 3.48
+cache_read = 0.348
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/glm-5.2.toml b/providers/digitalocean/models/glm-5.2.toml
index 3f01602c179..eb3b3d1b78d 100644
--- a/providers/digitalocean/models/glm-5.2.toml
+++ b/providers/digitalocean/models/glm-5.2.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["medium", "high", "xhigh"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.105
+input = 1.4
+output = 4.4
+cache_read = 0.21
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/glm-5.3.toml b/providers/digitalocean/models/glm-5.3.toml
index 9ba7811a008..6172a61a5ab 100644
--- a/providers/digitalocean/models/glm-5.3.toml
+++ b/providers/digitalocean/models/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.95
-output = 3.4
-cache_read = 0.2
+input = 1.4
+output = 4.4
+cache_read = 0.26
 
 [limit]
 context = 1_048_576
diff --git a/providers/digitalocean/models/kimi-k3.toml b/providers/digitalocean/models/kimi-k3.toml
index 7533a7423e2..62d0f7b7404 100644
--- a/providers/digitalocean/models/kimi-k3.toml
+++ b/providers/digitalocean/models/kimi-k3.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 2.55
-output = 12.95
-cache_read = 0.285
+input = 3
+output = 15
+cache_read = 0.3
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/digitalocean/models/mimo-v2.5-pro.toml b/providers/digitalocean/models/mimo-v2.5-pro.toml
index 20f6f6db3ea..365235fb39f 100644
--- a/providers/digitalocean/models/mimo-v2.5-pro.toml
+++ b/providers/digitalocean/models/mimo-v2.5-pro.toml
@@ -9,9 +9,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.4
-output = 1.5
-cache_read = 0.08
+input = 0.8
+output = 3
+cache_read = 0.16
 
 [limit]
 context = 262_144
diff --git a/providers/digitalocean/models/openai-gpt-oss-120b.toml b/providers/digitalocean/models/openai-gpt-oss-120b.toml
index fc45d3aac8b..4af4fef37ad 100644
--- a/providers/digitalocean/models/openai-gpt-oss-120b.toml
+++ b/providers/digitalocean/models/openai-gpt-oss-120b.toml
@@ -19,8 +19,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.055
-output = 0.385
+input = 0.1
+output = 0.7
 cache_read = 0.02
 
 [limit]

From a16b62ad1be404f4f10308b71807787b8afa2732 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 00:59:46 +0000
Subject: [PATCH 1845/2012] chore(sync): update Kilo model catalog (#7687)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml                 |  6 +++---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 4 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..c3469c1e9dc 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.13
+output = 0.53
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 022394b5a93..6caf71dbf01 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.53196
-output = 1.59588
-cache_read = 0.016926
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 8aac86d2dac..ca12f499908 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.4
-cache_read = 0.01
+input = 0.03
+output = 1
+cache_read = 0.012
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2237628e2fc..6a2c43df8a6 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7728
-output = 2.4288
-cache_read = 0.14352
+input = 0.7839
+output = 2.6532
+cache_read = 0.15678
 
 [limit]
-context = 1_048_576
-output = 131_072
+context = 262_144
+output = 235_929
 
 [modalities]
 input = ["text"]

From 59590e31fb40ac41a9ec4fb81af471c0e4883bb9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 01:35:29 +0000
Subject: [PATCH 1846/2012] chore(sync): update Kilo model catalog (#7690)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/~deepseek/deepseek-v4-flash-latest.toml     |  4 ++--
 providers/kilo/models/~z-ai/glm-latest.toml            | 10 +++++-----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index ca12f499908..0412502ff5d 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,8 +16,8 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 1
-cache_read = 0.012
+output = 0.8
+cache_read = 0.008
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 6a2c43df8a6..51780ddf2fb 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7839
-output = 2.6532
-cache_read = 0.15678
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
-context = 262_144
-output = 235_929
+context = 1_048_576
+output = 131_072
 
 [modalities]
 input = ["text"]

From d577dc96de1773f5849f093eb1a4c05088ef7276 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 01:35:33 +0000
Subject: [PATCH 1847/2012] chore(sync): update OpenRouter model catalog
 (#7689)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml   | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml        | 2 +-
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 85c55eed894..81420955903 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 1
+output = 0.8
 cache_read = 0.012
 
 [limit]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 12ade84d329..afa82f5ee86 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7839
-output = 2.6532
-cache_read = 0.15678
+input = 0.7826
+output = 2.4596
+cache_read = 0.14534
 
 [limit]
 context = 1_310_720
-output = 235_929
+output = 131_072
 
 [modalities]
 input = ["text"]

From d0a0af81e26a783730ba502b6fbaf7b9ec52714b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 01:35:49 +0000
Subject: [PATCH 1848/2012] chore(sync): update NanoGPT model catalog (#7691)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From 63b9be07808ebb2da0f0c158456d892d56663e6e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:24:58 -0500
Subject: [PATCH 1849/2012] chore(sync): update CoreWeave model catalog (#7643)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek-ai/DeepSeek-V4.1-Flash.toml   | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/wandb/models/deepseek-ai/DeepSeek-V4.1-Flash.toml

diff --git a/providers/wandb/models/deepseek-ai/DeepSeek-V4.1-Flash.toml b/providers/wandb/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
new file mode 100644
index 00000000000..0c25c1c3eec
--- /dev/null
+++ b/providers/wandb/models/deepseek-ai/DeepSeek-V4.1-Flash.toml
@@ -0,0 +1,15 @@
+base_model = "deepseek/deepseek-v4.1-flash"
+description = "DeepSeek V4.1 Flash is a multimodal MoE model for coding, reasoning, and agentic workloads with long contexts."
+family = "deepseek"
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.2
+output = 0.65
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 1_048_576

From 7fb291189f66de31dd1edb34dbc3d25597995d38 Mon Sep 17 00:00:00 2001
From: Divy <divy@coralbricks.ai>
Date: Mon, 21 Sep 2026 19:25:26 -0700
Subject: [PATCH 1850/2012] coralbricks: retire Kimi K3, add DeepSeek V4.1
 Flash, publish cache-write rates (#7599)

* remove(coralbricks): retire Kimi K3

Issue: CoralBricks retired kimi-k3 on 2026-09-20 with a hard 404.
Fix: delete the stale provider model entry.

* add(coralbricks): DeepSeek V4.1 Flash (deepseek-v4.1-flash-fast-fp4)

Public on inference.coralbricks.ai since 2026-09-21: 1M context, $0.22 in /
$0.66 out per 1M tokens, cached input free, cache writes $0.07 (0.3x input),
text only, served as an MXFP4 checkpoint.

* coralbricks: DeepSeek V4.1 Flash at $0.30 / $1.20, declare reasoning options

CoralBricks set the published rate for deepseek-v4.1-flash-fast-fp4 to
$0.30 input / $1.20 output per 1M tokens, with cache writes at $0.09.

The entry inherits reasoning = true from the lab baseline but declared no
reasoning_options, which failed validate. Declare the thinking toggle and
low / high / max effort, each verified against the CoralBricks endpoint.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* coralbricks: DeepSeek V4.1 Flash reasoning via reasoning_effort, with sources

On the CoralBricks endpoint reasoning is off unless reasoning_effort is
set, and thinking.type has no effect, so the control is effort
none | low | high | max rather than a toggle. Adds a header with the
pricing sources and the verified wire path.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* coralbricks: DeepSeek V4.1 Flash interleaved reasoning, no attachments

Reasoning comes back in reasoning_content, as on the other CoralBricks
reasoners, so interleaved = true. The endpoint serves this model
text-only and rejects image parts with a 400, so attachment = false
rather than the base model's true.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* coralbricks: publish cache-write rates for every model

https://www.coralbricks.ai/pricing bills the novel tokens of a request at
the cache-write rate, 1.5x input: GLM 5.3 $1.68, GLM 5.3 Flash $0.23,
GPT-OSS 120B $0.18 per 1M. Add cache_write to those three entries.
DeepSeek V4.1 Flash already carries its $0.09. Its header now states the
effort field only, without claims about the default state.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* coralbricks: DeepSeek V4.1 Flash reasoning is opt-in; add medium effort

Reasoning on this endpoint is opt-in by design: with no reasoning_effort,
or none, the model answers without reasoning; low, medium, high and max
turn it on. State that in the header and declare medium, which the
endpoint accepts.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* coralbricks: DeepSeek V4.1 Flash effort ladder back to low / high / max

The endpoint accepts medium, but acceptance is not a distinct depth; keep
the lab ladder plus the host's none (off).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 .../models/deepseek-v4.1-flash-fast-fp4.toml  | 25 +++++++++++++++++++
 .../coralbricks/models/glm-5.3-flash-fp4.toml |  1 +
 providers/coralbricks/models/glm-5.3-fp4.toml |  1 +
 .../coralbricks/models/gpt-oss-120b.toml      |  1 +
 providers/coralbricks/models/kimi-k3.toml     | 10 --------
 5 files changed, 28 insertions(+), 10 deletions(-)
 create mode 100644 providers/coralbricks/models/deepseek-v4.1-flash-fast-fp4.toml
 delete mode 100644 providers/coralbricks/models/kimi-k3.toml

diff --git a/providers/coralbricks/models/deepseek-v4.1-flash-fast-fp4.toml b/providers/coralbricks/models/deepseek-v4.1-flash-fast-fp4.toml
new file mode 100644
index 00000000000..9c63cb4a2db
--- /dev/null
+++ b/providers/coralbricks/models/deepseek-v4.1-flash-fast-fp4.toml
@@ -0,0 +1,25 @@
+# CoralBricks pricing per https://www.coralbricks.ai/pricing and https://www.coralbricks.ai/api/public/models (accessed 2026-09-21):
+# input $0.30, output $1.20, cache write $0.09 per 1M tokens; cached input is free.
+# Effort: reasoning_effort = none|low|high|max. Reasoning is opt-in: with no reasoning_effort, or none, the model
+# answers without reasoning. thinking.type is ignored by this endpoint.
+# Verified live against POST https://inference.coralbricks.ai/v1/chat/completions (2026-09-21): low, high and max
+# return HTTP 200 with reasoning_content; none and no field return 0 reasoning tokens.
+base_model = "deepseek/deepseek-v4.1-flash"
+name = "DeepSeek V4.1 Flash FP4"
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
+interleaved = true
+attachment = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0
+cache_write = 0.09
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/coralbricks/models/glm-5.3-flash-fp4.toml b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
index a7836490ddd..2c52bfdcb00 100644
--- a/providers/coralbricks/models/glm-5.3-flash-fp4.toml
+++ b/providers/coralbricks/models/glm-5.3-flash-fp4.toml
@@ -12,6 +12,7 @@ output = 131_072
 input = 0.15
 output = 0.5
 cache_read = 0
+cache_write = 0.23
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/coralbricks/models/glm-5.3-fp4.toml b/providers/coralbricks/models/glm-5.3-fp4.toml
index 2488afc97cb..90e669e9d54 100644
--- a/providers/coralbricks/models/glm-5.3-fp4.toml
+++ b/providers/coralbricks/models/glm-5.3-fp4.toml
@@ -12,3 +12,4 @@ output = 131_072
 input = 1.12
 output = 4.4
 cache_read = 0
+cache_write = 1.68
diff --git a/providers/coralbricks/models/gpt-oss-120b.toml b/providers/coralbricks/models/gpt-oss-120b.toml
index 8fd3f30e11e..80b28acc1b5 100644
--- a/providers/coralbricks/models/gpt-oss-120b.toml
+++ b/providers/coralbricks/models/gpt-oss-120b.toml
@@ -8,3 +8,4 @@ interleaved = true
 input = 0.12
 output = 0.6
 cache_read = 0
+cache_write = 0.18
diff --git a/providers/coralbricks/models/kimi-k3.toml b/providers/coralbricks/models/kimi-k3.toml
deleted file mode 100644
index 77c2aabdc7b..00000000000
--- a/providers/coralbricks/models/kimi-k3.toml
+++ /dev/null
@@ -1,10 +0,0 @@
-base_model = "moonshotai/kimi-k3"
-
-reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["minimal", "low", "medium", "high"] }]
-interleaved = true
-
-[cost]
-input = 3.0
-output = 15.0
-cache_read = 0

From c6db867cb5b2b2d601b29a32d5705ab9e78d9685 Mon Sep 17 00:00:00 2001
From: chenxue <17203886+0genlab@users.noreply.github.com>
Date: Tue, 22 Sep 2026 10:26:28 +0800
Subject: [PATCH 1851/2012] feat(aihubmix): add gemini-3.6-flash (#7430)

AIHubMix serves this route but the catalog carried no aihubmix entry for
it. Cost comes from the provider's public model listing and the reasoning
controls from its published model-data index; limit, modalities,
attachment and tool_call inherit from models/google/gemini-3.6-flash.toml rather than being
repeated here.

Co-authored-by: chenxue <DDU1222@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 providers/aihubmix/models/gemini-3.6-flash.toml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/aihubmix/models/gemini-3.6-flash.toml

diff --git a/providers/aihubmix/models/gemini-3.6-flash.toml b/providers/aihubmix/models/gemini-3.6-flash.toml
new file mode 100644
index 00000000000..d9c9722442d
--- /dev/null
+++ b/providers/aihubmix/models/gemini-3.6-flash.toml
@@ -0,0 +1,14 @@
+# Effort: minimal|low|medium|high
+# $.reasoning_effort on /v1/chat/completions (alias $.reasoning.effort, which is also the Responses field);
+# $.output_config.effort on /v1/messages, subject to model support.
+# https://docs.aihubmix.com/cn/api/unified-inference
+base_model = "google/gemini-3.6-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high"]
+
+[cost]
+input = 1.5
+output = 7.5
+cache_read = 0.15

From 226a53f287a2ca8e18803398f321bc14b9b95866 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:29:23 -0500
Subject: [PATCH 1852/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7678)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): normalize MiMo V2.6 models

* fix(vercel): add MiMo V2.6 reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/xiaomi/mimo-v2.6-pro-ultraspeed.toml    | 18 ++++++++++++++++++
 .../models/deepseek/deepseek-v4.1-flash.toml   |  2 +-
 .../vercel/models/xiaomi/mimo-v2.6-flash.toml  | 13 +++++++++++++
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml       | 13 +++++++++++++
 .../vercel/models/xiaomi/mimo-v2.6-pro.toml    | 13 +++++++++++++
 5 files changed, 58 insertions(+), 1 deletion(-)
 create mode 100644 models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
 create mode 100644 providers/vercel/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/vercel/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
 create mode 100644 providers/vercel/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..a4ef4e05975
--- /dev/null
+++ b/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,18 @@
+name = "MiMo-V2.6-Pro-UltraSpeed"
+description = "MiMo pro model for strong multimodal reasoning and agent execution"
+family = "mimo"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "audio"]
+output = ["text"]
diff --git a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
index c185e36af68..5d9ec77c911 100644
--- a/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/vercel/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,7 +12,7 @@ values = ["high", "xhigh"]
 [cost]
 input = 0.3
 output = 1.2
-cache_read = 0.03
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/vercel/models/xiaomi/mimo-v2.6-flash.toml b/providers/vercel/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..bb5496a5085
--- /dev/null
+++ b/providers/vercel/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning.effort = none|minimal|low|medium|high|xhigh|max
+# https://mimo.mi.com/docs/en-US/api/chat/responses
+base_model = "xiaomi/mimo-v2.6-flash"
+name = "MiMo V2.6 Flash"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/vercel/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/vercel/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..dc1eb9fa8fa
--- /dev/null
+++ b/providers/vercel/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning.effort = none|minimal|low|medium|high|xhigh|max
+# https://mimo.mi.com/docs/en-US/api/chat/responses
+base_model = "xiaomi/mimo-v2.6-pro-ultraspeed"
+name = "MiMo V2.6 Pro UltraSpeed"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4.35
+output = 8.7
+cache_read = 0.036
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/vercel/models/xiaomi/mimo-v2.6-pro.toml b/providers/vercel/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..11f32cb1976
--- /dev/null
+++ b/providers/vercel/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,13 @@
+# Effort: reasoning.effort = none|minimal|low|medium|high|xhigh|max
+# https://mimo.mi.com/docs/en-US/api/chat/responses
+base_model = "xiaomi/mimo-v2.6-pro"
+name = "MiMo V2.6 Pro"
+reasoning_options = [{ type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[modalities]
+input = ["text", "image"]

From 15d02bdb540bc047ef8943d41bf59214477a14e8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:30:37 -0500
Subject: [PATCH 1853/2012] fix(sync): preserve Workers AI pricing overrides
 (#7651)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../sync/providers/cloudflare-workers-ai.ts   |  8 +--
 .../core/test/cloudflare-workers-ai.test.ts   | 56 +++++++++++++++++++
 2 files changed, 57 insertions(+), 7 deletions(-)
 create mode 100644 packages/core/test/cloudflare-workers-ai.test.ts

diff --git a/packages/core/src/sync/providers/cloudflare-workers-ai.ts b/packages/core/src/sync/providers/cloudflare-workers-ai.ts
index 252e9ae9cba..e68dd2b8e36 100644
--- a/packages/core/src/sync/providers/cloudflare-workers-ai.ts
+++ b/packages/core/src/sync/providers/cloudflare-workers-ai.ts
@@ -41,13 +41,7 @@ const CloudflareModel = z.object({
   max_output_length: z.number().nullable().optional(),
   input_modalities: z.array(z.string()).optional(),
   output_modalities: z.array(z.string()).optional(),
-  pricing: z.object({
-    prompt: z.string(),
-    completion: z.string(),
-    internal_reasoning: z.string().optional(),
-    input_cache_read: z.string().optional(),
-    input_cache_write: z.string().optional(),
-  }),
+  pricing: OpenRouterModel.shape.pricing,
   supported_features: z.array(z.string()).optional(),
   supported_sampling_parameters: z.array(z.string()).optional(),
 }).passthrough();
diff --git a/packages/core/test/cloudflare-workers-ai.test.ts b/packages/core/test/cloudflare-workers-ai.test.ts
new file mode 100644
index 00000000000..e15416cb7ec
--- /dev/null
+++ b/packages/core/test/cloudflare-workers-ai.test.ts
@@ -0,0 +1,56 @@
+import { expect, test } from "bun:test";
+
+import { cloudflareWorkersAi } from "../src/sync/providers/cloudflare-workers-ai.js";
+
+test("preserves OpenRouter pricing overrides as context tiers", () => {
+  const [source] = cloudflareWorkersAi.parseModels({
+    result: {
+      data: [{
+        id: "workers-ai/@cf/test/context-priced",
+        name: "Context Priced",
+        created: 1_782_777_600,
+        hugging_face_id: null,
+        knowledge_cutoff: null,
+        context_length: 256_000,
+        architecture: {
+          input_modalities: ["text"],
+          output_modalities: ["text"],
+        },
+        pricing: {
+          prompt: "0.000001",
+          completion: "0.000002",
+          overrides: [{
+            min_prompt_tokens: 128_000,
+            prompt: "0.000003",
+            completion: "0.000004",
+          }],
+        },
+        top_provider: {
+          context_length: 256_000,
+          max_completion_tokens: 8_192,
+        },
+        supported_parameters: [],
+      }],
+    },
+  });
+
+  expect(source?.pricing.overrides).toEqual([{
+    min_prompt_tokens: 128_000,
+    prompt: "0.000003",
+    completion: "0.000004",
+  }]);
+
+  const translated = cloudflareWorkersAi.translateModel(source!, {
+    existing: () => undefined,
+    authored: () => undefined,
+  });
+  expect(translated.model.cost).toEqual({
+    input: 1,
+    output: 2,
+    tiers: [{
+      tier: { type: "context", size: 128_000 },
+      input: 3,
+      output: 4,
+    }],
+  });
+});

From 8a0fe46cc5e854bea2668da39a9a5faf267b3482 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:34:42 -0500
Subject: [PATCH 1854/2012] fix(alibaba-cn): correct Qwen context pricing
 (#7692)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Use Alibaba’s published Beijing USD tiers and align qwen3-max with the current 2026-01-23 alias metadata.

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
Co-authored-by: tbourrillon <7868998+tbourrillon@users.noreply.github.com>
---
 providers/alibaba-cn/models/qwen3-max.toml    | 40 +++++++++++++++----
 .../alibaba-cn/models/qwen3.5-flash.toml      | 29 ++++++++++----
 providers/alibaba-cn/models/qwen3.5-plus.toml | 30 ++++++++++----
 3 files changed, 78 insertions(+), 21 deletions(-)

diff --git a/providers/alibaba-cn/models/qwen3-max.toml b/providers/alibaba-cn/models/qwen3-max.toml
index ffb4eb4582b..6f9229234b8 100644
--- a/providers/alibaba-cn/models/qwen3-max.toml
+++ b/providers/alibaba-cn/models/qwen3-max.toml
@@ -1,23 +1,49 @@
+# Sources (accessed 2026-09-22):
+# https://help.aliyun.com/en/model-studio/model-pricing (China (Beijing) USD list prices)
+# https://help.aliyun.com/en/model-studio/model-qwen3-max (alias, capabilities, and limits)
+# https://help.aliyun.com/en/model-studio/deep-thinking
+# qwen3-max currently aliases qwen3-max-2026-01-23.
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens, max 81,920)
 name = "Qwen3 Max"
 description = "Flagship Qwen model for complex reasoning, coding, and agentic workflows"
 family = "qwen"
-release_date = "2025-09-23"
-last_updated = "2026-09-11"
+release_date = "2026-01-23"
+last_updated = "2026-09-22"
 attachment = false
-reasoning = false
+reasoning = true
+structured_output = true
 temperature = true
 knowledge = "2025-04"
 tool_call = true
 open_weights = false
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+max = 81_920
+
+[interleaved]
+field = "reasoning_content"
+
 [cost]
-input = 1.291
-output = 7.749
+input = 0.359
+output = 1.434
+reasoning = 1.434
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.574
+output = 2.294
+reasoning = 2.294
 
 [[cost.tiers]]
 tier = { type = "context", size = 128_000 }
-input = 2.153
-output = 12.915
+input = 1.004
+output = 4.014
+reasoning = 4.014
 
 [limit]
 context = 262_144
diff --git a/providers/alibaba-cn/models/qwen3.5-flash.toml b/providers/alibaba-cn/models/qwen3.5-flash.toml
index 51651ed2e75..4e8481bd07f 100644
--- a/providers/alibaba-cn/models/qwen3.5-flash.toml
+++ b/providers/alibaba-cn/models/qwen3.5-flash.toml
@@ -1,8 +1,14 @@
+# Sources (accessed 2026-09-22):
+# https://help.aliyun.com/en/model-studio/model-pricing (China (Beijing) USD list prices)
+# https://help.aliyun.com/en/model-studio/qwen3-5-flash (capabilities and limits)
+# https://help.aliyun.com/en/model-studio/deep-thinking
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens, max 81,920)
 name = "Qwen3.5 Flash"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
 release_date = "2026-02-23"
-last_updated = "2026-09-11"
+last_updated = "2026-09-22"
 attachment = true
 reasoning = true
 structured_output = true
@@ -18,16 +24,25 @@ type = "toggle"
 type = "budget_tokens"
 max = 81_920
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
-input = 0.172
-output = 1.033
-reasoning = 1.033
+input = 0.029
+output = 0.287
+reasoning = 0.287
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 0.115
+output = 1.147
+reasoning = 1.147
 
 [[cost.tiers]]
 tier = { type = "context", size = 256_000 }
-input = 0.689
-output = 4.133
-reasoning = 4.133
+input = 0.172
+output = 1.72
+reasoning = 1.72
 
 [limit]
 context = 1_000_000
diff --git a/providers/alibaba-cn/models/qwen3.5-plus.toml b/providers/alibaba-cn/models/qwen3.5-plus.toml
index 56105ed505b..429ae670c66 100644
--- a/providers/alibaba-cn/models/qwen3.5-plus.toml
+++ b/providers/alibaba-cn/models/qwen3.5-plus.toml
@@ -1,10 +1,17 @@
+# Sources (accessed 2026-09-22):
+# https://help.aliyun.com/en/model-studio/model-pricing (China (Beijing) USD list prices)
+# https://help.aliyun.com/en/model-studio/qwen3-5-plus (capabilities and limits)
+# https://help.aliyun.com/en/model-studio/deep-thinking
+# Toggle: enable_thinking true|false
+# Budget: thinking_budget (integer reasoning tokens, max 81,920)
 name = "Qwen3.5 Plus"
 description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
 family = "qwen"
 release_date = "2026-02-16"
-last_updated = "2026-09-11"
-attachment = false
+last_updated = "2026-09-22"
+attachment = true
 reasoning = true
+structured_output = true
 temperature = true
 knowledge = "2025-04"
 tool_call = true
@@ -17,16 +24,25 @@ type = "toggle"
 type = "budget_tokens"
 max = 81_920
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
+input = 0.115
+output = 0.688
+reasoning = 0.688
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
 input = 0.287
-output = 1.722
-reasoning = 1.722
+output = 1.72
+reasoning = 1.72
 
 [[cost.tiers]]
 tier = { type = "context", size = 256_000 }
-input = 1.148
-output = 6.888
-reasoning = 6.888
+input = 0.573
+output = 3.44
+reasoning = 3.44
 
 [limit]
 context = 1_000_000

From ec4b93ddb6b1f6b9607c099259db651fd62c4a6f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 02:35:03 +0000
Subject: [PATCH 1855/2012] chore(sync): update OpenRouter model catalog
 (#7693)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml       | 18 +-----------------
 providers/openrouter/models/z-ai/glm-5.3.toml  |  6 +++---
 .../~deepseek/deepseek-v4-flash-latest.toml    |  2 +-
 .../openrouter/models/~z-ai/glm-latest.toml    |  6 +++---
 4 files changed, 8 insertions(+), 24 deletions(-)

diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
index 40cfae78edb..e7a245618f3 100644
--- a/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -1,16 +1,8 @@
 # Toggle: reasoning.enabled = true|false
 # https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "MiMo-V2.6-Pro-UltraSpeed"
+base_model = "xiaomi/mimo-v2.6-pro-ultraspeed"
 description = "MiMo pro model for strong multimodal reasoning and agent execution"
-family = "mimo"
-release_date = "2026-09-21"
-last_updated = "2026-09-21"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "toggle"
@@ -19,11 +11,3 @@ type = "toggle"
 input = 4.35
 output = 8.7
 cache_read = 0.036
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "audio"]
-output = ["text"]
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index b68509e1e67..358dbfe6dac 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.784
-output = 2.464
-cache_read = 0.1456
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 81420955903..6c267e8f600 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -22,7 +22,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.03
 output = 0.8
-cache_read = 0.012
+cache_read = 0.008
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index afa82f5ee86..6bcd917eec4 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7826
-output = 2.4596
-cache_read = 0.14534
+input = 0.7
+output = 2.2
+cache_read = 0.13
 
 [limit]
 context = 1_310_720

From f909dbf12263e28a5e28517967bf28f7d00b3912 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 02:35:08 +0000
Subject: [PATCH 1856/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7694)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/mimo-v2-6-pro-ultraspeed.toml           | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/empiriolabs/models/mimo-v2-6-pro-ultraspeed.toml

diff --git a/providers/empiriolabs/models/mimo-v2-6-pro-ultraspeed.toml b/providers/empiriolabs/models/mimo-v2-6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..e69d20993d0
--- /dev/null
+++ b/providers/empiriolabs/models/mimo-v2-6-pro-ultraspeed.toml
@@ -0,0 +1,14 @@
+base_model = "xiaomi/mimo-v2.6-pro-ultraspeed"
+name = "MiMo V2.6 Pro UltraSpeed"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 21.75
+output = 43.5
+cache_read = 21.75
+
+[limit]
+context = 1_000_000

From 9ef10005b290d2dbbd8edcf5eb0ba198e72cef50 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 02:35:48 +0000
Subject: [PATCH 1857/2012] chore(sync): update Kilo model catalog (#7695)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml       | 18 +-----------------
 1 file changed, 1 insertion(+), 17 deletions(-)

diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
index 08e541fac0d..ad9bc52de83 100644
--- a/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -1,14 +1,6 @@
-name = "Xiaomi: MiMo-V2.6-Pro-UltraSpeed"
+base_model = "xiaomi/mimo-v2.6-pro-ultraspeed"
 description = "MiMo-V2.6-Pro-UltraSpeed is the fast speed edition of Xiaomi's flagship foundation model, MiMo-V2.6-Pro. Built from the same 1T MiMo-V2.6-Pro checkpoint, it matches the original model in quality while delivering roughly 10x..."
-family = "mimo"
-release_date = "2026-09-21"
-last_updated = "2026-09-21"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
 structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -18,11 +10,3 @@ values = ["none", "high"]
 input = 4.35
 output = 8.7
 cache_read = 0.036
-
-[limit]
-context = 1_048_576
-output = 131_072
-
-[modalities]
-input = ["text", "image", "video", "audio"]
-output = ["text"]

From fd3dedfe12242774c50647a1fdfd928269e32292 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Mon, 21 Sep 2026 21:37:18 -0500
Subject: [PATCH 1858/2012] chore: use Muse Spark in workflows (#7696)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 .github/workflows/ci-fixer.yml    | 2 +-
 .github/workflows/issue-fixer.yml | 2 +-
 .github/workflows/opencode.yml    | 3 ++-
 .github/workflows/pr-reviewer.yml | 2 +-
 4 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ci-fixer.yml b/.github/workflows/ci-fixer.yml
index 9d1205a126c..846da3efff8 100644
--- a/.github/workflows/ci-fixer.yml
+++ b/.github/workflows/ci-fixer.yml
@@ -160,7 +160,7 @@ jobs:
           Failed log excerpt:
           EOF
             cat "$LOG_FILE"
-          } | opencode run --agent ci-fixer -m opencode/grok-4.5 | tee "$RESPONSE_FILE"
+          } | opencode run --agent ci-fixer -m opencode/muse-spark-1.3#xhigh | tee "$RESPONSE_FILE"
 
       - name: Check changed paths
         if: steps.budget.outputs.run == 'true' && steps.budget-cache.outputs.cache-hit != 'true'
diff --git a/.github/workflows/issue-fixer.yml b/.github/workflows/issue-fixer.yml
index 17a2ce9a544..e4b26e201d8 100644
--- a/.github/workflows/issue-fixer.yml
+++ b/.github/workflows/issue-fixer.yml
@@ -60,7 +60,7 @@ jobs:
             + "If it is a feature request, a request to track a new kind of information, a question, or any miscellaneous non-catalog-data request, do not edit files. Respond briefly that it needs maintainer review and no automated fix was opened."
           ' "$ISSUE_FILE" > "$PROMPT_FILE"
 
-          opencode run --agent issue-fixer -m opencode/grok-4.5 --format json < "$PROMPT_FILE" | tee "$EVENTS_FILE"
+          opencode run --agent issue-fixer -m opencode/muse-spark-1.3#xhigh --format json < "$PROMPT_FILE" | tee "$EVENTS_FILE"
 
           if ! jq -ers 'map(select(.type == "text") | .part.text) | last | select(length > 0)' "$EVENTS_FILE" > "$RESPONSE_FILE"; then
             echo "Issue fixer did not produce a final response." >&2
diff --git a/.github/workflows/opencode.yml b/.github/workflows/opencode.yml
index 24c819769a9..2e06c6beccb 100644
--- a/.github/workflows/opencode.yml
+++ b/.github/workflows/opencode.yml
@@ -27,4 +27,5 @@ jobs:
         env:
           OPENCODE_API_KEY: ${{ secrets.OPENCODE_API_KEY }}
         with:
-          model: opencode/grok-4.5
+          model: opencode/muse-spark-1.3
+          variant: xhigh
diff --git a/.github/workflows/pr-reviewer.yml b/.github/workflows/pr-reviewer.yml
index 48a5c853311..45336b5fdd6 100644
--- a/.github/workflows/pr-reviewer.yml
+++ b/.github/workflows/pr-reviewer.yml
@@ -78,7 +78,7 @@ jobs:
           export PR_REVIEW_READY_FILE
           rm -f "$PR_REVIEW_READY_FILE"
 
-          opencode run --agent pr-reviewer -m opencode/grok-4.5 --format json <<'EOF' | tee "$EVENTS_FILE"
+          opencode run --agent pr-reviewer -m opencode/muse-spark-1.3#xhigh --format json <<'EOF' | tee "$EVENTS_FILE"
           Review this pull request using the trusted reviewer instructions. Start with `.pr-review/pull-request.json`, `.pr-review/diff.patch`, `AGENTS.md`, and the contributing guidance in `README.md`. Read `sync.md`, the reasoning-options audit guide, schema code, and nearby base-revision files when relevant to the changed files. Use only the read, glob, grep, and mark-pr-ready tools. Return only the final review comment in the agent's required output format. Never include progress narration or passed-check summaries.
           EOF
 

From a38ec312a02a31f9b111bea47856ad8a994d4b25 Mon Sep 17 00:00:00 2001
From: thatdevguy <46411187+chrissalomon@users.noreply.github.com>
Date: Mon, 21 Sep 2026 22:37:44 -0400
Subject: [PATCH 1859/2012] Add Tempr's Google Gemini and Gemma models (#7674)

* Add Tempr's Google Gemini models

16 models Google currently serves that Tempr's Gateway can reach: 14
chat models and the 2 embedding models, base_model-ing the lab entries
with Google's own pricing. gemini-embedding-2 is text-only on Tempr's
/v1/embeddings, so its entry says so.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* tempr: drop the Gemini 2.5 models Google no longer serves to new users

gemini-2.5-flash, gemini-2.5-flash-lite and gemini-2.5-pro are still in
Google's model list, but a call answers 404 "This model ... is no longer
available to new users" (checked 2026-09-21), so a new Tempr customer
can't reach them. Their successors (gemini-3.6-flash,
gemini-3.5-flash-lite, gemini-3.1-pro-preview) are already listed.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* tempr: add the Gemma 4 models (gemma-4-31b-it, gemma-4-26b-a4b-it)

The two Gemma 4 models the Gemini API serves, on the same Google key as
the Gemini entries: base_model points at models/google/gemma-4-*, no
[cost] (same as providers/google), and the thinking toggle Tempr's
GET /v1/models reports. Tempr switches it with thinkingLevel
"minimal"/"high", the only two levels Gemma 4 takes (live 2026-09-21).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
---
 .../models/google/gemini-3-flash-preview.toml | 15 ++++++++++++++
 .../models/google/gemini-3.1-flash-lite.toml  | 15 ++++++++++++++
 .../gemini-3.1-pro-preview-customtools.toml   | 20 +++++++++++++++++++
 .../models/google/gemini-3.1-pro-preview.toml | 20 +++++++++++++++++++
 .../models/google/gemini-3.5-flash-lite.toml  | 14 +++++++++++++
 .../tempr/models/google/gemini-3.5-flash.toml | 15 ++++++++++++++
 .../tempr/models/google/gemini-3.6-flash.toml | 15 ++++++++++++++
 .../tempr/models/google/gemini-3.7-flash.toml | 15 ++++++++++++++
 .../tempr/models/google/gemini-3.8-flash.toml | 15 ++++++++++++++
 .../models/google/gemini-embedding-001.toml   |  5 +++++
 .../models/google/gemini-embedding-2.toml     | 15 ++++++++++++++
 .../models/google/gemini-flash-latest.toml    | 15 ++++++++++++++
 .../google/gemini-flash-lite-latest.toml      | 14 +++++++++++++
 .../models/google/gemma-4-26b-a4b-it.toml     |  9 +++++++++
 .../tempr/models/google/gemma-4-31b-it.toml   |  9 +++++++++
 15 files changed, 211 insertions(+)
 create mode 100644 providers/tempr/models/google/gemini-3-flash-preview.toml
 create mode 100644 providers/tempr/models/google/gemini-3.1-flash-lite.toml
 create mode 100644 providers/tempr/models/google/gemini-3.1-pro-preview-customtools.toml
 create mode 100644 providers/tempr/models/google/gemini-3.1-pro-preview.toml
 create mode 100644 providers/tempr/models/google/gemini-3.5-flash-lite.toml
 create mode 100644 providers/tempr/models/google/gemini-3.5-flash.toml
 create mode 100644 providers/tempr/models/google/gemini-3.6-flash.toml
 create mode 100644 providers/tempr/models/google/gemini-3.7-flash.toml
 create mode 100644 providers/tempr/models/google/gemini-3.8-flash.toml
 create mode 100644 providers/tempr/models/google/gemini-embedding-001.toml
 create mode 100644 providers/tempr/models/google/gemini-embedding-2.toml
 create mode 100644 providers/tempr/models/google/gemini-flash-latest.toml
 create mode 100644 providers/tempr/models/google/gemini-flash-lite-latest.toml
 create mode 100644 providers/tempr/models/google/gemma-4-26b-a4b-it.toml
 create mode 100644 providers/tempr/models/google/gemma-4-31b-it.toml

diff --git a/providers/tempr/models/google/gemini-3-flash-preview.toml b/providers/tempr/models/google/gemini-3-flash-preview.toml
new file mode 100644
index 00000000000..af36c2e66df
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3-flash-preview.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-3-flash-preview"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 0.5
+output = 3
+cache_read = 0.05
+input_audio = 1
diff --git a/providers/tempr/models/google/gemini-3.1-flash-lite.toml b/providers/tempr/models/google/gemini-3.1-flash-lite.toml
new file mode 100644
index 00000000000..56979e347f7
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.1-flash-lite.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-3.1-flash-lite"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+input_audio = 0.5
diff --git a/providers/tempr/models/google/gemini-3.1-pro-preview-customtools.toml b/providers/tempr/models/google/gemini-3.1-pro-preview-customtools.toml
new file mode 100644
index 00000000000..cc288f0d5ca
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.1-pro-preview-customtools.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high
+# /v1/responses: reasoning.effort = low|medium|high
+base_model = "google/gemini-3.1-pro-preview-customtools"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200000 }
+input = 4
+output = 18
+cache_read = 0.4
diff --git a/providers/tempr/models/google/gemini-3.1-pro-preview.toml b/providers/tempr/models/google/gemini-3.1-pro-preview.toml
new file mode 100644
index 00000000000..e08f204c81c
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.1-pro-preview.toml
@@ -0,0 +1,20 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high
+# /v1/responses: reasoning.effort = low|medium|high
+base_model = "google/gemini-3.1-pro-preview"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200000 }
+input = 4
+output = 18
+cache_read = 0.4
diff --git a/providers/tempr/models/google/gemini-3.5-flash-lite.toml b/providers/tempr/models/google/gemini-3.5-flash-lite.toml
new file mode 100644
index 00000000000..6863ddd6378
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.5-flash-lite.toml
@@ -0,0 +1,14 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-3.5-flash-lite"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 0.3
+output = 2.5
+cache_read = 0.03
diff --git a/providers/tempr/models/google/gemini-3.5-flash.toml b/providers/tempr/models/google/gemini-3.5-flash.toml
new file mode 100644
index 00000000000..bb1b51b4d5e
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.5-flash.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-3.5-flash"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 1.5
+output = 9
+cache_read = 0.15
+input_audio = 1.5
diff --git a/providers/tempr/models/google/gemini-3.6-flash.toml b/providers/tempr/models/google/gemini-3.6-flash.toml
new file mode 100644
index 00000000000..aa1d7ce7069
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.6-flash.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-3.6-flash"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75
diff --git a/providers/tempr/models/google/gemini-3.7-flash.toml b/providers/tempr/models/google/gemini-3.7-flash.toml
new file mode 100644
index 00000000000..135e4ffe07a
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.7-flash.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high
+# /v1/responses: reasoning.effort = low|medium|high
+base_model = "google/gemini-3.7-flash"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75
diff --git a/providers/tempr/models/google/gemini-3.8-flash.toml b/providers/tempr/models/google/gemini-3.8-flash.toml
new file mode 100644
index 00000000000..44ee2192f92
--- /dev/null
+++ b/providers/tempr/models/google/gemini-3.8-flash.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high
+# /v1/responses: reasoning.effort = low|medium|high
+base_model = "google/gemini-3.8-flash"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75
diff --git a/providers/tempr/models/google/gemini-embedding-001.toml b/providers/tempr/models/google/gemini-embedding-001.toml
new file mode 100644
index 00000000000..eb9ce08109f
--- /dev/null
+++ b/providers/tempr/models/google/gemini-embedding-001.toml
@@ -0,0 +1,5 @@
+base_model = "google/gemini-embedding-001"
+
+[cost]
+input = 0.15
+output = 0
diff --git a/providers/tempr/models/google/gemini-embedding-2.toml b/providers/tempr/models/google/gemini-embedding-2.toml
new file mode 100644
index 00000000000..a9b158bc61f
--- /dev/null
+++ b/providers/tempr/models/google/gemini-embedding-2.toml
@@ -0,0 +1,15 @@
+# Tempr serves this model on /v1/embeddings, which takes OpenAI's embeddings shape, and
+# Gemini gets text input only there: image, audio, video and PDF input isn't reachable.
+base_model = "google/gemini-embedding-2"
+attachment = false
+
+[cost]
+input = 0.2
+output = 0
+
+[limit]
+output = 1
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/tempr/models/google/gemini-flash-latest.toml b/providers/tempr/models/google/gemini-flash-latest.toml
new file mode 100644
index 00000000000..2aa8e6ce3bd
--- /dev/null
+++ b/providers/tempr/models/google/gemini-flash-latest.toml
@@ -0,0 +1,15 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = low|medium|high
+# /v1/responses: reasoning.effort = low|medium|high
+base_model = "google/gemini-flash-latest"
+
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high"] },
+]
+
+[cost]
+input = 0.75
+output = 3.75
+cache_read = 0.075
+input_audio = 0.75
diff --git a/providers/tempr/models/google/gemini-flash-lite-latest.toml b/providers/tempr/models/google/gemini-flash-lite-latest.toml
new file mode 100644
index 00000000000..0c9e9f3e9c8
--- /dev/null
+++ b/providers/tempr/models/google/gemini-flash-lite-latest.toml
@@ -0,0 +1,14 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Effort: reasoning.effort = minimal|low|medium|high (alias: top-level reasoning_effort)
+# /v1/messages: output_config.effort = minimal|low|medium|high
+# /v1/responses: reasoning.effort = minimal|low|medium|high
+base_model = "google/gemini-flash-lite-latest"
+
+reasoning_options = [
+  { type = "effort", values = ["minimal", "low", "medium", "high"] },
+]
+
+[cost]
+input = 0.3
+output = 2.5
+cache_read = 0.03
diff --git a/providers/tempr/models/google/gemma-4-26b-a4b-it.toml b/providers/tempr/models/google/gemma-4-26b-a4b-it.toml
new file mode 100644
index 00000000000..89c5a30b22b
--- /dev/null
+++ b/providers/tempr/models/google/gemma-4-26b-a4b-it.toml
@@ -0,0 +1,9 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# /v1/messages: thinking.type = enabled|disabled
+# /v1/responses: reasoning.effort = "none" turns it off, any other level turns it on
+base_model = "google/gemma-4-26b-a4b-it"
+
+reasoning_options = [
+  { type = "toggle" },
+]
diff --git a/providers/tempr/models/google/gemma-4-31b-it.toml b/providers/tempr/models/google/gemma-4-31b-it.toml
new file mode 100644
index 00000000000..94549c23965
--- /dev/null
+++ b/providers/tempr/models/google/gemma-4-31b-it.toml
@@ -0,0 +1,9 @@
+# Tempr Gateway: https://temprhq.io/docs/gateway-chat-completions#reasoning
+# Toggle: reasoning.enabled = true|false (reasoning.effort = "none" also turns it off)
+# /v1/messages: thinking.type = enabled|disabled
+# /v1/responses: reasoning.effort = "none" turns it off, any other level turns it on
+base_model = "google/gemma-4-31b-it"
+
+reasoning_options = [
+  { type = "toggle" },
+]

From d5bed5aa97cd45fd2221e1f86342f9f33e3775ca Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 03:31:21 +0000
Subject: [PATCH 1860/2012] chore(sync): update Kilo model catalog (#7702)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~z-ai/glm-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 51780ddf2fb..a1f67c33d90 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 0.7826
+output = 2.4596
+cache_read = 0.14534
 
 [limit]
 context = 1_048_576

From 76aceeb8f429801d5cb278f81ab1e9ec56cc00a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 03:31:27 +0000
Subject: [PATCH 1861/2012] chore(sync): update OpenRouter model catalog
 (#7703)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3.toml     | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 358dbfe6dac..8b9af97ef26 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 0.84
+output = 2.64
+cache_read = 0.156
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 6bcd917eec4..afa82f5ee86 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7
-output = 2.2
-cache_read = 0.13
+input = 0.7826
+output = 2.4596
+cache_read = 0.14534
 
 [limit]
 context = 1_310_720

From 409f66dc1588624d781ef566b44e34a10f83bc97 Mon Sep 17 00:00:00 2001
From: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Date: Mon, 21 Sep 2026 22:38:19 -0500
Subject: [PATCH 1862/2012] fix(workflows): pass OpenCode variant separately
 (#7705)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 .github/workflows/ci-fixer.yml    | 2 +-
 .github/workflows/issue-fixer.yml | 2 +-
 .github/workflows/pr-reviewer.yml | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/ci-fixer.yml b/.github/workflows/ci-fixer.yml
index 846da3efff8..5a94d2a3db2 100644
--- a/.github/workflows/ci-fixer.yml
+++ b/.github/workflows/ci-fixer.yml
@@ -160,7 +160,7 @@ jobs:
           Failed log excerpt:
           EOF
             cat "$LOG_FILE"
-          } | opencode run --agent ci-fixer -m opencode/muse-spark-1.3#xhigh | tee "$RESPONSE_FILE"
+          } | opencode run --agent ci-fixer -m opencode/muse-spark-1.3 --variant xhigh | tee "$RESPONSE_FILE"
 
       - name: Check changed paths
         if: steps.budget.outputs.run == 'true' && steps.budget-cache.outputs.cache-hit != 'true'
diff --git a/.github/workflows/issue-fixer.yml b/.github/workflows/issue-fixer.yml
index e4b26e201d8..4e1f144f164 100644
--- a/.github/workflows/issue-fixer.yml
+++ b/.github/workflows/issue-fixer.yml
@@ -60,7 +60,7 @@ jobs:
             + "If it is a feature request, a request to track a new kind of information, a question, or any miscellaneous non-catalog-data request, do not edit files. Respond briefly that it needs maintainer review and no automated fix was opened."
           ' "$ISSUE_FILE" > "$PROMPT_FILE"
 
-          opencode run --agent issue-fixer -m opencode/muse-spark-1.3#xhigh --format json < "$PROMPT_FILE" | tee "$EVENTS_FILE"
+          opencode run --agent issue-fixer -m opencode/muse-spark-1.3 --variant xhigh --format json < "$PROMPT_FILE" | tee "$EVENTS_FILE"
 
           if ! jq -ers 'map(select(.type == "text") | .part.text) | last | select(length > 0)' "$EVENTS_FILE" > "$RESPONSE_FILE"; then
             echo "Issue fixer did not produce a final response." >&2
diff --git a/.github/workflows/pr-reviewer.yml b/.github/workflows/pr-reviewer.yml
index 45336b5fdd6..fe824e9d48d 100644
--- a/.github/workflows/pr-reviewer.yml
+++ b/.github/workflows/pr-reviewer.yml
@@ -78,7 +78,7 @@ jobs:
           export PR_REVIEW_READY_FILE
           rm -f "$PR_REVIEW_READY_FILE"
 
-          opencode run --agent pr-reviewer -m opencode/muse-spark-1.3#xhigh --format json <<'EOF' | tee "$EVENTS_FILE"
+          opencode run --agent pr-reviewer -m opencode/muse-spark-1.3 --variant xhigh --format json <<'EOF' | tee "$EVENTS_FILE"
           Review this pull request using the trusted reviewer instructions. Start with `.pr-review/pull-request.json`, `.pr-review/diff.patch`, `AGENTS.md`, and the contributing guidance in `README.md`. Read `sync.md`, the reasoning-options audit guide, schema code, and nearby base-revision files when relevant to the changed files. Use only the read, glob, grep, and mark-pr-ready tools. Return only the final review comment in the agent's required output format. Never include progress narration or passed-check summaries.
           EOF
 

From f02d68065194a8fcb2e39b035ef7d4e5460d91bf Mon Sep 17 00:00:00 2001
From: Kastan Day <kastanvday@gmail.com>
Date: Mon, 21 Sep 2026 20:50:47 -0700
Subject: [PATCH 1863/2012] fix(sync): safely refresh Workers AI reasoning from
 search (#7453)

Co-authored-by: opencode-agent <opencode-agent@users.noreply.github.com>
---
 .../sync/providers/cloudflare-workers-ai.ts   | 296 +++++++++++++-----
 1 file changed, 223 insertions(+), 73 deletions(-)

diff --git a/packages/core/src/sync/providers/cloudflare-workers-ai.ts b/packages/core/src/sync/providers/cloudflare-workers-ai.ts
index e68dd2b8e36..0ed0e391513 100644
--- a/packages/core/src/sync/providers/cloudflare-workers-ai.ts
+++ b/packages/core/src/sync/providers/cloudflare-workers-ai.ts
@@ -1,16 +1,25 @@
 import { z } from "zod";
-import { readdirSync } from "node:fs";
+import { readFileSync, readdirSync } from "node:fs";
 import path from "node:path";
 
-import type { ExistingModel, SyncedModel, SyncProvider } from "../index.js";
-import {
-  buildOpenRouterModel,
-  OpenRouterModel,
-  OpenRouterResponse,
-} from "./openrouter.js";
+import type { ExistingModel, SyncedFullModel, SyncedModel, SyncProvider } from "../index.js";
+import { MissingReasoningOptionsError } from "../missing-reasoning-options.js";
+import { buildOpenRouterModel, OpenRouterModel } from "./openrouter.js";
 
 const API_BASE = "https://api.cloudflare.com/client/v4/accounts";
-const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+const TOGGLE_HEADER = "# Toggle: chat_template_kwargs.enable_thinking = true|false\n";
+// Models with a verified enable_thinking control. DeepSeek/Kimi use other wire paths;
+// their authored comments are preserved below, never inferred from generic schemas.
+// These are exact model IDs, not a default for future models or whole publishers.
+const ENABLE_THINKING_MODELS = new Set([
+  "@cf/google/gemma-4-26b-a4b-it",
+  "@cf/nvidia/nemotron-3-120b-a12b",
+  "@cf/qwen/qwen3.8-27b",
+  "@cf/zai-org/glm-4.7-flash",
+  "@cf/zai-org/glm-5.2",
+]);
+const ROOT_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..");
+const MODELS_DIR = path.join(ROOT_DIR, "models");
 const metadataFilesByPublisher = new Map<string, string[]>();
 const METADATA_PUBLISHERS: Record<string, string> = {
   "deepseek-ai": "deepseek",
@@ -24,31 +33,50 @@ const METADATA_PUBLISHERS: Record<string, string> = {
   "zai-org": "zhipuai",
 };
 
-const CloudflareOpenRouterResponse = z.object({
-  result: z.union([OpenRouterResponse, z.array(OpenRouterModel)]).optional(),
-  result_info: z.object({
-    page: z.number().optional(),
-    total_pages: z.number().optional(),
-  }).passthrough().optional(),
-}).passthrough();
+const WorkersAiReasoning = OpenRouterModel.shape.reasoning.unwrap().partial({ mandatory: true });
+const WorkersAiModel = OpenRouterModel.extend({ reasoning: WorkersAiReasoning.optional() });
+type WorkersAiModel = z.infer<typeof WorkersAiModel>;
 
 const CloudflareModel = z.object({
-  id: z.string(),
-  name: z.string(),
-  created: z.number(),
+  id: z.string().trim().min(1),
+  name: z.string().trim().min(1),
+  created: z.number().int().nonnegative(),
   hugging_face_id: z.string().nullable().optional(),
-  context_length: z.number(),
-  max_output_length: z.number().nullable().optional(),
-  input_modalities: z.array(z.string()).optional(),
-  output_modalities: z.array(z.string()).optional(),
-  pricing: OpenRouterModel.shape.pricing,
+  context_length: z.number().int().positive(),
+  max_output_length: z.number().int().positive().nullable().optional(),
+  input_modalities: z.array(z.string().min(1)).min(1).optional(),
+  output_modalities: z.array(z.string().min(1)).min(1).optional(),
+  pricing: OpenRouterModel.shape.pricing.extend({
+    prompt: z.string().refine(validPrice),
+    completion: z.string().refine(validPrice),
+  }),
   supported_features: z.array(z.string()).optional(),
   supported_sampling_parameters: z.array(z.string()).optional(),
+  // Validate reasoning separately so malformed controls do not discard the model.
+  reasoning: z.unknown().optional(),
 }).passthrough();
 
 const CloudflareResponse = z.object({
-  data: z.array(CloudflareModel),
-}).passthrough();
+  data: z.array(z.unknown()).optional(),
+  result: z.union([
+    z.array(z.unknown()),
+    z.object({ data: z.array(z.unknown()) }),
+  ]).optional(),
+  success: z.literal(true).optional(),
+  result_info: z.object({
+    total_pages: z.number().int().positive().optional(),
+  }).optional(),
+}).refine((response) => response.data !== undefined || response.result !== undefined, {
+  message: "Cloudflare Workers AI response did not include model data",
+});
+
+function modelRows(response: z.infer<typeof CloudflareResponse>) {
+  return response.data ?? (Array.isArray(response.result) ? response.result : response.result!.data);
+}
+
+function validPrice(value: string) {
+  return value.trim() !== "" && Number.isFinite(Number(value)) && Number(value) >= 0;
+}
 
 type CloudflareModel = z.infer<typeof CloudflareModel>;
 
@@ -56,6 +84,8 @@ export const cloudflareWorkersAi = {
   id: "cloudflare-workers-ai",
   name: "Cloudflare Workers AI",
   modelsDir: "providers/cloudflare-workers-ai/models",
+  deleteMissing: false,
+  authoritativeHeaders: true,
   async fetchModels() {
     const accountID = process.env.CLOUDFLARE_WORKERS_AI_SYNC_ACCOUNT_ID;
     const token = process.env.CLOUDFLARE_WORKERS_AI_SYNC_API_TOKEN;
@@ -66,50 +96,104 @@ export const cloudflareWorkersAi = {
     }
 
     const first = await fetchPage(accountID, token, 1);
-    const models = parseCloudflareModels(first);
-    const pageInfo = CloudflareOpenRouterResponse.safeParse(first).success
-      ? CloudflareOpenRouterResponse.parse(first).result_info
-      : undefined;
-
-    for (let page = 2; page <= (pageInfo?.total_pages ?? 1); page++) {
-      models.push(...parseCloudflareModels(await fetchPage(accountID, token, page)));
+    if (first === undefined) throw new Error("Cloudflare Workers AI search returned no usable models");
+    const rows = modelRows(first);
+    for (let page = 2; page <= (first.result_info?.total_pages ?? 1); page++) {
+      const response = await fetchPage(accountID, token, page);
+      // Keep successful pages; deleteMissing: false retains models on failed pages.
+      if (response !== undefined) rows.push(...modelRows(response));
     }
 
-    return { data: models };
+    return { data: rows };
   },
   parseModels(raw) {
-    return parseCloudflareModels(raw);
+    const models = parseCloudflareModels(raw);
+    if (models.length === 0) throw new Error("Cloudflare Workers AI search returned no usable models");
+    return models;
   },
   translateModel(model, context) {
-    const normalized = normalizeModel(model);
-    const id = normalized.id.replace(/^workers-ai\//, "");
+    const id = model.id;
+    const existing = context.existing(id);
+    if (existing === undefined && hasReasoning(model) && workersAiReasoningOptions(model) === undefined) {
+      throw new MissingReasoningOptionsError(id, "Workers AI Search does not specify concrete reasoning controls; manual authoring is needed");
+    }
+    const translated = buildWorkersAiModel(model, existing);
+    // Only read paths already present in the sync runner's catalogue map.
+    const header = existing === undefined ? "" : modelHeader(path.resolve(ROOT_DIR, this.modelsDir, `${id}.toml`));
+    if (header === undefined) return undefined;
+    // Remove only the exact generated line; preserve every other leading comment.
+    const preservedHeader = header.replace(TOGGLE_HEADER, "");
+    // Respect curated model-specific wire instructions, including Qwen's split lines.
+    const hasToggleWire = /\b(?:thinking\.type|chat_template_kwargs\.thinking|enable_thinking)\b/.test(preservedHeader);
+    const toggleHeader = ENABLE_THINKING_MODELS.has(id) ? TOGGLE_HEADER : "";
+    if (translated.reasoning_options?.some((option) => option.type === "toggle") && !hasToggleWire && !toggleHeader) {
+      if (existing === undefined) {
+        throw new MissingReasoningOptionsError(id, "Workers AI toggle wire path needs manual verification");
+      }
+      console.warn(`Keeping catalogue reasoning for ${id}: Workers AI toggle wire path is unknown`);
+      // Keep the reasoning boundary small: other usable properties still update.
+      return {
+        id,
+        model: buildWorkersAiModel({ ...model, reasoning: undefined }, existing),
+        header,
+      };
+    }
     return {
       id,
-      model: buildWorkersAiModel(normalized, context.existing(id)),
+      model: translated,
+      header: (translated.reasoning_options?.some((option) => option.type === "toggle") && !hasToggleWire ? toggleHeader : "")
+        + preservedHeader,
     };
   },
-} satisfies SyncProvider<CloudflareModel>;
+} satisfies SyncProvider<WorkersAiModel>;
+
+function hasReasoning(model: WorkersAiModel) {
+  return model.supported_parameters.includes("reasoning") || model.supported_parameters.includes("include_reasoning")
+    || model.reasoning?.mandatory !== undefined || model.reasoning?.supported_efforts !== undefined
+    || model.reasoning?.supports_max_tokens === true;
+}
+
+function modelHeader(file: string) {
+  try {
+    const lines = readFileSync(file, "utf8").split("\n");
+    const firstKey = lines.findIndex((line) => line.trim() !== "" && !line.trim().startsWith("#"));
+    return lines.slice(0, firstKey === -1 ? lines.length : firstKey).join("\n") + "\n";
+  } catch (error) {
+    console.warn(`Skipping Workers AI model with an unreadable catalogue header: ${file}`, error);
+    return undefined;
+  }
+}
 
 export function buildWorkersAiModel(
-  model: z.infer<typeof OpenRouterModel>,
+  model: WorkersAiModel,
   existing: ExistingModel | undefined,
 ): SyncedModel {
+  const reasoningOptions = workersAiReasoningOptions(model);
+  const reasoning = reasoningOptions !== undefined
+    ? true
+    : existing?.reasoning ?? hasReasoning(model);
   const source = {
     ...model,
+    reasoning: undefined,
+    supported_parameters: [
+      ...model.supported_parameters.filter((parameter) => !["reasoning", "include_reasoning"].includes(parameter)),
+      ...(reasoning ? ["reasoning"] : []),
+    ],
     name: existing?.name ?? model.name,
     top_provider: {
       ...model.top_provider,
       max_completion_tokens: existing?.limit?.output ?? model.top_provider.max_completion_tokens,
     },
   };
-  const synced = {
-    ...buildOpenRouterModel(
-      source,
-      existing,
-      existing?.base_model ?? resolveCloudflareBaseModel(model),
-    ),
-    reasoning_options: existing?.reasoning_options,
-  };
+  // The shared builder uses these options when source.reasoning is omitted.
+  const existingWithReasoningOptions = reasoningOptions === undefined
+    ? existing
+    : { ...existing, reasoning_options: reasoningOptions };
+  const synced = buildOpenRouterModel(
+    source,
+    existingWithReasoningOptions,
+    existing?.base_model ?? resolveCloudflareBaseModel(model),
+  );
   if ("base_model" in synced) return synced;
   return {
     ...synced,
@@ -123,7 +207,50 @@ export function buildWorkersAiModel(
   };
 }
 
-export function resolveCloudflareBaseModel(model: z.infer<typeof OpenRouterModel>) {
+function workersAiReasoningOptions({ id, reasoning }: WorkersAiModel): SyncedFullModel["reasoning_options"] {
+  // A null gateway allowlist does not identify concrete model controls.
+  // Preserve the catalogue instead of expanding it to every effort in the schema.
+  if (reasoning === undefined || reasoning.supported_efforts === null) return undefined;
+
+  const options: NonNullable<SyncedFullModel["reasoning_options"]> = [];
+  const efforts = reasoning.supported_efforts;
+  if (efforts?.length === 0) return undefined;
+  if (efforts === undefined && reasoning.supports_max_tokens !== true) {
+    // GLM-4.7-Flash's verified ConfigAPI/Search shape is { mandatory: false,
+    // default_enabled: true }; its only control is enable_thinking. The generic
+    // input schema's low/medium/high enum is not a model capability.
+    // Creator: https://huggingface.co/zai-org/GLM-4.7-Flash/blob/main/chat_template.jinja
+    // Require the explicit off-capability flag; absence of efforts alone says nothing.
+    return id === "@cf/zai-org/glm-4.7-flash" && reasoning.mandatory === false
+      ? [{ type: "toggle" }]
+      : undefined;
+  }
+  // Without either an explicit mandatory flag or a named off setting, a partial
+  // effort list cannot tell us whether replacing the catalogue would lose a toggle.
+  if (reasoning.mandatory === undefined && !efforts?.includes("none")) return undefined;
+
+  if (reasoning.mandatory === false && !efforts?.includes("none")) {
+    options.push({ type: "toggle" });
+  }
+
+  const values = reasoning.mandatory ? efforts?.filter((value) => value !== "none") : efforts;
+  // One mandatory effective effort offers no caller choice. Unlike missing or
+  // empty metadata, a concrete singleton establishes this explicitly.
+  const fixedEffort = reasoning.mandatory === true && values?.length === 1;
+  if (values?.length && !fixedEffort) {
+    options.push({ type: "effort", values: [...values] });
+  }
+
+  if (reasoning.supports_max_tokens === true) {
+    // Explicit reasoning.max_tokens support, never inferred from output limits.
+    options.push({ type: "budget_tokens" });
+  }
+
+  // Empty control metadata never clears authored controls; a known fixed effort can.
+  return options.length > 0 || fixedEffort ? options : undefined;
+}
+
+export function resolveCloudflareBaseModel(model: WorkersAiModel) {
   const [, publisher] = model.id.replace(/^workers-ai\//, "").split("/");
   if (publisher === undefined) return undefined;
 
@@ -157,39 +284,61 @@ async function fetchPage(accountID: string, token: string, page: number) {
   url.searchParams.set("per_page", "1000");
   url.searchParams.set("page", String(page));
 
-  const response = await fetch(url, {
-    headers: { Authorization: `Bearer ${token}` },
-  });
-  if (!response.ok) {
-    throw new Error(
-      `Cloudflare Workers AI models request failed: ${response.status} ${response.statusText}${await responseDetails(response)}`,
-    );
+  for (let attempt = 1; attempt <= 4; attempt++) {
+    try {
+      const response = await fetch(url, {
+        headers: { Authorization: `Bearer ${token}` },
+        signal: AbortSignal.timeout(30_000),
+      });
+      if (!response.ok) {
+        throw new Error(`${response.status} ${response.statusText}${await responseDetails(response)}`);
+      }
+      return CloudflareResponse.parse(await response.json());
+    } catch (error) {
+      console.warn(
+        `Workers AI search page ${page}, attempt ${attempt}/4 failed: ${error instanceof Error ? error.message : String(error)}`,
+      );
+      if (attempt < 4) await Bun.sleep(30_000);
+    }
   }
-  return response.json();
 }
 
-function parseCloudflareModels(raw: unknown): CloudflareModel[] {
-  const cloudflare = CloudflareResponse.safeParse(raw);
-  if (cloudflare.success) return cloudflare.data.data;
-
-  const direct = OpenRouterResponse.safeParse(raw);
-  if (direct.success) return direct.data.data.map((model) => CloudflareModel.parse(model));
-
-  const wrapped = CloudflareOpenRouterResponse.parse(raw);
-  if (wrapped.result === undefined) {
-    throw new Error("Cloudflare Workers AI response did not include model data");
-  }
-  const models = Array.isArray(wrapped.result) ? wrapped.result : wrapped.result.data;
-  return models.map((model) => CloudflareModel.parse(model));
+function parseCloudflareModels(raw: unknown): WorkersAiModel[] {
+  const response = CloudflareResponse.parse(raw);
+  return modelRows(response).flatMap((row) => {
+    const parsed = CloudflareModel.safeParse(row);
+    if (!parsed.success) {
+      console.warn(`Skipping invalid Workers AI model: ${parsed.error.message}`);
+      return [];
+    }
+    try {
+      return [normalizeModel(parsed.data)];
+    } catch (error) {
+      console.warn(`Skipping invalid Workers AI model ${parsed.data.id}: ${error instanceof Error ? error.message : String(error)}`);
+      return [];
+    }
+  });
 }
 
 function normalizeModel(model: CloudflareModel) {
-  if ("architecture" in model && "top_provider" in model && "supported_parameters" in model) {
-    return OpenRouterModel.parse(model);
+  const parsed = WorkersAiReasoning.safeParse(model.reasoning);
+  const reasoning = parsed.success ? parsed.data : undefined;
+  if (!parsed.success && model.reasoning != null) {
+    console.warn(`Ignoring invalid Workers AI reasoning for ${model.id}: ${parsed.error.message}`);
+  }
+  const id = model.id.replace(/^workers-ai\//, "");
+  const normalizedID = id.startsWith("@cf/") ? id : `@cf/${id}`;
+  if ("architecture" in model || "top_provider" in model || "supported_parameters" in model) {
+    const normalized = WorkersAiModel.parse({ ...model, id: normalizedID, reasoning });
+    z.number().int().positive().nullable().parse(normalized.top_provider.max_completion_tokens);
+    if (normalized.architecture.input_modalities.length === 0 || normalized.architecture.output_modalities.length === 0) {
+      throw new Error("Model modalities must not be empty");
+    }
+    return normalized;
   }
 
-  return OpenRouterModel.parse({
-    id: model.id.startsWith("@cf/") ? model.id : `@cf/${model.id.replace(/^@cf\//, "")}`,
+  return WorkersAiModel.parse({
+    id: normalizedID,
     name: model.name,
     created: model.created,
     hugging_face_id: model.hugging_face_id ?? null,
@@ -208,6 +357,7 @@ function normalizeModel(model: CloudflareModel) {
       ...model.supported_sampling_parameters ?? [],
       ...model.supported_features ?? [],
     ],
+    reasoning,
   });
 }
 

From a483df9c693998f16f3042fcbdd66a9e43bbe840 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Mon, 21 Sep 2026 23:09:23 -0500
Subject: [PATCH 1864/2012] chore(sync): update Pioneer model catalog (#7704)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/fastino/gliguard-LLMGuardrails-300M.toml   | 11 +++++++++--
 providers/pioneer/models/fastino/gliner2-base-v1.toml | 11 +++++++++--
 .../pioneer/models/fastino/gliner2-large-v1.toml      | 11 +++++++++--
 .../models/fastino/gliner2-multi-large-v1.toml        | 11 +++++++++--
 .../pioneer/models/fastino/gliner2-multi-v1.toml      | 11 +++++++++--
 .../fastino/gliner2-privacy-filter-PII-multi.toml     | 11 +++++++++--
 6 files changed, 54 insertions(+), 12 deletions(-)

diff --git a/providers/pioneer/models/fastino/gliguard-LLMGuardrails-300M.toml b/providers/pioneer/models/fastino/gliguard-LLMGuardrails-300M.toml
index de754b40e20..f455ac6d116 100644
--- a/providers/pioneer/models/fastino/gliguard-LLMGuardrails-300M.toml
+++ b/providers/pioneer/models/fastino/gliguard-LLMGuardrails-300M.toml
@@ -3,11 +3,18 @@ description = "Tool-capable chat model for instruction following and agentic app
 release_date = "2026-04-30"
 last_updated = "2026-04-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/pioneer/models/fastino/gliner2-base-v1.toml b/providers/pioneer/models/fastino/gliner2-base-v1.toml
index ef23d31a6ad..060cd035d56 100644
--- a/providers/pioneer/models/fastino/gliner2-base-v1.toml
+++ b/providers/pioneer/models/fastino/gliner2-base-v1.toml
@@ -3,11 +3,18 @@ description = "Tool-capable chat model for instruction following and agentic app
 release_date = "2025-06-30"
 last_updated = "2025-06-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/pioneer/models/fastino/gliner2-large-v1.toml b/providers/pioneer/models/fastino/gliner2-large-v1.toml
index 086e826101d..93f13f5d9c0 100644
--- a/providers/pioneer/models/fastino/gliner2-large-v1.toml
+++ b/providers/pioneer/models/fastino/gliner2-large-v1.toml
@@ -3,11 +3,18 @@ description = "Flagship model for demanding analysis, coding, and production age
 release_date = "2025-06-30"
 last_updated = "2025-06-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/pioneer/models/fastino/gliner2-multi-large-v1.toml b/providers/pioneer/models/fastino/gliner2-multi-large-v1.toml
index 74367df5b6c..5e3d3050671 100644
--- a/providers/pioneer/models/fastino/gliner2-multi-large-v1.toml
+++ b/providers/pioneer/models/fastino/gliner2-multi-large-v1.toml
@@ -3,11 +3,18 @@ description = "Flagship model for demanding analysis, coding, and production age
 release_date = "2025-11-30"
 last_updated = "2025-11-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/pioneer/models/fastino/gliner2-multi-v1.toml b/providers/pioneer/models/fastino/gliner2-multi-v1.toml
index 70b2d9553eb..76f6cdfd84a 100644
--- a/providers/pioneer/models/fastino/gliner2-multi-v1.toml
+++ b/providers/pioneer/models/fastino/gliner2-multi-v1.toml
@@ -3,11 +3,18 @@ description = "Tool-capable chat model for instruction following and agentic app
 release_date = "2025-11-30"
 last_updated = "2025-11-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]
diff --git a/providers/pioneer/models/fastino/gliner2-privacy-filter-PII-multi.toml b/providers/pioneer/models/fastino/gliner2-privacy-filter-PII-multi.toml
index 707d407fa98..af0a2ddc10a 100644
--- a/providers/pioneer/models/fastino/gliner2-privacy-filter-PII-multi.toml
+++ b/providers/pioneer/models/fastino/gliner2-privacy-filter-PII-multi.toml
@@ -3,11 +3,18 @@ description = "Tool-capable chat model for instruction following and agentic app
 release_date = "2026-04-30"
 last_updated = "2026-04-30"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
 [cost]
 input = 0.15
 output = 0.15
@@ -16,7 +23,7 @@ cache_write = 0.15
 
 [limit]
 context = 8_192
-output = 4_096
+output = 8_192
 
 [modalities]
 input = ["text"]

From f3fe9e21d073c676dc712961bb3c8470d0c42739 Mon Sep 17 00:00:00 2001
From: suxcode <1729303158@qq.com>
Date: Tue, 22 Sep 2026 12:09:36 +0800
Subject: [PATCH 1865/2012] =?UTF-8?q?feat(xiaomi)=20:=20=E5=A2=9E=E5=8A=A0?=
 =?UTF-8?q?=20MiMo=20V2.6=20=E5=8E=9F=E7=94=9F=E4=B8=8E=E8=AE=A2=E9=98=85?=
 =?UTF-8?q?=E6=A8=A1=E5=9E=8B=20(#7700)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

edit by gpt-5
---
 .../models/mimo-v2.6-flash.toml               | 12 +++++++
 .../models/mimo-v2.6-pro.toml                 | 12 +++++++
 .../models/mimo-v2.6-flash.toml               | 12 +++++++
 .../models/mimo-v2.6-pro.toml                 | 12 +++++++
 .../models/mimo-v2.6-flash.toml               | 12 +++++++
 .../models/mimo-v2.6-pro.toml                 | 12 +++++++
 providers/xiaomi/models/mimo-v2.6-flash.toml  | 32 +++++++++++++++++++
 .../models/mimo-v2.6-pro-ultraspeed.toml      | 31 ++++++++++++++++++
 providers/xiaomi/models/mimo-v2.6-pro.toml    | 32 +++++++++++++++++++
 9 files changed, 167 insertions(+)
 create mode 100644 providers/xiaomi-token-plan-ams/models/mimo-v2.6-flash.toml
 create mode 100644 providers/xiaomi-token-plan-ams/models/mimo-v2.6-pro.toml
 create mode 100644 providers/xiaomi-token-plan-cn/models/mimo-v2.6-flash.toml
 create mode 100644 providers/xiaomi-token-plan-cn/models/mimo-v2.6-pro.toml
 create mode 100644 providers/xiaomi-token-plan-sgp/models/mimo-v2.6-flash.toml
 create mode 100644 providers/xiaomi-token-plan-sgp/models/mimo-v2.6-pro.toml
 create mode 100644 providers/xiaomi/models/mimo-v2.6-flash.toml
 create mode 100644 providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
 create mode 100644 providers/xiaomi/models/mimo-v2.6-pro.toml

diff --git a/providers/xiaomi-token-plan-ams/models/mimo-v2.6-flash.toml b/providers/xiaomi-token-plan-ams/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..4129697bc51
--- /dev/null
+++ b/providers/xiaomi-token-plan-ams/models/mimo-v2.6-flash.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access
+base_model = "xiaomi/mimo-v2.6-flash"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi-token-plan-ams/models/mimo-v2.6-pro.toml b/providers/xiaomi-token-plan-ams/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..62ef14092f6
--- /dev/null
+++ b/providers/xiaomi-token-plan-ams/models/mimo-v2.6-pro.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access
+base_model = "xiaomi/mimo-v2.6-pro"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi-token-plan-cn/models/mimo-v2.6-flash.toml b/providers/xiaomi-token-plan-cn/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..091c178cc51
--- /dev/null
+++ b/providers/xiaomi-token-plan-cn/models/mimo-v2.6-flash.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/zh-CN/api/chat/openai-api
+base_model = "xiaomi/mimo-v2.6-flash"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi-token-plan-cn/models/mimo-v2.6-pro.toml b/providers/xiaomi-token-plan-cn/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..27c7d777207
--- /dev/null
+++ b/providers/xiaomi-token-plan-cn/models/mimo-v2.6-pro.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/zh-CN/api/chat/openai-api
+base_model = "xiaomi/mimo-v2.6-pro"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-flash.toml b/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..4129697bc51
--- /dev/null
+++ b/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-flash.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access
+base_model = "xiaomi/mimo-v2.6-flash"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-pro.toml b/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..62ef14092f6
--- /dev/null
+++ b/providers/xiaomi-token-plan-sgp/models/mimo-v2.6-pro.toml
@@ -0,0 +1,12 @@
+# Toggle: thinking.type = enabled|disabled
+# Source: https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access
+base_model = "xiaomi/mimo-v2.6-pro"
+reasoning_options = [{ type = "toggle" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
diff --git a/providers/xiaomi/models/mimo-v2.6-flash.toml b/providers/xiaomi/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..338e97b8d17
--- /dev/null
+++ b/providers/xiaomi/models/mimo-v2.6-flash.toml
@@ -0,0 +1,32 @@
+# Toggle: thinking.type = enabled|disabled
+# Sources: https://mimo.mi.com/models/zh-CN/mimo-v2.6-flash and https://mimo.mi.com/docs/zh-CN/api/chat/openai-api
+name = "MiMo-V2.6-Flash"
+description = "MiMo Flash model for multimodal coding agents and long-context automation"
+family = "mimo"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2026-09-22"
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml b/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..f1d91260627
--- /dev/null
+++ b/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,31 @@
+# Toggle: thinking.type = enabled|disabled
+# Sources: https://mimo.mi.com/models/zh-CN/mimo-v2.6-pro-ultraspeed and https://mimo.mi.com/docs/zh-CN/api/chat/openai-api
+name = "MiMo-V2.6-Pro-UltraSpeed"
+description = "MiMo pro model for strong multimodal reasoning and agent execution"
+family = "mimo"
+release_date = "2026-09-21"
+last_updated = "2026-09-21"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 4.35
+output = 8.7
+cache_read = 0.036
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "video", "audio"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/xiaomi/models/mimo-v2.6-pro.toml b/providers/xiaomi/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..8debc33d06f
--- /dev/null
+++ b/providers/xiaomi/models/mimo-v2.6-pro.toml
@@ -0,0 +1,32 @@
+# Toggle: thinking.type = enabled|disabled
+# Sources: https://mimo.mi.com/models/zh-CN/mimo-v2.6-pro and https://mimo.mi.com/docs/zh-CN/api/chat/openai-api
+name = "MiMo-V2.6-Pro"
+description = "MiMo Pro model for multimodal coding agents and long-context automation"
+family = "mimo"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2026-09-22"
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"

From ee0566f601c9036ae67782a8f2e16f4b67a12123 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 04:30:22 +0000
Subject: [PATCH 1866/2012] chore(sync): update OpenRouter model catalog
 (#7706)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  6 ++--
 .../models/deepseek/deepseek-v4.1-flash.toml  |  6 ++--
 .../models/nex-agi/nex-n2.5-mini.toml         | 28 +++++++++++++++++++
 .../models/nex-agi/nex-n2.5-pro.toml          | 28 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 .../models/~z-ai/glm-flash-latest.toml        |  4 +--
 .../openrouter/models/~z-ai/glm-latest.toml   |  8 +++---
 7 files changed, 71 insertions(+), 15 deletions(-)
 create mode 100644 providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
 create mode 100644 providers/openrouter/models/nex-agi/nex-n2.5-pro.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml b/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
new file mode 100644
index 00000000000..3869a240dd3
--- /dev/null
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
@@ -0,0 +1,28 @@
+name = "Nex-N2.5-Mini"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = false
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0.025
+output = 0.1
+cache_read = 0.0025
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml b/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml
new file mode 100644
index 00000000000..b3b848a836e
--- /dev/null
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml
@@ -0,0 +1,28 @@
+name = "Nex-N2.5-Pro"
+description = "Multimodal reasoning model for visual analysis, planning, and tool use"
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = false
+structured_output = false
+open_weights = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1db14b9b08b..f88779c2deb 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.638616
+output = 1.915848
+cache_read = 0.021287
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index ac623f94e52..8f90bbaaf93 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -17,11 +17,11 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.075
 output = 0.25
-cache_read = 0.02
+cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 102_400
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index afa82f5ee86..167f0470b75 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7826
-output = 2.4596
-cache_read = 0.14534
+input = 0.6545
+output = 2.057
+cache_read = 0.107525
 
 [limit]
 context = 1_310_720
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 9c941701896ebe52f7a368c54a6df79471d97737 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 04:30:30 +0000
Subject: [PATCH 1867/2012] chore(sync): update NanoGPT model catalog (#7708)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 848a1b9ea6ba0fe4f568b2b38b4e694645a0230e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 04:30:38 +0000
Subject: [PATCH 1868/2012] chore(sync): update Kilo model catalog (#7709)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 providers/kilo/models/~z-ai/glm-flash-latest.toml        | 4 ++--
 providers/kilo/models/~z-ai/glm-latest.toml              | 8 ++++----
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6caf71dbf01..47b086887a0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.638616
+output = 1.915848
+cache_read = 0.021287
 
 [limit]
 context = 1_024_000
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index 5035574d3d5..aee70deebb0 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -17,11 +17,11 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.075
 output = 0.25
-cache_read = 0.02
+cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 102_400
+output = 943_718
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index a1f67c33d90..3a29b91f006 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.7826
-output = 2.4596
-cache_read = 0.14534
+input = 0.6545
+output = 2.057
+cache_read = 0.107525
 
 [limit]
 context = 1_048_576
-output = 131_072
+output = 943_718
 
 [modalities]
 input = ["text"]

From 2cea6220cdadc68b8c680077ecd3fe99ab42a14a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 05:27:31 +0000
Subject: [PATCH 1869/2012] chore(sync): update Eden AI model catalog (#7713)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/amazon/google.gemma-3-12b-it.toml     |  3 +--
 .../models/amazon/google.gemma-3-12b-it@us.toml  |  3 +--
 .../models/amazon/google.gemma-3-27b-it.toml     |  3 +--
 .../models/amazon/google.gemma-3-27b-it@us.toml  |  3 +--
 .../models/amazon/google.gemma-3-4b-it.toml      |  1 -
 .../models/amazon/google.gemma-3-4b-it@us.toml   |  1 -
 .../models/amazon/moonshot.kimi-k2-thinking.toml |  5 ++---
 .../amazon/openai.gpt-oss-safeguard-20b.toml     |  2 --
 .../amazon/openai.gpt-oss-safeguard-20b@us.toml  |  2 --
 .../edenai/models/amazon/zai.glm-4.7-flash.toml  |  5 ++++-
 .../models/amazon/zai.glm-4.7-flash@us.toml      |  5 ++++-
 .../edenai/models/deepinfra/tencent/Hy3.toml     |  6 +++---
 providers/edenai/models/xai/grok-4.7.toml        | 16 ++++++++++++++++
 13 files changed, 33 insertions(+), 22 deletions(-)
 create mode 100644 providers/edenai/models/xai/grok-4.7.toml

diff --git a/providers/edenai/models/amazon/google.gemma-3-12b-it.toml b/providers/edenai/models/amazon/google.gemma-3-12b-it.toml
index 76ea0c2327a..3889a7e378e 100644
--- a/providers/edenai/models/amazon/google.gemma-3-12b-it.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-12b-it.toml
@@ -1,7 +1,6 @@
 base_model = "google/gemma-3-12b-it"
 name = "Gemma 3 12B IT (Amazon Bedrock)"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.09
diff --git a/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
index 91ebef21c30..332de3d4ed7 100644
--- a/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-12b-it@us.toml
@@ -1,7 +1,6 @@
 base_model = "google/gemma-3-12b-it"
 name = "Gemma 3 12B IT (Amazon Bedrock, US)"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.09
diff --git a/providers/edenai/models/amazon/google.gemma-3-27b-it.toml b/providers/edenai/models/amazon/google.gemma-3-27b-it.toml
index b99b2b38ac2..938c2319714 100644
--- a/providers/edenai/models/amazon/google.gemma-3-27b-it.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-27b-it.toml
@@ -1,7 +1,6 @@
 base_model = "google/gemma-3-27b-it"
 name = "Gemma 3 27B IT (Amazon Bedrock)"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.23
diff --git a/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
index 4c9f67f300e..a1637470411 100644
--- a/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-27b-it@us.toml
@@ -1,7 +1,6 @@
 base_model = "google/gemma-3-27b-it"
 name = "Gemma 3 27B IT (Amazon Bedrock, US)"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [cost]
 input = 0.23
diff --git a/providers/edenai/models/amazon/google.gemma-3-4b-it.toml b/providers/edenai/models/amazon/google.gemma-3-4b-it.toml
index f502a9a96cd..536c47c8d73 100644
--- a/providers/edenai/models/amazon/google.gemma-3-4b-it.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-4b-it.toml
@@ -1,6 +1,5 @@
 base_model = "google/gemma-3-4b-it"
 name = "Gemma 3 4B IT (Amazon Bedrock)"
-tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml b/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
index 6d658417b4f..2d4e482e0c4 100644
--- a/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
+++ b/providers/edenai/models/amazon/google.gemma-3-4b-it@us.toml
@@ -1,6 +1,5 @@
 base_model = "google/gemma-3-4b-it"
 name = "Gemma 3 4B IT (Amazon Bedrock, US)"
-tool_call = false
 structured_output = false
 
 [cost]
diff --git a/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml b/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
index 1fdc4e3a58c..7413aad8307 100644
--- a/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
+++ b/providers/edenai/models/amazon/moonshot.kimi-k2-thinking.toml
@@ -1,7 +1,6 @@
 base_model = "moonshotai/kimi-k2-thinking"
 name = "Kimi K2 Thinking (Amazon Bedrock)"
-tool_call = false
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
@@ -9,4 +8,4 @@ input = 0.6
 output = 2.5
 
 [limit]
-context = 128_000
+context = 256_000
diff --git a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
index 536758dfd0e..c159c3bb673 100644
--- a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
+++ b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b.toml
@@ -1,7 +1,5 @@
 base_model = "openai/gpt-oss-safeguard-20b"
 name = "GPT OSS Safeguard 20B (Amazon Bedrock)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
index 7992978de7d..98893d21912 100644
--- a/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
+++ b/providers/edenai/models/amazon/openai.gpt-oss-safeguard-20b@us.toml
@@ -1,7 +1,5 @@
 base_model = "openai/gpt-oss-safeguard-20b"
 name = "GPT OSS Safeguard 20B (Amazon Bedrock, US)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
diff --git a/providers/edenai/models/amazon/zai.glm-4.7-flash.toml b/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
index 2a31f0a6612..bf215607f73 100644
--- a/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
+++ b/providers/edenai/models/amazon/zai.glm-4.7-flash.toml
@@ -1,8 +1,11 @@
 base_model = "zhipuai/glm-4.7-flash"
 name = "GLM-4.7-Flash (Amazon Bedrock)"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
 input = 0.07
 output = 0.4
+
+[limit]
+context = 203_000
diff --git a/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml b/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
index ec6e4df6db1..1aa3c944b18 100644
--- a/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
+++ b/providers/edenai/models/amazon/zai.glm-4.7-flash@us.toml
@@ -1,8 +1,11 @@
 base_model = "zhipuai/glm-4.7-flash"
 name = "GLM-4.7-Flash (Amazon Bedrock, US)"
-structured_output = false
+structured_output = true
 reasoning_options = []
 
 [cost]
 input = 0.07
 output = 0.4
+
+[limit]
+context = 203_000
diff --git a/providers/edenai/models/deepinfra/tencent/Hy3.toml b/providers/edenai/models/deepinfra/tencent/Hy3.toml
index c08482653b4..6cb87c36ee1 100644
--- a/providers/edenai/models/deepinfra/tencent/Hy3.toml
+++ b/providers/edenai/models/deepinfra/tencent/Hy3.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.14
-output = 0.58
-cache_read = 0.035
+input = 0.13
+output = 0.53
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/edenai/models/xai/grok-4.7.toml b/providers/edenai/models/xai/grok-4.7.toml
new file mode 100644
index 00000000000..65cbf72670b
--- /dev/null
+++ b/providers/edenai/models/xai/grok-4.7.toml
@@ -0,0 +1,16 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 3.2
+output = 9.6
+cache_read = 0.8

From f904d8c5907b90a50ad3774e99c64d29dd6c3afc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 05:27:39 +0000
Subject: [PATCH 1870/2012] chore(sync): update Kilo model catalog (#7711)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/nex-agi/nex-n2.5-mini.toml    | 28 +++++++++++++++++++
 .../kilo/models/nex-agi/nex-n2.5-pro.toml     | 28 +++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 providers/kilo/models/nex-agi/nex-n2.5-mini.toml
 create mode 100644 providers/kilo/models/nex-agi/nex-n2.5-pro.toml

diff --git a/providers/kilo/models/nex-agi/nex-n2.5-mini.toml b/providers/kilo/models/nex-agi/nex-n2.5-mini.toml
new file mode 100644
index 00000000000..04ca31fff1a
--- /dev/null
+++ b/providers/kilo/models/nex-agi/nex-n2.5-mini.toml
@@ -0,0 +1,28 @@
+name = "Nex AGI: Nex-N2.5-Mini"
+description = "Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes. Its core strength is agentic coding within a visual feedback loop: it can explore codebases, implement multi-file..."
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0.025
+output = 0.1
+cache_read = 0.0025
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/nex-agi/nex-n2.5-pro.toml b/providers/kilo/models/nex-agi/nex-n2.5-pro.toml
new file mode 100644
index 00000000000..e565d70e36f
--- /dev/null
+++ b/providers/kilo/models/nex-agi/nex-n2.5-pro.toml
@@ -0,0 +1,28 @@
+name = "Nex AGI: Nex-N2.5-Pro"
+description = "Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes. Its core strength is agentic coding within a visual feedback loop: it can explore codebases, implement multi-file..."
+family = "agi"
+release_date = "2026-09-08"
+last_updated = "2026-09-08"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "medium", "high"]
+
+[cost]
+input = 0.075
+output = 0.25
+cache_read = 0.015
+
+[limit]
+context = 262_144
+output = 235_929
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

From 3e8b207f61ac82bff1b9dd5c9575b1f533e8eebb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 05:27:50 +0000
Subject: [PATCH 1871/2012] chore(sync): update OpenRouter model catalog
 (#7712)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/nex-agi/nex-n2.5-mini.toml | 2 +-
 providers/openrouter/models/nex-agi/nex-n2.5-pro.toml  | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml b/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
index 3869a240dd3..539fc6a6686 100644
--- a/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-mini.toml
@@ -7,7 +7,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = false
-structured_output = false
+structured_output = true
 open_weights = true
 
 [[reasoning_options]]
diff --git a/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml b/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml
index b3b848a836e..0cb11065de4 100644
--- a/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml
+++ b/providers/openrouter/models/nex-agi/nex-n2.5-pro.toml
@@ -6,8 +6,8 @@ last_updated = "2026-09-08"
 attachment = true
 reasoning = true
 temperature = true
-tool_call = false
-structured_output = false
+tool_call = true
+structured_output = true
 open_weights = true
 
 [[reasoning_options]]

From 80db4c2dcae857d9c3c4f489c01d86cee7313bcc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 06:41:25 +0000
Subject: [PATCH 1872/2012] chore(sync): update NanoGPT model catalog (#7717)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../deepseek-v4-flash-vision-exp.toml         |  6 +++---
 .../models/xiaomi/mimo-v2.6-flash.toml        | 19 +++++++++++++++++++
 .../xiaomi/mimo-v2.6-pro-ultraspeed.toml      | 16 ++++++++++++++++
 .../nano-gpt/models/xiaomi/mimo-v2.6-pro.toml | 19 +++++++++++++++++++
 4 files changed, 57 insertions(+), 3 deletions(-)
 create mode 100644 providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/nano-gpt/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
 create mode 100644 providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..bf3082ff91f
--- /dev/null
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,19 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+name = "MiMo V2.6 Flash"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+cache_write = 0
+
+[limit]
+input = 1_048_576
+
+[modalities]
+input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
new file mode 100644
index 00000000000..8950d4913de
--- /dev/null
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -0,0 +1,16 @@
+base_model = "xiaomi/mimo-v2.6-pro-ultraspeed"
+name = "MiMo V2.6 Pro UltraSpeed"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 4.35
+output = 8.7
+cache_read = 0.036
+cache_write = 0
+
+[limit]
+input = 1_048_576
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..ea8e65a904c
--- /dev/null
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,19 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+name = "MiMo V2.6 Pro"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+cache_write = 0
+
+[limit]
+input = 1_048_576
+
+[modalities]
+input = ["text", "image", "video", "audio"]

From d179fe47750299d90db707f4d050e6f6581d5ecf Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 06:41:27 +0000
Subject: [PATCH 1873/2012] chore(sync): update OpenRouter model catalog
 (#7716)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml | 1 +
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 2 +-
 4 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..854b27b6a72 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.3
+output = 1.2
+cache_read = 0.006
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 14e222d894d..35f13bee686 100644
--- a/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/openrouter/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -8,3 +8,4 @@ output = 1.2
 
 [limit]
 context = 262_144
+output = 235_929
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 6c267e8f600..8c5f6c380c1 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.8
+output = 1
 cache_read = 0.008
 
 [limit]

From 39d81f59ca3c8bc150b6c54868e6d3a6d346a440 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 06:41:31 +0000
Subject: [PATCH 1874/2012] chore(sync): update Kilo model catalog (#7718)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml   | 4 ++++
 providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml | 2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
index 7c76fb082af..5d67b671b7a 100644
--- a/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
+++ b/providers/kilo/models/qwen/qwen3-next-80b-a3b-thinking.toml
@@ -9,3 +9,7 @@ values = ["high"]
 [cost]
 input = 0.15
 output = 1.2
+
+[limit]
+context = 262_144
+output = 235_929
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 0412502ff5d..6200a01f4d0 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 0.8
+output = 1
 cache_read = 0.008
 
 [limit]

From a7a1850b68d87a73d7893775aa3749462c07acb3 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Tue, 22 Sep 2026 15:21:58 +0800
Subject: [PATCH 1875/2012] chore(opencode): deprecate MiMo V2.5 Free

---
 providers/opencode/models/mimo-v2.5-free.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/opencode/models/mimo-v2.5-free.toml b/providers/opencode/models/mimo-v2.5-free.toml
index 3a1d9a269a6..d799e874914 100644
--- a/providers/opencode/models/mimo-v2.5-free.toml
+++ b/providers/opencode/models/mimo-v2.5-free.toml
@@ -10,6 +10,7 @@ temperature = true
 tool_call = true
 knowledge = "2024-12"
 open_weights = true
+status = "deprecated"
 
 [interleaved]
 field = "reasoning_content"

From c167bacf65d44d05a3c777459a7edc7f710c6c36 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 07:30:22 +0000
Subject: [PATCH 1876/2012] chore(sync): update Kilo model catalog (#7723)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  1 +
 .../models/kwaipilot/kat-coder-pro-v2.toml    | 24 -------------------
 .../models/~deepseek/deepseek-pro-latest.toml | 10 ++++----
 3 files changed, 6 insertions(+), 29 deletions(-)
 delete mode 100644 providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..fa4ef3087d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml b/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml
deleted file mode 100644
index 8e2499749ae..00000000000
--- a/providers/kilo/models/kwaipilot/kat-coder-pro-v2.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "Kwaipilot: KAT-Coder-Pro V2"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-03-27"
-last_updated = "2026-03-27"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.06
-
-[limit]
-context = 262_144
-output = 144_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 47b086887a0..5dc58c6a638 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.638616
-output = 1.915848
-cache_read = 0.021287
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From 62558513ae315f477aea08cfea2b017b3020ebcb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 07:30:37 +0000
Subject: [PATCH 1877/2012] chore(sync): update OpenRouter model catalog
 (#7722)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro-0813.toml |  7 +++---
 .../models/kwaipilot/kat-coder-pro-v2.toml    | 24 -------------------
 .../models/~deepseek/deepseek-pro-latest.toml |  8 +++----
 3 files changed, 8 insertions(+), 31 deletions(-)
 delete mode 100644 providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..f96833d38c7 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml b/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml
deleted file mode 100644
index 80b0f00e52d..00000000000
--- a/providers/openrouter/models/kwaipilot/kat-coder-pro-v2.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-name = "KAT-Coder-Pro V2"
-description = "Coding model for repository understanding, refactors, and agentic engineering tasks"
-family = "kat-coder"
-release_date = "2026-03-27"
-last_updated = "2026-03-27"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.3
-output = 1.2
-cache_read = 0.06
-
-[limit]
-context = 262_144
-output = 144_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index f88779c2deb..1e614023f3e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.638616
-output = 1.915848
-cache_read = 0.021287
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 3406591910d2f2430e766a6fc2f91ff08d71bcc6 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 07:30:44 +0000
Subject: [PATCH 1878/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7725)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/mimo-v2.6-flash.toml | 13 +++++++++++++
 providers/llmgateway/models/mimo-v2.6-pro.toml   | 13 +++++++++++++
 2 files changed, 26 insertions(+)
 create mode 100644 providers/llmgateway/models/mimo-v2.6-flash.toml
 create mode 100644 providers/llmgateway/models/mimo-v2.6-pro.toml

diff --git a/providers/llmgateway/models/mimo-v2.6-flash.toml b/providers/llmgateway/models/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..a71af774946
--- /dev/null
+++ b/providers/llmgateway/models/mimo-v2.6-flash.toml
@@ -0,0 +1,13 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_000_000
diff --git a/providers/llmgateway/models/mimo-v2.6-pro.toml b/providers/llmgateway/models/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..36f93eaabd5
--- /dev/null
+++ b/providers/llmgateway/models/mimo-v2.6-pro.toml
@@ -0,0 +1,13 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[limit]
+context = 1_000_000

From 8e350677b4c9c6f3649c2bdffd51647ea6e5299b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 07:30:48 +0000
Subject: [PATCH 1879/2012] chore(sync): update LLM Gateway model catalog
 (#7724)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/xiaomi/mimo-v2.6-flash.toml            | 15 +++++++++++++++
 .../models/xiaomi/mimo-v2.6-pro.toml              | 15 +++++++++++++++
 2 files changed, 30 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/llmgateway-providers/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-flash.toml b/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..718a69cf405
--- /dev/null
+++ b/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,15 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+name = "MiMo V2.6 Flash (Xiaomi)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_000_000
diff --git a/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-pro.toml b/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..3fe108bfc5a
--- /dev/null
+++ b/providers/llmgateway-providers/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,15 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+name = "MiMo V2.6 Pro (Xiaomi)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
+
+[limit]
+context = 1_000_000

From 4fd8087c870d4132757c7a66b1e24b177d6dda3c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 08:34:32 +0000
Subject: [PATCH 1880/2012] chore(sync): update CrossModel model catalog
 (#7732)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/xiaomi/mimo-v2.6-flash.toml  | 14 ++++++++++++++
 .../crossmodel/models/xiaomi/mimo-v2.6-pro.toml    | 14 ++++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100644 providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
 create mode 100644 providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml

diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml b/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..51fd77c6aa8
--- /dev/null
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
@@ -0,0 +1,14 @@
+base_model = "xiaomi/mimo-v2.6-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.16
+output = 0.32
+cache_read = 0.004
+cache_write = 0.16
+
+[modalities]
+input = ["text", "image", "audio", "video"]
diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml b/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..e0fca43e997
--- /dev/null
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml
@@ -0,0 +1,14 @@
+base_model = "xiaomi/mimo-v2.6-pro"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.47
+output = 0.94
+cache_read = 0.005
+cache_write = 0.47
+
+[modalities]
+input = ["text", "image", "audio", "video"]

From c342e16fa8ea3b663ab79f5c421174379a0e6303 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 08:34:51 +0000
Subject: [PATCH 1881/2012] chore(sync): update NanoGPT model catalog (#7731)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/abliteration-ai/abliterated-model-large-v2.toml    | 4 ++--
 .../models/abliteration-ai/abliterated-model-large.toml       | 4 ++--
 .../nano-gpt/models/abliteration-ai/abliterated-model.toml    | 4 ++--
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
index 2228e8805bd..1a9af392dac 100644
--- a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large-v2.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 5
+input = 3
 output = 5
-cache_read = 0.5
+cache_read = 0.3
 
 [limit]
 context = 1_000_000
diff --git a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
index 71a9a723881..58ef18e1471 100644
--- a/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model-large.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
+input = 3
 output = 5
-cache_read = 0.5
+cache_read = 0.3
 
 [limit]
 context = 1_000_000
diff --git a/providers/nano-gpt/models/abliteration-ai/abliterated-model.toml b/providers/nano-gpt/models/abliteration-ai/abliterated-model.toml
index ac86f6f5738..46dfe7fcc9f 100644
--- a/providers/nano-gpt/models/abliteration-ai/abliterated-model.toml
+++ b/providers/nano-gpt/models/abliteration-ai/abliterated-model.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 3
+input = 1
 output = 3
-cache_read = 0.3
+cache_read = 0.1
 
 [limit]
 context = 262_144

From a4c8f9fd66da7eeccae8cb84dae3802993d99bc7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 09:28:13 +0000
Subject: [PATCH 1882/2012] chore(sync): update Kilo model catalog (#7734)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  3 +--
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index fa4ef3087d9..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,5 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 5dc58c6a638..6caf71dbf01 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From c44ddd7aa48f99d660e4e549c3fbe129065a94c2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 09:28:29 +0000
Subject: [PATCH 1883/2012] chore(sync): update OpenRouter model catalog
 (#7735)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index f96833d38c7..c387c37a3f7 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1e614023f3e..1db14b9b08b 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.558624
+output = 1.675872
+cache_read = 0.018621
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 12a5301ede55a89a2e0d3c003e8bf16f378d1146 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 09:29:44 +0000
Subject: [PATCH 1884/2012] chore(sync): update CrossModel model catalog
 (#7736)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../crossmodel/models/x-ai/grok-4.7.toml      | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/crossmodel/models/x-ai/grok-4.7.toml

diff --git a/providers/crossmodel/models/x-ai/grok-4.7.toml b/providers/crossmodel/models/x-ai/grok-4.7.toml
new file mode 100644
index 00000000000..99495b45cae
--- /dev/null
+++ b/providers/crossmodel/models/x-ai/grok-4.7.toml
@@ -0,0 +1,21 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+cache_write = 2
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+cache_write = 4
+
+[modalities]
+input = ["text", "image"]

From 4bcee582b0287f80995b9eab05a73d4681f209e8 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 10:28:41 +0000
Subject: [PATCH 1885/2012] chore(sync): update Kilo model catalog (#7740)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6caf71dbf01..685f520f471 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.59862
+output = 1.79586
+cache_read = 0.019954
 
 [limit]
 context = 1_024_000

From 7a6af0dd799013b320d0a8fb38aec01d7a564e5d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 10:29:00 +0000
Subject: [PATCH 1886/2012] chore(sync): update OpenRouter model catalog
 (#7739)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index c387c37a3f7..67ed7ee6779 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.59862
+output = 1.79586
+cache_read = 0.019954
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 854b27b6a72..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.3
-output = 1.2
-cache_read = 0.006
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1db14b9b08b..7033c741dee 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.558624
-output = 1.675872
-cache_read = 0.018621
+input = 0.59862
+output = 1.79586
+cache_read = 0.019954
 
 [limit]
 context = 1_048_576

From 614c8224cff67d671f419dd43771601183130573 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 10:29:05 +0000
Subject: [PATCH 1887/2012] chore(sync): update NanoGPT model catalog (#7741)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From 2f7cfc6e35ccc4b9503c68aa0e8ab4d031483a59 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 11:25:30 +0000
Subject: [PATCH 1888/2012] chore(sync): update Kilo model catalog (#7743)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  2 +-
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 01800e18158..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_024_000
+context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 685f520f471..5dc58c6a638 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.59862
-output = 1.79586
-cache_read = 0.019954
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 943_718
 
 [modalities]
 input = ["text"]

From be72b1cf5fc58e8ede2c3bad9675bc0b435e5adb Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 11:25:35 +0000
Subject: [PATCH 1889/2012] chore(sync): update OpenRouter model catalog
 (#7744)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 .../openrouter/models/moonshotai/kimi-k2.7-code.toml      | 2 +-
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 67ed7ee6779..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.59862
-output = 1.79586
-cache_read = 0.019954
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
index 307383582d9..3fb61dea6a2 100644
--- a/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/openrouter/models/moonshotai/kimi-k2.7-code.toml
@@ -4,7 +4,7 @@ reasoning_options = []
 
 [cost]
 input = 0.7062
-output = 3.21
+output = 3.3
 cache_read = 0.18
 
 [limit]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 7033c741dee..1e614023f3e 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.59862
-output = 1.79586
-cache_read = 0.019954
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 943_718
 
 [modalities]
 input = ["text"]

From 83c0d82263bf118e87ad19bbae94503353109434 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 12:36:50 +0000
Subject: [PATCH 1890/2012] chore(sync): update Kilo model catalog (#7746)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..fa4ef3087d9 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,3 +12,4 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
+output = 943_718

From e0cb541fcc24796de9f04f6f8e563c11c21a924e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 12:37:08 +0000
Subject: [PATCH 1891/2012] chore(sync): update OpenRouter model catalog
 (#7745)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..f96833d38c7 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 0.624
+output = 2.88
+cache_read = 0.088
 
 [limit]
 context = 1_048_576
+output = 943_718

From 332109faa1d55b8e16a8da53d74354e5e97b7422 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:28:32 +0000
Subject: [PATCH 1892/2012] chore(sync): update OpenRouter model catalog
 (#7750)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 7 +++----
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 4 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..141d4b41ca6 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.049
+output = 0.098
+cache_read = 0.0098
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index f96833d38c7..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,10 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..2a431cd5648 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.951432
+output = 1.902864
+cache_read = 0.079286
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 1e614023f3e..9d1738cedb4 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.61908
+output = 1.85724
+cache_read = 0.020636
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 384_000
 
 [modalities]
 input = ["text"]

From 7c08237c0e50fa9657fa0ea41c3eae5aeb3e3a1b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:28:51 +0000
Subject: [PATCH 1893/2012] chore(sync): update Kilo model catalog (#7749)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/deepseek/deepseek-v4-pro-0813.toml     |  1 -
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index fa4ef3087d9..007b654034e 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -12,4 +12,3 @@ cache_read = 0.044
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 5dc58c6a638..78a5dc0748a 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.624
-output = 2.88
-cache_read = 0.088
+input = 0.61908
+output = 1.85724
+cache_read = 0.020636
 
 [limit]
-context = 1_048_576
-output = 943_718
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From a8fe72fbf5dbcc280ea7f500bd06fc4cb8e2e594 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:29:35 +0000
Subject: [PATCH 1894/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7753)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3.toml           |  2 +-
 providers/llmgateway/models/inkling-small.toml     | 13 +++++++++++++
 providers/llmgateway/models/inkling.toml           | 13 +++++++++++++
 providers/llmgateway/models/muse-glimmer-30b.toml  | 10 ++++++++++
 .../llmgateway/models/nemotron-3.5-lightning.toml  |  9 +++++++++
 providers/llmgateway/models/qwen3.8-2.4t-a95b.toml | 13 +++++++++++++
 providers/llmgateway/models/step-3.7-flash.toml    | 14 ++++++++++++++
 7 files changed, 73 insertions(+), 1 deletion(-)
 create mode 100644 providers/llmgateway/models/inkling-small.toml
 create mode 100644 providers/llmgateway/models/inkling.toml
 create mode 100644 providers/llmgateway/models/muse-glimmer-30b.toml
 create mode 100644 providers/llmgateway/models/nemotron-3.5-lightning.toml
 create mode 100644 providers/llmgateway/models/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/llmgateway/models/step-3.7-flash.toml

diff --git a/providers/llmgateway/models/glm-5.3.toml b/providers/llmgateway/models/glm-5.3.toml
index 66cc365baaa..6be69ce3acf 100644
--- a/providers/llmgateway/models/glm-5.3.toml
+++ b/providers/llmgateway/models/glm-5.3.toml
@@ -2,7 +2,7 @@ base_model = "zhipuai/glm-5.3"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "high", "max"]
+values = ["none", "low", "high", "max"]
 
 [cost]
 input = 1.2
diff --git a/providers/llmgateway/models/inkling-small.toml b/providers/llmgateway/models/inkling-small.toml
new file mode 100644
index 00000000000..954d4c25f03
--- /dev/null
+++ b/providers/llmgateway/models/inkling-small.toml
@@ -0,0 +1,13 @@
+base_model = "thinkingmachines/inkling-small"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.45
+output = 1.2
+cache_read = 0.1
+
+[limit]
+context = 524_288
diff --git a/providers/llmgateway/models/inkling.toml b/providers/llmgateway/models/inkling.toml
new file mode 100644
index 00000000000..1d07005aa33
--- /dev/null
+++ b/providers/llmgateway/models/inkling.toml
@@ -0,0 +1,13 @@
+base_model = "thinkingmachines/inkling"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.95
+output = 4.05
+cache_read = 0.16
+
+[limit]
+context = 524_288
diff --git a/providers/llmgateway/models/muse-glimmer-30b.toml b/providers/llmgateway/models/muse-glimmer-30b.toml
new file mode 100644
index 00000000000..9d06a10be3b
--- /dev/null
+++ b/providers/llmgateway/models/muse-glimmer-30b.toml
@@ -0,0 +1,10 @@
+base_model = "meta/muse-glimmer-30b"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.04
diff --git a/providers/llmgateway/models/nemotron-3.5-lightning.toml b/providers/llmgateway/models/nemotron-3.5-lightning.toml
new file mode 100644
index 00000000000..6e7dba809e3
--- /dev/null
+++ b/providers/llmgateway/models/nemotron-3.5-lightning.toml
@@ -0,0 +1,9 @@
+base_model = "nvidia/nemotron-3.5-lightning"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.08
+output = 0.2
diff --git a/providers/llmgateway/models/qwen3.8-2.4t-a95b.toml b/providers/llmgateway/models/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..02fe65d078f
--- /dev/null
+++ b/providers/llmgateway/models/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,13 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_010_000
diff --git a/providers/llmgateway/models/step-3.7-flash.toml b/providers/llmgateway/models/step-3.7-flash.toml
new file mode 100644
index 00000000000..ad0acb372ea
--- /dev/null
+++ b/providers/llmgateway/models/step-3.7-flash.toml
@@ -0,0 +1,14 @@
+base_model = "stepfun/step-3.7-flash"
+base_model_omit = ["limit.input"]
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.15
+cache_read = 0.04
+
+[limit]
+context = 262_144

From 6f41b78e33c939a7b9d836fb880a854058ca856b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:30:30 +0000
Subject: [PATCH 1895/2012] chore(sync): update OpenRouter model catalog
 (#7752)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 2a431cd5648..799137f52c2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.951432
-output = 1.902864
-cache_read = 0.079286
+input = 0.946386
+output = 1.892772
+cache_read = 0.078866
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 8b9af97ef26..7c1154ec1bd 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.84
-output = 2.64
-cache_read = 0.156
+input = 0.6538
+output = 2.0548
+cache_read = 0.12142
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 9d1738cedb4..d4c1d5ba628 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.61908
-output = 1.85724
-cache_read = 0.020636
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 167f0470b75..e0e1e1ca329 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.6545
-output = 2.057
-cache_read = 0.107525
+input = 0.6538
+output = 2.0548
+cache_read = 0.12142
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 6651aab362256017c2b80910c57b96a9b567b215 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:30:53 +0000
Subject: [PATCH 1896/2012] chore(sync): update Kilo model catalog (#7754)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 providers/kilo/models/~z-ai/glm-latest.toml            |  8 ++++----
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 78a5dc0748a..27422e128e0 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.61908
-output = 1.85724
-cache_read = 0.020636
+input = 0.57948
+output = 1.73844
+cache_read = 0.018438
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 3a29b91f006..0facfa991b5 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.6545
-output = 2.057
-cache_read = 0.107525
+input = 0.6538
+output = 2.0548
+cache_read = 0.12142
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text"]

From 0609b4f5bddabaad04bf0c56e723dbda8121dc1d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 15:27:00 +0000
Subject: [PATCH 1897/2012] chore(sync): update OpenRouter model catalog
 (#7759)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/aion-labs/aion-2.0.toml       | 2 +-
 providers/openrouter/models/aion-labs/aion-3.0-mini.toml  | 2 +-
 providers/openrouter/models/aion-labs/aion-3.0.toml       | 2 +-
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/openai/gpt-oss-20b.toml       | 8 ++------
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 6 files changed, 12 insertions(+), 16 deletions(-)

diff --git a/providers/openrouter/models/aion-labs/aion-2.0.toml b/providers/openrouter/models/aion-labs/aion-2.0.toml
index b54b64899f3..8e499e073cf 100644
--- a/providers/openrouter/models/aion-labs/aion-2.0.toml
+++ b/providers/openrouter/models/aion-labs/aion-2.0.toml
@@ -16,7 +16,7 @@ output = 1.6
 cache_read = 0.2
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/aion-labs/aion-3.0-mini.toml b/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
index af968c32457..33c5d843701 100644
--- a/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
+++ b/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
@@ -16,7 +16,7 @@ output = 1.4
 cache_read = 0.18
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/aion-labs/aion-3.0.toml b/providers/openrouter/models/aion-labs/aion-3.0.toml
index e25d916b2b3..f16de215471 100644
--- a/providers/openrouter/models/aion-labs/aion-3.0.toml
+++ b/providers/openrouter/models/aion-labs/aion-3.0.toml
@@ -16,7 +16,7 @@ output = 6
 cache_read = 0.75
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 799137f52c2..a705f18ca62 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.946386
-output = 1.892772
-cache_read = 0.078866
+input = 0.936294
+output = 1.872588
+cache_read = 0.078025
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-oss-20b.toml b/providers/openrouter/models/openai/gpt-oss-20b.toml
index b65dd3718e0..98aeb108997 100644
--- a/providers/openrouter/models/openai/gpt-oss-20b.toml
+++ b/providers/openrouter/models/openai/gpt-oss-20b.toml
@@ -6,9 +6,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.03
-output = 0.13
-cache_read = 0.03
-
-[limit]
-output = 117_964
+input = 0.018
+output = 0.09
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index d4c1d5ba628..0ea95b2339f 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.57156
+output = 1.71468
+cache_read = 0.019052
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]

From 8f37270269201a2e8d280327eef9f7e47bc8c37f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 15:27:04 +0000
Subject: [PATCH 1898/2012] chore(sync): update Eden AI model catalog (#7757)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml   | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml       | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml        | 4 ++--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml  | 4 ++--
 providers/edenai/models/scaleway/gpt-oss-120b.toml            | 4 ++--
 providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml  | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index beffb59e2ed..73ebbfe4b80 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.3447
-output = 0.4596
+input = 0.34389
+output = 0.45852
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 11dee7e0f74..92757e9fa56 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.74685
-output = 0.74685
+input = 0.745095
+output = 0.745095
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index 2bf7b7f0f21..cad05c7d393 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17235
-output = 0.74685
+input = 0.171945
+output = 0.745095
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index d761571f13a..d2f9f4d763d 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4596
-output = 0.9192
+input = 0.45852
+output = 0.91704
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index dbff07167cc..8f735ccfed6 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.17235
-output = 0.6894
+input = 0.171945
+output = 0.68778
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 8d733529a6c..55662b4313f 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.0341
-output = 1.0341
+input = 1.03167
+output = 1.03167

From d07f6dce7adfd0e4aa07ceffa9cadd7e059d2455 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 15:27:20 +0000
Subject: [PATCH 1899/2012] chore(sync): update Kilo model catalog (#7758)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/aion-labs/aion-2.0.toml          |  2 +-
 providers/kilo/models/aion-labs/aion-3.0-mini.toml     |  2 +-
 providers/kilo/models/aion-labs/aion-3.0.toml          |  2 +-
 providers/kilo/models/openai/gpt-oss-20b.toml          |  3 ---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml     | 10 +++++-----
 5 files changed, 8 insertions(+), 11 deletions(-)

diff --git a/providers/kilo/models/aion-labs/aion-2.0.toml b/providers/kilo/models/aion-labs/aion-2.0.toml
index 4bbbba9076b..46cc9f272f7 100644
--- a/providers/kilo/models/aion-labs/aion-2.0.toml
+++ b/providers/kilo/models/aion-labs/aion-2.0.toml
@@ -19,7 +19,7 @@ output = 1.6
 cache_read = 0.2
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/aion-labs/aion-3.0-mini.toml b/providers/kilo/models/aion-labs/aion-3.0-mini.toml
index 4987c065dc7..84f2979d62c 100644
--- a/providers/kilo/models/aion-labs/aion-3.0-mini.toml
+++ b/providers/kilo/models/aion-labs/aion-3.0-mini.toml
@@ -19,7 +19,7 @@ output = 1.4
 cache_read = 0.18
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/aion-labs/aion-3.0.toml b/providers/kilo/models/aion-labs/aion-3.0.toml
index caa24ee8921..ac0a56ff9f9 100644
--- a/providers/kilo/models/aion-labs/aion-3.0.toml
+++ b/providers/kilo/models/aion-labs/aion-3.0.toml
@@ -19,7 +19,7 @@ output = 6
 cache_read = 0.75
 
 [limit]
-context = 131_072
+context = 1_048_576
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/openai/gpt-oss-20b.toml b/providers/kilo/models/openai/gpt-oss-20b.toml
index fe3d2508db6..98aeb108997 100644
--- a/providers/kilo/models/openai/gpt-oss-20b.toml
+++ b/providers/kilo/models/openai/gpt-oss-20b.toml
@@ -8,6 +8,3 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.018
 output = 0.09
-
-[limit]
-output = 117_964
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 27422e128e0..9da59d77044 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57948
-output = 1.73844
-cache_read = 0.018438
+input = 0.57156
+output = 1.71468
+cache_read = 0.019052
 
 [limit]
-context = 1_048_576
-output = 393_216
+context = 1_024_000
+output = 384_000
 
 [modalities]
 input = ["text"]

From 173ae60de164b79b830a9cabc8a920c32cf7e0f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 16:29:18 +0000
Subject: [PATCH 1900/2012] chore(sync): update Kilo model catalog (#7760)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.6-27b.toml              | 3 +++
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 3 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.6-27b.toml b/providers/kilo/models/qwen/qwen3.6-27b.toml
index 61085517bb0..6ca4eb2df17 100644
--- a/providers/kilo/models/qwen/qwen3.6-27b.toml
+++ b/providers/kilo/models/qwen/qwen3.6-27b.toml
@@ -9,5 +9,8 @@ values = ["none", "high"]
 input = 0.45
 output = 2.7
 
+[limit]
+output = 262_140
+
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index c3469c1e9dc..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.13
-output = 0.53
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 9da59d77044..6565f5a8e9b 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.57156
-output = 1.71468
-cache_read = 0.019052
+input = 0.56364
+output = 1.69092
+cache_read = 0.018788
 
 [limit]
 context = 1_024_000

From 9255381af9015a988ca3155e23b6d0742a445f19 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 16:29:31 +0000
Subject: [PATCH 1901/2012] chore(sync): update OpenRouter model catalog
 (#7761)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro.toml      | 6 +++---
 providers/openrouter/models/qwen/qwen3.6-27b.toml        | 9 ++++++---
 providers/openrouter/models/tencent/hy3.toml             | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml | 6 +++---
 4 files changed, 15 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index a705f18ca62..8ce9eb4af6a 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.936294
-output = 1.872588
-cache_read = 0.078025
+input = 0.927768
+output = 1.855536
+cache_read = 0.077314
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.6-27b.toml b/providers/openrouter/models/qwen/qwen3.6-27b.toml
index 15ca4007021..c188473afdb 100644
--- a/providers/openrouter/models/qwen/qwen3.6-27b.toml
+++ b/providers/openrouter/models/qwen/qwen3.6-27b.toml
@@ -6,9 +6,12 @@ base_model = "alibaba/qwen3.6-27b"
 type = "toggle"
 
 [cost]
-input = 0.3
-output = 2
-cache_read = 0.03
+input = 0.32
+output = 2.7
+cache_read = 0.15
+
+[limit]
+output = 262_140
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 0ea95b2339f..ff538327fb0 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.57156
-output = 1.71468
-cache_read = 0.019052
+input = 0.56364
+output = 1.69092
+cache_read = 0.018788
 
 [limit]
 context = 1_048_576

From 1ba7a9dff9a2be54e824d7211475b8d9d3b4c5de Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 11:47:14 -0500
Subject: [PATCH 1902/2012] feat: add Claude Opus 5.5 (#7763)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/anthropic/claude-opus-5-5.toml         | 19 +++++++++++++++++++
 .../models/anthropic.claude-opus-5-5.toml     | 11 +++++++++++
 .../models/au.anthropic.claude-opus-5-5.toml  | 13 +++++++++++++
 .../models/eu.anthropic.claude-opus-5-5.toml  | 13 +++++++++++++
 .../global.anthropic.claude-opus-5-5.toml     | 12 ++++++++++++
 .../models/jp.anthropic.claude-opus-5-5.toml  | 13 +++++++++++++
 .../models/us.anthropic.claude-opus-5-5.toml  | 13 +++++++++++++
 .../anthropic/models/claude-opus-5-5.toml     | 13 +++++++++++++
 .../models/claude-opus-5-5.toml               | 13 +++++++++++++
 providers/azure/models/claude-opus-5-5.toml   | 13 +++++++++++++
 .../models/claude-opus-5-5@default.toml       |  9 +++++++++
 .../models/claude-opus-5-5@default.toml       | 12 ++++++++++++
 12 files changed, 154 insertions(+)
 create mode 100644 models/anthropic/claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/anthropic.claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/au.anthropic.claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/eu.anthropic.claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/global.anthropic.claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/jp.anthropic.claude-opus-5-5.toml
 create mode 100644 providers/amazon-bedrock/models/us.anthropic.claude-opus-5-5.toml
 create mode 100644 providers/anthropic/models/claude-opus-5-5.toml
 create mode 100644 providers/azure-cognitive-services/models/claude-opus-5-5.toml
 create mode 100644 providers/azure/models/claude-opus-5-5.toml
 create mode 100644 providers/google-vertex-anthropic/models/claude-opus-5-5@default.toml
 create mode 100644 providers/google-vertex/models/claude-opus-5-5@default.toml

diff --git a/models/anthropic/claude-opus-5-5.toml b/models/anthropic/claude-opus-5-5.toml
new file mode 100644
index 00000000000..0f1cfeba4ed
--- /dev/null
+++ b/models/anthropic/claude-opus-5-5.toml
@@ -0,0 +1,19 @@
+name = "Claude Opus 5.5"
+description = "Claude model for long-running agentic coding and knowledge work"
+family = "claude-opus"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+open_weights = false
+knowledge = "2026-06"
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/amazon-bedrock/models/anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..a9684824bf1
--- /dev/null
+++ b/providers/amazon-bedrock/models/anthropic.claude-opus-5-5.toml
@@ -0,0 +1,11 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/amazon-bedrock/models/au.anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/au.anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..25e28fdb4b6
--- /dev/null
+++ b/providers/amazon-bedrock/models/au.anthropic.claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Bedrock regional endpoints carry a 10% pricing premium over global endpoints.
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 5.5 (AU)"
+
+[cost]
+input = 4.4
+output = 22
+cache_read = 0.22
+cache_write = 5.5
diff --git a/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..8c9d8b97cdd
--- /dev/null
+++ b/providers/amazon-bedrock/models/eu.anthropic.claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Bedrock regional endpoints carry a 10% pricing premium over global endpoints.
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 5.5 (EU)"
+
+[cost]
+input = 4.4
+output = 22
+cache_read = 0.22
+cache_write = 5.5
diff --git a/providers/amazon-bedrock/models/global.anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/global.anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..c36a3b86f74
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.anthropic.claude-opus-5-5.toml
@@ -0,0 +1,12 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 5.5 (Global)"
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..78ad0c8319e
--- /dev/null
+++ b/providers/amazon-bedrock/models/jp.anthropic.claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Bedrock regional endpoints carry a 10% pricing premium over global endpoints.
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 5.5 (JP)"
+
+[cost]
+input = 4.4
+output = 22
+cache_read = 0.22
+cache_write = 5.5
diff --git a/providers/amazon-bedrock/models/us.anthropic.claude-opus-5-5.toml b/providers/amazon-bedrock/models/us.anthropic.claude-opus-5-5.toml
new file mode 100644
index 00000000000..f74ececd884
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.anthropic.claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+# Sources: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
+# https://platform.claude.com/docs/en/about-claude/pricing
+# Bedrock regional endpoints carry a 10% pricing premium over global endpoints.
+# Effort: additionalModelRequestFields.output_config.effort (Converse); output_config.effort (InvokeModel).
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+name = "Claude Opus 5.5 (US)"
+
+[cost]
+input = 4.4
+output = 22
+cache_read = 0.22
+cache_write = 5.5
diff --git a/providers/anthropic/models/claude-opus-5-5.toml b/providers/anthropic/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..7341878caeb
--- /dev/null
+++ b/providers/anthropic/models/claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[experimental.modes.fast]
+cost = { input = 8, output = 40, cache_read = 0.4, cache_write = 10 }
+provider = { body = { speed = "fast" }, headers = { anthropic-beta = "fast-mode-2026-02-01" } }
diff --git a/providers/azure-cognitive-services/models/claude-opus-5-5.toml b/providers/azure-cognitive-services/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..c4fedc9752f
--- /dev/null
+++ b/providers/azure-cognitive-services/models/claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"
diff --git a/providers/azure/models/claude-opus-5-5.toml b/providers/azure/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..cc062a28b81
--- /dev/null
+++ b/providers/azure/models/claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"
diff --git a/providers/google-vertex-anthropic/models/claude-opus-5-5@default.toml b/providers/google-vertex-anthropic/models/claude-opus-5-5@default.toml
new file mode 100644
index 00000000000..cede54eca95
--- /dev/null
+++ b/providers/google-vertex-anthropic/models/claude-opus-5-5@default.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/google-vertex/models/claude-opus-5-5@default.toml b/providers/google-vertex/models/claude-opus-5-5@default.toml
new file mode 100644
index 00000000000..8a2a18a47c1
--- /dev/null
+++ b/providers/google-vertex/models/claude-opus-5-5@default.toml
@@ -0,0 +1,12 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/google-vertex/anthropic"

From 3ede55347c32de6a4cee88e7c68bd31ef76370a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 11:47:41 -0500
Subject: [PATCH 1903/2012] chore(sync): update LLM Gateway model catalog
 (#7751)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepinfra/glm-5.3.toml               | 14 ++++++++++++++
 .../models/deepinfra/inkling-small.toml         | 16 ++++++++++++++++
 .../models/deepinfra/inkling.toml               | 17 +++++++++++++++++
 .../models/deepinfra/muse-glimmer-30b.toml      | 14 ++++++++++++++
 .../deepinfra/nemotron-3.5-lightning.toml       | 13 +++++++++++++
 .../models/deepinfra/qwen3.8-2.4t-a95b.toml     | 11 +++++++++++
 .../models/deepinfra/qwen3.8-27b.toml           | 14 ++++++++++++++
 .../models/deepinfra/step-3.7-flash.toml        | 17 +++++++++++++++++
 .../models/novita/qwen3.8-2.4t-a95b.toml        | 14 ++++++++++++++
 .../models/novita/step-3.7-flash.toml           | 16 ++++++++++++++++
 .../models/together-ai/glm-5.3-flash.toml       | 15 +++++++++++++++
 .../models/together-ai/glm-5.3.toml             | 15 +++++++++++++++
 .../models/together-ai/inkling.toml             | 16 ++++++++++++++++
 .../models/together-ai/muse-glimmer-30b.toml    | 14 ++++++++++++++
 .../models/together-ai/qwen3.8-2.4t-a95b.toml   | 15 +++++++++++++++
 15 files changed, 221 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/deepinfra/glm-5.3.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/inkling-small.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/inkling.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/muse-glimmer-30b.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/nemotron-3.5-lightning.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/qwen3.8-27b.toml
 create mode 100644 providers/llmgateway-providers/models/deepinfra/step-3.7-flash.toml
 create mode 100644 providers/llmgateway-providers/models/novita/qwen3.8-2.4t-a95b.toml
 create mode 100644 providers/llmgateway-providers/models/novita/step-3.7-flash.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/glm-5.3-flash.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/glm-5.3.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/inkling.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/muse-glimmer-30b.toml
 create mode 100644 providers/llmgateway-providers/models/together-ai/qwen3.8-2.4t-a95b.toml

diff --git a/providers/llmgateway-providers/models/deepinfra/glm-5.3.toml b/providers/llmgateway-providers/models/deepinfra/glm-5.3.toml
new file mode 100644
index 00000000000..49f3f113f39
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/glm-5.3.toml
@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 1.2
+output = 4
+cache_read = 0.2
+
+[limit]
+context = 1_048_576
diff --git a/providers/llmgateway-providers/models/deepinfra/inkling-small.toml b/providers/llmgateway-providers/models/deepinfra/inkling-small.toml
new file mode 100644
index 00000000000..0109639620f
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/inkling-small.toml
@@ -0,0 +1,16 @@
+base_model = "thinkingmachines/inkling-small"
+name = "Inkling Small (DeepInfra)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.45
+output = 1.2
+cache_read = 0.1
+
+[limit]
+context = 524_288
+output = 262_144
diff --git a/providers/llmgateway-providers/models/deepinfra/inkling.toml b/providers/llmgateway-providers/models/deepinfra/inkling.toml
new file mode 100644
index 00000000000..41e528e739e
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/inkling.toml
@@ -0,0 +1,17 @@
+base_model = "thinkingmachines/inkling"
+name = "Inkling (DeepInfra)"
+tool_call = false
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.95
+output = 4.05
+cache_read = 0.16
+
+[limit]
+context = 524_288
+output = 262_144
diff --git a/providers/llmgateway-providers/models/deepinfra/muse-glimmer-30b.toml b/providers/llmgateway-providers/models/deepinfra/muse-glimmer-30b.toml
new file mode 100644
index 00000000000..0c26147df21
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/muse-glimmer-30b.toml
@@ -0,0 +1,14 @@
+base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.04
+
+[limit]
+output = 16_384
diff --git a/providers/llmgateway-providers/models/deepinfra/nemotron-3.5-lightning.toml b/providers/llmgateway-providers/models/deepinfra/nemotron-3.5-lightning.toml
new file mode 100644
index 00000000000..aa669cc8be8
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/nemotron-3.5-lightning.toml
@@ -0,0 +1,13 @@
+base_model = "nvidia/nemotron-3.5-lightning"
+name = "Nemotron 3.5 Lightning (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.08
+output = 0.2
+
+[limit]
+output = 131_072
diff --git a/providers/llmgateway-providers/models/deepinfra/qwen3.8-2.4t-a95b.toml b/providers/llmgateway-providers/models/deepinfra/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..09c8ba4899d
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,11 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8 2.4T A95B (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.2
diff --git a/providers/llmgateway-providers/models/deepinfra/qwen3.8-27b.toml b/providers/llmgateway-providers/models/deepinfra/qwen3.8-27b.toml
new file mode 100644
index 00000000000..3b37e4d93ab
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/qwen3.8-27b.toml
@@ -0,0 +1,14 @@
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8 27B (DeepInfra)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.2
+output = 2.5
+cache_read = 0.05
+
+[limit]
+output = 235_929
diff --git a/providers/llmgateway-providers/models/deepinfra/step-3.7-flash.toml b/providers/llmgateway-providers/models/deepinfra/step-3.7-flash.toml
new file mode 100644
index 00000000000..d9f47a5da91
--- /dev/null
+++ b/providers/llmgateway-providers/models/deepinfra/step-3.7-flash.toml
@@ -0,0 +1,17 @@
+base_model = "stepfun/step-3.7-flash"
+base_model_omit = ["limit.input"]
+name = "Step 3.7 Flash (DeepInfra)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.15
+cache_read = 0.04
+
+[limit]
+context = 262_144
+output = 32_768
diff --git a/providers/llmgateway-providers/models/novita/qwen3.8-2.4t-a95b.toml b/providers/llmgateway-providers/models/novita/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..6cac460cd05
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,14 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8 2.4T A95B (NovitaAI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_000_000
diff --git a/providers/llmgateway-providers/models/novita/step-3.7-flash.toml b/providers/llmgateway-providers/models/novita/step-3.7-flash.toml
new file mode 100644
index 00000000000..ca5843d5fb8
--- /dev/null
+++ b/providers/llmgateway-providers/models/novita/step-3.7-flash.toml
@@ -0,0 +1,16 @@
+base_model = "stepfun/step-3.7-flash"
+base_model_omit = ["limit.input"]
+name = "Step 3.7 Flash (NovitaAI)"
+structured_output = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1.15
+cache_read = 0.04
+
+[limit]
+context = 262_144
diff --git a/providers/llmgateway-providers/models/together-ai/glm-5.3-flash.toml b/providers/llmgateway-providers/models/together-ai/glm-5.3-flash.toml
new file mode 100644
index 00000000000..af0987b5584
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/glm-5.3-flash.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3-flash"
+name = "GLM-5.3 Flash (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.15
+output = 0.5
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 943_717
diff --git a/providers/llmgateway-providers/models/together-ai/glm-5.3.toml b/providers/llmgateway-providers/models/together-ai/glm-5.3.toml
new file mode 100644
index 00000000000..67971580a73
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/glm-5.3.toml
@@ -0,0 +1,15 @@
+base_model = "zhipuai/glm-5.3"
+name = "GLM-5.3 (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "high", "max"]
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 943_717
diff --git a/providers/llmgateway-providers/models/together-ai/inkling.toml b/providers/llmgateway-providers/models/together-ai/inkling.toml
new file mode 100644
index 00000000000..1d6f4d10c7c
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/inkling.toml
@@ -0,0 +1,16 @@
+base_model = "thinkingmachines/inkling"
+name = "Inkling (Together AI)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1
+output = 4.05
+cache_read = 0.17
+
+[limit]
+context = 524_288
+output = 471_859
diff --git a/providers/llmgateway-providers/models/together-ai/muse-glimmer-30b.toml b/providers/llmgateway-providers/models/together-ai/muse-glimmer-30b.toml
new file mode 100644
index 00000000000..99056ca7d36
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/muse-glimmer-30b.toml
@@ -0,0 +1,14 @@
+base_model = "meta/muse-glimmer-30b"
+name = "Muse Glimmer 30B (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.35
+output = 1.5
+cache_read = 0.04
+
+[limit]
+output = 117_964
diff --git a/providers/llmgateway-providers/models/together-ai/qwen3.8-2.4t-a95b.toml b/providers/llmgateway-providers/models/together-ai/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..cd2eecc8a54
--- /dev/null
+++ b/providers/llmgateway-providers/models/together-ai/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8 2.4T A95B (Together AI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.25
+
+[limit]
+context = 1_010_000
+output = 909_000

From 397169451b894daf7605a3ef07341dad5ed8f062 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 17:24:42 +0000
Subject: [PATCH 1904/2012] chore(sync): update Kilo model catalog (#7764)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/anthropic/claude-opus-5.5.toml     | 14 ++++++++++++++
 .../kilo/models/~anthropic/claude-opus-latest.toml | 10 +++++-----
 .../kilo/models/~deepseek/deepseek-pro-latest.toml |  6 +++---
 3 files changed, 22 insertions(+), 8 deletions(-)
 create mode 100644 providers/kilo/models/anthropic/claude-opus-5.5.toml

diff --git a/providers/kilo/models/anthropic/claude-opus-5.5.toml b/providers/kilo/models/anthropic/claude-opus-5.5.toml
new file mode 100644
index 00000000000..53b1a0d9fb2
--- /dev/null
+++ b/providers/kilo/models/anthropic/claude-opus-5.5.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-opus-5-5"
+description = "Claude Opus 5.5 is Anthropic's flagship model for demanding reasoning, coding, and long-horizon agentic work, succeeding Claude Opus 5. It is particularly strong at multi-step changes in large codebases, code..."
+temperature = true
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/kilo/models/~anthropic/claude-opus-latest.toml b/providers/kilo/models/~anthropic/claude-opus-latest.toml
index 7eae0b14770..709f3bb1f1b 100644
--- a/providers/kilo/models/~anthropic/claude-opus-latest.toml
+++ b/providers/kilo/models/~anthropic/claude-opus-latest.toml
@@ -12,13 +12,13 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high", "xhigh", "max"]
+values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
 
 [limit]
 context = 1_000_000
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6565f5a8e9b..0d1469b0157 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.56364
-output = 1.69092
-cache_read = 0.018788
+input = 0.55572
+output = 1.66716
+cache_read = 0.018524
 
 [limit]
 context = 1_024_000

From f310c935bf304ed5f2e2e4fc24e4c61a2802b6ee Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 17:24:48 +0000
Subject: [PATCH 1905/2012] chore(sync): update Venice model catalog (#7766)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/venice/models/claude-opus-5-5.toml | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 providers/venice/models/claude-opus-5-5.toml

diff --git a/providers/venice/models/claude-opus-5-5.toml b/providers/venice/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..ff60fa3a6ef
--- /dev/null
+++ b/providers/venice/models/claude-opus-5-5.toml
@@ -0,0 +1,17 @@
+base_model = "anthropic/claude-opus-5-5"
+description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
+release_date = "2026-09-18"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4.8
+output = 24
+cache_read = 0.24
+cache_write = 6
+
+[modalities]
+input = ["text", "image"]

From c8129bec6c110bac7f769b784c19465f11e1a016 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 17:24:50 +0000
Subject: [PATCH 1906/2012] chore(sync): update OpenRouter model catalog
 (#7767)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-5.5.toml          | 14 ++++++++++++++
 .../models/deepseek/deepseek-v4-pro.toml           |  6 +++---
 .../models/~anthropic/claude-opus-latest.toml      | 11 ++++-------
 .../models/~deepseek/deepseek-pro-latest.toml      |  8 ++++----
 4 files changed, 25 insertions(+), 14 deletions(-)
 create mode 100644 providers/openrouter/models/anthropic/claude-opus-5.5.toml

diff --git a/providers/openrouter/models/anthropic/claude-opus-5.5.toml b/providers/openrouter/models/anthropic/claude-opus-5.5.toml
new file mode 100644
index 00000000000..70709244ff6
--- /dev/null
+++ b/providers/openrouter/models/anthropic/claude-opus-5.5.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-opus-5-5"
+description = "Flagship Claude model for deep reasoning, coding, and long-horizon agents"
+temperature = true
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 8ce9eb4af6a..844ff795f34 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.927768
-output = 1.855536
-cache_read = 0.077314
+input = 0.915936
+output = 1.831872
+cache_read = 0.076328
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~anthropic/claude-opus-latest.toml b/providers/openrouter/models/~anthropic/claude-opus-latest.toml
index f4dce5e6b4e..8057256dd8c 100644
--- a/providers/openrouter/models/~anthropic/claude-opus-latest.toml
+++ b/providers/openrouter/models/~anthropic/claude-opus-latest.toml
@@ -12,18 +12,15 @@ tool_call = true
 structured_output = true
 open_weights = false
 
-[[reasoning_options]]
-type = "toggle"
-
 [[reasoning_options]]
 type = "effort"
 values = ["low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 5
-output = 25
-cache_read = 0.5
-cache_write = 6.25
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
 
 [limit]
 context = 1_000_000
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index ff538327fb0..2a909029289 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.56364
-output = 1.69092
-cache_read = 0.018788
+input = 0.5544
+output = 1.6632
+cache_read = 0.01764
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 1bc95312f70441e339d5aa010ad656b9c7706952 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 17:25:07 +0000
Subject: [PATCH 1907/2012] chore(sync): update Merge Gateway model catalog
 (#7768)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../merge-gateway/models/anthropic/claude-opus-5-5.toml  | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/merge-gateway/models/anthropic/claude-opus-5-5.toml

diff --git a/providers/merge-gateway/models/anthropic/claude-opus-5-5.toml b/providers/merge-gateway/models/anthropic/claude-opus-5-5.toml
new file mode 100644
index 00000000000..5f5b832e6f8
--- /dev/null
+++ b/providers/merge-gateway/models/anthropic/claude-opus-5-5.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+reasoning_options = []
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5

From 1023d2a188ba0c65c3535fab3847d1ab45ed1669 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 18:32:21 +0000
Subject: [PATCH 1908/2012] chore(sync): update OpenRouter model catalog
 (#7774)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/openai/gpt-6-luna-pro.toml         | 36 +++++++++++++++++++
 .../openrouter/models/openai/gpt-6-luna.toml  | 36 +++++++++++++++++++
 .../models/openai/gpt-6-sol-pro.toml          | 36 +++++++++++++++++++
 .../openrouter/models/openai/gpt-6-sol.toml   | 36 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml |  6 ++--
 .../models/~openai/gpt-luna-latest.toml       | 16 ++++-----
 7 files changed, 158 insertions(+), 14 deletions(-)
 create mode 100644 providers/openrouter/models/openai/gpt-6-luna-pro.toml
 create mode 100644 providers/openrouter/models/openai/gpt-6-luna.toml
 create mode 100644 providers/openrouter/models/openai/gpt-6-sol-pro.toml
 create mode 100644 providers/openrouter/models/openai/gpt-6-sol.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 844ff795f34..33415e50e77 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.915936
-output = 1.831872
-cache_read = 0.076328
+input = 0.904104
+output = 1.808208
+cache_read = 0.075342
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-6-luna-pro.toml b/providers/openrouter/models/openai/gpt-6-luna-pro.toml
new file mode 100644
index 00000000000..f3f1dd66bc1
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-luna-pro.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Luna Pro"
+description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-6-luna.toml b/providers/openrouter/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..d4a1dfaba03
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-luna.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Luna"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-6-sol-pro.toml b/providers/openrouter/models/openai/gpt-6-sol-pro.toml
new file mode 100644
index 00000000000..a34a53e3bb1
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-sol-pro.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Sol Pro"
+description = "Frontier GPT model for professional reasoning, coding, and multimodal work"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-6-sol.toml b/providers/openrouter/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..0994e9d8867
--- /dev/null
+++ b/providers/openrouter/models/openai/gpt-6-sol.toml
@@ -0,0 +1,36 @@
+name = "GPT-6 Sol"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 2a909029289..786d9386078 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5544
-output = 1.6632
-cache_read = 0.01764
+input = 0.54384
+output = 1.63152
+cache_read = 0.017304
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~openai/gpt-luna-latest.toml b/providers/openrouter/models/~openai/gpt-luna-latest.toml
index e349af31375..5324cf22735 100644
--- a/providers/openrouter/models/~openai/gpt-luna-latest.toml
+++ b/providers/openrouter/models/~openai/gpt-luna-latest.toml
@@ -16,17 +16,17 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.2
-output = 1.2
-cache_read = 0.02
-cache_write = 0.25
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
 
 [[cost.tiers]]
 tier = { type = "context", size = 272_000 }
-input = 0.4
-output = 1.8
-cache_read = 0.04
-cache_write = 0.5
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
 
 [limit]
 context = 1_050_000

From e5299b8996d2bdc4399001a28b91e42dd198898d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 18:32:24 +0000
Subject: [PATCH 1909/2012] chore(sync): update NanoGPT model catalog (#7773)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-5.5.toml         | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 providers/nano-gpt/models/anthropic/claude-opus-5.5.toml

diff --git a/providers/nano-gpt/models/anthropic/claude-opus-5.5.toml b/providers/nano-gpt/models/anthropic/claude-opus-5.5.toml
new file mode 100644
index 00000000000..5948dbab4f8
--- /dev/null
+++ b/providers/nano-gpt/models/anthropic/claude-opus-5.5.toml
@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[limit]
+input = 1_000_000

From 18a5916204b7422110327e62536b72d24784c9d7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 18:32:39 +0000
Subject: [PATCH 1910/2012] chore(sync): update Kilo model catalog (#7775)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/openai/gpt-6-luna-pro.toml    | 29 +++++++++++++++++++
 providers/kilo/models/openai/gpt-6-luna.toml  | 29 +++++++++++++++++++
 .../kilo/models/openai/gpt-6-sol-pro.toml     | 29 +++++++++++++++++++
 providers/kilo/models/openai/gpt-6-sol.toml   | 29 +++++++++++++++++++
 .../models/~deepseek/deepseek-pro-latest.toml | 10 +++----
 .../kilo/models/~openai/gpt-luna-latest.toml  |  8 ++---
 6 files changed, 125 insertions(+), 9 deletions(-)
 create mode 100644 providers/kilo/models/openai/gpt-6-luna-pro.toml
 create mode 100644 providers/kilo/models/openai/gpt-6-luna.toml
 create mode 100644 providers/kilo/models/openai/gpt-6-sol-pro.toml
 create mode 100644 providers/kilo/models/openai/gpt-6-sol.toml

diff --git a/providers/kilo/models/openai/gpt-6-luna-pro.toml b/providers/kilo/models/openai/gpt-6-luna-pro.toml
new file mode 100644
index 00000000000..3e57329bcf1
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-luna-pro.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Luna Pro"
+description = "GPT-6 Luna Pro is the same underlying model as [GPT-6 Luna](https://openrouter.ai/openai/gpt-6-luna), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/openai/gpt-6-luna.toml b/providers/kilo/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..c6af65eba0a
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-luna.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Luna"
+description = "GPT-6 Luna is the fast, cost-efficient model in OpenAI's GPT-6 series, positioned below GPT-6 Sol. It is suited for high-volume and latency-sensitive workloads such as chat, classification, and lightweight agentic..."
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/openai/gpt-6-sol-pro.toml b/providers/kilo/models/openai/gpt-6-sol-pro.toml
new file mode 100644
index 00000000000..f1d263d64fa
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-sol-pro.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Sol Pro"
+description = "GPT-6 Sol Pro is the same underlying model as [GPT-6 Sol](https://openrouter.ai/openai/gpt-6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode"
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/openai/gpt-6-sol.toml b/providers/kilo/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..1fe6fe83662
--- /dev/null
+++ b/providers/kilo/models/openai/gpt-6-sol.toml
@@ -0,0 +1,29 @@
+name = "OpenAI: GPT-6 Sol"
+description = "GPT-6 Sol is the cost-efficient high-end model in OpenAI's GPT-6 series, positioned below the flagship GPT-6 Astra and above the fast GPT-6 Luna tier. It is suited for demanding professional..."
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["pdf", "image", "text"]
+output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 0d1469b0157..c3ebc81adc2 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.55572
-output = 1.66716
-cache_read = 0.018524
+input = 0.54384
+output = 1.63152
+cache_read = 0.017304
 
 [limit]
-context = 1_024_000
-output = 384_000
+context = 1_048_576
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~openai/gpt-luna-latest.toml b/providers/kilo/models/~openai/gpt-luna-latest.toml
index 6f3028a0af2..d593f0aca35 100644
--- a/providers/kilo/models/~openai/gpt-luna-latest.toml
+++ b/providers/kilo/models/~openai/gpt-luna-latest.toml
@@ -15,10 +15,10 @@ type = "effort"
 values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.2
-output = 1.2
-cache_read = 0.02
-cache_write = 0.25
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
 
 [limit]
 context = 1_050_000

From 5473d5b9f0484328e119d5a0215b4105cd5002eb Mon Sep 17 00:00:00 2001
From: R44VC0RP <89211796+R44VC0RP@users.noreply.github.com>
Date: Tue, 22 Sep 2026 18:33:19 +0000
Subject: [PATCH 1911/2012] feat(opencode): add Claude Opus 5.5 and GPT-6 Sol

---
 models/openai/gpt-6-sol.toml                  | 21 +++++++++++++++++++
 .../opencode/models/claude-opus-5-5.toml      | 12 +++++++++++
 providers/opencode/models/gpt-6-sol.toml      | 19 +++++++++++++++++
 3 files changed, 52 insertions(+)
 create mode 100644 models/openai/gpt-6-sol.toml
 create mode 100644 providers/opencode/models/claude-opus-5-5.toml
 create mode 100644 providers/opencode/models/gpt-6-sol.toml

diff --git a/models/openai/gpt-6-sol.toml b/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..824b5d86bed
--- /dev/null
+++ b/models/openai/gpt-6-sol.toml
@@ -0,0 +1,21 @@
+name = "GPT-6 Sol"
+description = "OpenAI model for complex coding and agentic workflows"
+family = "gpt-sol"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-04-20"
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/opencode/models/claude-opus-5-5.toml b/providers/opencode/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..e39cc456019
--- /dev/null
+++ b/providers/opencode/models/claude-opus-5-5.toml
@@ -0,0 +1,12 @@
+# Pricing: https://platform.claude.com/docs/en/models/opus-5-5/overview
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/anthropic"
diff --git a/providers/opencode/models/gpt-6-sol.toml b/providers/opencode/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..eefd5c4f407
--- /dev/null
+++ b/providers/opencode/models/gpt-6-sol.toml
@@ -0,0 +1,19 @@
+# Pricing and controls: https://developers.openai.com/api/docs/models/gpt-6-sol
+base_model = "openai/gpt-6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/openai"

From bfafa895d71dff8c2be417f85dc0d0713d6a27b7 Mon Sep 17 00:00:00 2001
From: R44VC0RP <89211796+R44VC0RP@users.noreply.github.com>
Date: Tue, 22 Sep 2026 18:34:47 +0000
Subject: [PATCH 1912/2012] feat(opencode): add GPT-6 Luna

---
 models/openai/gpt-6-luna.toml             | 21 +++++++++++++++++++++
 providers/opencode/models/gpt-6-luna.toml | 19 +++++++++++++++++++
 2 files changed, 40 insertions(+)
 create mode 100644 models/openai/gpt-6-luna.toml
 create mode 100644 providers/opencode/models/gpt-6-luna.toml

diff --git a/models/openai/gpt-6-luna.toml b/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..0a43442c0a5
--- /dev/null
+++ b/models/openai/gpt-6-luna.toml
@@ -0,0 +1,21 @@
+name = "GPT-6 Luna"
+description = "OpenAI's most efficient model for focused, high-volume tasks"
+family = "gpt-luna"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-05-18"
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/opencode/models/gpt-6-luna.toml b/providers/opencode/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..1263a99bdc4
--- /dev/null
+++ b/providers/opencode/models/gpt-6-luna.toml
@@ -0,0 +1,19 @@
+# Pricing and controls: https://developers.openai.com/api/docs/models/gpt-6-luna
+base_model = "openai/gpt-6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[provider]
+npm = "@ai-sdk/openai"

From ef12935def70ea07ba6c98964e94aa15e7148768 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:34:49 -0500
Subject: [PATCH 1913/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7765)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): add Claude Opus 5.5 effort options

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/anthropic/claude-opus-5.5-fast.toml    | 9 +++++++++
 providers/vercel/models/anthropic/claude-opus-5.5.toml   | 8 ++++++++
 2 files changed, 17 insertions(+)
 create mode 100644 providers/vercel/models/anthropic/claude-opus-5.5-fast.toml
 create mode 100644 providers/vercel/models/anthropic/claude-opus-5.5.toml

diff --git a/providers/vercel/models/anthropic/claude-opus-5.5-fast.toml b/providers/vercel/models/anthropic/claude-opus-5.5-fast.toml
new file mode 100644
index 00000000000..9e516c8d185
--- /dev/null
+++ b/providers/vercel/models/anthropic/claude-opus-5.5-fast.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-5-5"
+name = "Claude Opus 5.5 (Fast)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 8
+output = 40
+cache_read = 0.4
+cache_write = 10
diff --git a/providers/vercel/models/anthropic/claude-opus-5.5.toml b/providers/vercel/models/anthropic/claude-opus-5.5.toml
new file mode 100644
index 00000000000..8c3442bfe4b
--- /dev/null
+++ b/providers/vercel/models/anthropic/claude-opus-5.5.toml
@@ -0,0 +1,8 @@
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5

From bc4f954e23b43c27c590319f1081928d4f22f2a2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:35:23 -0500
Subject: [PATCH 1914/2012] chore(sync): update Requesty model catalog (#7769)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/claude-opus-5-5.toml   | 15 +++++++++++++++
 .../requesty/models/claude-opus-5-5@eu.toml      | 16 ++++++++++++++++
 2 files changed, 31 insertions(+)
 create mode 100644 providers/requesty/models/claude-opus-5-5.toml
 create mode 100644 providers/requesty/models/claude-opus-5-5@eu.toml

diff --git a/providers/requesty/models/claude-opus-5-5.toml b/providers/requesty/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..1c8fca5fbe5
--- /dev/null
+++ b/providers/requesty/models/claude-opus-5-5.toml
@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/requesty/models/claude-opus-5-5@eu.toml b/providers/requesty/models/claude-opus-5-5@eu.toml
new file mode 100644
index 00000000000..eaa19cb4ca0
--- /dev/null
+++ b/providers/requesty/models/claude-opus-5-5@eu.toml
@@ -0,0 +1,16 @@
+base_model = "anthropic/claude-opus-5-5"
+name = "Claude Opus 5.5 (EU)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 5.5
+output = 27.5
+cache_read = 0.55
+cache_write = 6.875

From 38ac62d15f1c70923d67fe4b622267037f9e5f49 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:39:47 -0500
Subject: [PATCH 1915/2012] fix: [missing-model] ofox:
 anthropic/claude-opus-5.5 (#7777)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../ofox/models/anthropic/claude-opus-5.5.toml   | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/anthropic/claude-opus-5.5.toml

diff --git a/providers/ofox/models/anthropic/claude-opus-5.5.toml b/providers/ofox/models/anthropic/claude-opus-5.5.toml
new file mode 100644
index 00000000000..23287398bb0
--- /dev/null
+++ b/providers/ofox/models/anthropic/claude-opus-5.5.toml
@@ -0,0 +1,16 @@
+# Sources:
+# https://ofox.ai/models/anthropic/claude-opus-5.5
+# https://platform.claude.com/docs/en/models/opus-5-5/overview
+# https://www.anthropic.com/claude-opus-5-5
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.ofox.ai/anthropic/v1"

From 5c476477d041eeb90e226971911ea0f8c47d3dee Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:40:36 -0500
Subject: [PATCH 1916/2012] fix: DeepSeek provider: mark retired aliases
 `deepseek-v4-flash` and `deepseek-v4-flash-vision-exp` as `status =
 deprecated` (#7727)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/deepseek/models/deepseek-v4-flash-vision-exp.toml | 3 +--
 providers/deepseek/models/deepseek-v4-flash.toml            | 3 +--
 2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml b/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
index e1227c7a467..aa1da80696b 100644
--- a/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
+++ b/providers/deepseek/models/deepseek-v4-flash-vision-exp.toml
@@ -6,8 +6,7 @@
 # https://api-docs.deepseek.com/quick_start/pricing (accessed 2026-09-10)
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4 Flash Vision Exp"
-## mark as deprecated soon
-# status = "deprecated"
+status = "deprecated"
 [[reasoning_options]]
 type = "toggle"
 
diff --git a/providers/deepseek/models/deepseek-v4-flash.toml b/providers/deepseek/models/deepseek-v4-flash.toml
index ae7ff2fc79b..3ebe57414da 100644
--- a/providers/deepseek/models/deepseek-v4-flash.toml
+++ b/providers/deepseek/models/deepseek-v4-flash.toml
@@ -6,8 +6,7 @@
 # https://api-docs.deepseek.com/quick_start/pricing (accessed 2026-09-10)
 base_model = "deepseek/deepseek-v4.1-flash"
 name = "DeepSeek V4 Flash"
-## mark as deprecated soon
-# status = "deprecated"
+status = "deprecated"
 
 [[reasoning_options]]
 type = "toggle"

From b165bd596d27e3fa271e5b859753939fc8962cb2 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:54:44 -0500
Subject: [PATCH 1917/2012] chore(sync): update Cloudflare AI Gateway model
 catalog (#7776)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/xai/grok-4.7.toml                  | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
 create mode 100644 providers/cloudflare-ai-gateway/models/xai/grok-4.7.toml

diff --git a/providers/cloudflare-ai-gateway/models/xai/grok-4.7.toml b/providers/cloudflare-ai-gateway/models/xai/grok-4.7.toml
new file mode 100644
index 00000000000..48df4f8248b
--- /dev/null
+++ b/providers/cloudflare-ai-gateway/models/xai/grok-4.7.toml
@@ -0,0 +1,19 @@
+base_model = "xai/grok-4.7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 4
+output = 12
+cache_read = 1
+
+[limit]
+context = 500_000

From 67cb71f2b5ea59a6d076a70a509cdbe663e152c7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:10:50 -0500
Subject: [PATCH 1918/2012] feat: add GPT-6 Sol and Luna (#7781)

Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 models/openai/gpt-6-luna.toml           | 21 +++++++++++++++++++++
 models/openai/gpt-6-sol.toml            | 21 +++++++++++++++++++++
 providers/azure/models/gpt-6-luna.toml  | 17 +++++++++++++++++
 providers/azure/models/gpt-6-sol.toml   | 17 +++++++++++++++++
 providers/openai/models/gpt-6-luna.toml | 24 ++++++++++++++++++++++++
 providers/openai/models/gpt-6-sol.toml  | 24 ++++++++++++++++++++++++
 6 files changed, 124 insertions(+)
 create mode 100644 models/openai/gpt-6-luna.toml
 create mode 100644 models/openai/gpt-6-sol.toml
 create mode 100644 providers/azure/models/gpt-6-luna.toml
 create mode 100644 providers/azure/models/gpt-6-sol.toml
 create mode 100644 providers/openai/models/gpt-6-luna.toml
 create mode 100644 providers/openai/models/gpt-6-sol.toml

diff --git a/models/openai/gpt-6-luna.toml b/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..0a43442c0a5
--- /dev/null
+++ b/models/openai/gpt-6-luna.toml
@@ -0,0 +1,21 @@
+name = "GPT-6 Luna"
+description = "OpenAI's most efficient model for focused, high-volume tasks"
+family = "gpt-luna"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-05-18"
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/models/openai/gpt-6-sol.toml b/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..824b5d86bed
--- /dev/null
+++ b/models/openai/gpt-6-sol.toml
@@ -0,0 +1,21 @@
+name = "GPT-6 Sol"
+description = "OpenAI model for complex coding and agentic workflows"
+family = "gpt-sol"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-04-20"
+open_weights = false
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/azure/models/gpt-6-luna.toml b/providers/azure/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..c105b8ef975
--- /dev/null
+++ b/providers/azure/models/gpt-6-luna.toml
@@ -0,0 +1,17 @@
+# Availability: https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability
+# Rates: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+base_model = "openai/gpt-6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.10
+output = 0.50
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.20
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/azure/models/gpt-6-sol.toml b/providers/azure/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..e0ad0d723d0
--- /dev/null
+++ b/providers/azure/models/gpt-6-sol.toml
@@ -0,0 +1,17 @@
+# Availability: https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability
+# Rates: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/
+base_model = "openai/gpt-6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.00
+output = 10.00
+cache_read = 0.20
+cache_write = 2.50
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4.00
+output = 15.00
+cache_read = 0.40
+cache_write = 5.00
diff --git a/providers/openai/models/gpt-6-luna.toml b/providers/openai/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..3a132f7fabb
--- /dev/null
+++ b/providers/openai/models/gpt-6-luna.toml
@@ -0,0 +1,24 @@
+# Pricing: https://developers.openai.com/api/docs/models/gpt-6-luna
+# Reasoning and modes: https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-luna
+base_model = "openai/gpt-6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.10
+output = 0.50
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.20
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[experimental.modes.fast]
+cost = { input = 0.20, output = 1.00, cache_read = 0.02, cache_write = 0.25 }
+provider = { body = { service_tier = "priority" } }
+
+[experimental.modes.pro]
+provider = { body = { reasoning = { mode = "pro" } } }
diff --git a/providers/openai/models/gpt-6-sol.toml b/providers/openai/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..e141e9653ad
--- /dev/null
+++ b/providers/openai/models/gpt-6-sol.toml
@@ -0,0 +1,24 @@
+# Pricing: https://developers.openai.com/api/docs/models/gpt-6-sol
+# Reasoning and modes: https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-sol
+base_model = "openai/gpt-6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 2.00
+output = 10.00
+cache_read = 0.20
+cache_write = 2.50
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4.00
+output = 15.00
+cache_read = 0.40
+cache_write = 5.00
+
+[experimental.modes.fast]
+cost = { input = 4.00, output = 20.00, cache_read = 0.40, cache_write = 5.00 }
+provider = { body = { service_tier = "priority" } }
+
+[experimental.modes.pro]
+provider = { body = { reasoning = { mode = "pro" } } }

From 921b79fcb0452721f464a79cca6e229475bbb277 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:11:38 -0500
Subject: [PATCH 1919/2012] chore(sync): update CoreWeave model catalog (#7762)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../wandb/models/google/gemma-4-26B-A4B-it.toml    | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
 create mode 100644 providers/wandb/models/google/gemma-4-26B-A4B-it.toml

diff --git a/providers/wandb/models/google/gemma-4-26B-A4B-it.toml b/providers/wandb/models/google/gemma-4-26B-A4B-it.toml
new file mode 100644
index 00000000000..8014a72bac7
--- /dev/null
+++ b/providers/wandb/models/google/gemma-4-26B-A4B-it.toml
@@ -0,0 +1,14 @@
+base_model = "google/gemma-4-26b-a4b-it"
+name = "Gemma 4 26B A4B"
+description = "Gemma 4 26B A4B is a multimodal MoE model with LoRA support and function calling for agentic workflows."
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.1
+output = 0.3
+cache_read = 0.05
+
+[limit]
+output = 262_144

From 9d18788bfe315796b5902a098b91359fcb7300dd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:17 +0000
Subject: [PATCH 1920/2012] chore(sync): update Kilo model catalog (#7787)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/openai/gpt-6-luna.toml  | 19 +------------------
 providers/kilo/models/openai/gpt-6-sol.toml   | 19 +------------------
 .../models/~deepseek/deepseek-pro-latest.toml |  8 ++++----
 .../~deepseek/deepseek-v4-flash-latest.toml   |  2 +-
 .../kilo/models/~moonshotai/kimi-latest.toml  |  6 +++---
 5 files changed, 10 insertions(+), 44 deletions(-)

diff --git a/providers/kilo/models/openai/gpt-6-luna.toml b/providers/kilo/models/openai/gpt-6-luna.toml
index c6af65eba0a..9e9940bc126 100644
--- a/providers/kilo/models/openai/gpt-6-luna.toml
+++ b/providers/kilo/models/openai/gpt-6-luna.toml
@@ -1,14 +1,5 @@
-name = "OpenAI: GPT-6 Luna"
+base_model = "openai/gpt-6-luna"
 description = "GPT-6 Luna is the fast, cost-efficient model in OpenAI's GPT-6 series, positioned below GPT-6 Sol. It is suited for high-volume and latency-sensitive workloads such as chat, classification, and lightweight agentic..."
-family = "gpt"
-release_date = "2026-09-22"
-last_updated = "2026-09-22"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,11 +10,3 @@ input = 0.1
 output = 0.5
 cache_read = 0.01
 cache_write = 0.125
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]
diff --git a/providers/kilo/models/openai/gpt-6-sol.toml b/providers/kilo/models/openai/gpt-6-sol.toml
index 1fe6fe83662..5960d039080 100644
--- a/providers/kilo/models/openai/gpt-6-sol.toml
+++ b/providers/kilo/models/openai/gpt-6-sol.toml
@@ -1,14 +1,5 @@
-name = "OpenAI: GPT-6 Sol"
+base_model = "openai/gpt-6-sol"
 description = "GPT-6 Sol is the cost-efficient high-end model in OpenAI's GPT-6 series, positioned below the flagship GPT-6 Astra and above the fast GPT-6 Luna tier. It is suited for demanding professional..."
-family = "gpt"
-release_date = "2026-09-22"
-last_updated = "2026-09-22"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -19,11 +10,3 @@ input = 2
 output = 10
 cache_read = 0.2
 cache_write = 2.5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index c3ebc81adc2..7666171face 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.54384
-output = 1.63152
-cache_read = 0.017304
+input = 0.4
+output = 4.3
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 6200a01f4d0..0412502ff5d 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -16,7 +16,7 @@ values = ["none", "low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 1
+output = 0.8
 cache_read = 0.008
 
 [limit]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index afff7105d92..366b401a83c 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
+input = 1.4989
+output = 10.758
+cache_read = 0.3
 
 [limit]
 context = 1_048_576

From d4374392a8ba4fb01ab330a1ad96a260cf78b62c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:20 +0000
Subject: [PATCH 1921/2012] chore(sync): update Merge Gateway model catalog
 (#7784)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-6-astra.toml              |  1 +
 .../merge-gateway/models/openai/gpt-6-luna.toml | 17 +++++++++++++++++
 .../merge-gateway/models/openai/gpt-6-sol.toml  | 17 +++++++++++++++++
 3 files changed, 35 insertions(+)
 create mode 100644 providers/merge-gateway/models/openai/gpt-6-luna.toml
 create mode 100644 providers/merge-gateway/models/openai/gpt-6-sol.toml

diff --git a/providers/merge-gateway/models/openai/gpt-6-astra.toml b/providers/merge-gateway/models/openai/gpt-6-astra.toml
index 223b48abe9e..da7cc6ca051 100644
--- a/providers/merge-gateway/models/openai/gpt-6-astra.toml
+++ b/providers/merge-gateway/models/openai/gpt-6-astra.toml
@@ -8,6 +8,7 @@ values = ["low", "medium", "high", "xhigh", "max"]
 input = 10
 output = 50
 cache_read = 1
+cache_write = 12.5
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/merge-gateway/models/openai/gpt-6-luna.toml b/providers/merge-gateway/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..7b7f12548e8
--- /dev/null
+++ b/providers/merge-gateway/models/openai/gpt-6-luna.toml
@@ -0,0 +1,17 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/merge-gateway/models/openai/gpt-6-sol.toml b/providers/merge-gateway/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..e0b3cd3f91b
--- /dev/null
+++ b/providers/merge-gateway/models/openai/gpt-6-sol.toml
@@ -0,0 +1,17 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[modalities]
+input = ["text", "image"]

From 74e8b5db5ab2c3d0fd136a9a202fb7429b3d5c07 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:22 +0000
Subject: [PATCH 1922/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7790)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/claude-opus-5-5.toml | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/llmgateway/models/claude-opus-5-5.toml

diff --git a/providers/llmgateway/models/claude-opus-5-5.toml b/providers/llmgateway/models/claude-opus-5-5.toml
new file mode 100644
index 00000000000..85ea7179304
--- /dev/null
+++ b/providers/llmgateway/models/claude-opus-5-5.toml
@@ -0,0 +1,11 @@
+base_model = "anthropic/claude-opus-5-5"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5

From bf1fc1b21f2e22a03e77db8262d6297d71756255 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:26 +0000
Subject: [PATCH 1923/2012] chore(sync): update NanoGPT model catalog (#7788)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/lumen-stealth.toml  | 23 +++++++++++++++
 .../nano-gpt/models/openai/gpt-6-sol-pro.toml | 29 +++++++++++++++++++
 .../nano-gpt/models/openai/gpt-6-sol.toml     | 15 ++++++++++
 3 files changed, 67 insertions(+)
 create mode 100644 providers/nano-gpt/models/lumen-stealth.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-sol.toml

diff --git a/providers/nano-gpt/models/lumen-stealth.toml b/providers/nano-gpt/models/lumen-stealth.toml
new file mode 100644
index 00000000000..13c7ea58999
--- /dev/null
+++ b/providers/nano-gpt/models/lumen-stealth.toml
@@ -0,0 +1,23 @@
+name = "Lumen Stealth"
+description = "Experimental multimodal model focused on reasoning, creative writing, roleplay, and agentic workflows. Available temporarily for evaluation ahead of public release. During this evaluation, prompts and responses are logged and may be reviewed to evaluate and improve the model. Do not send sensitive or confidential information."
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+tool_call = false
+structured_output = false
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.05
+output = 0
+
+[limit]
+context = 200_000
+input = 200_000
+output = 16_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml b/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
new file mode 100644
index 00000000000..daa2b93f111
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
@@ -0,0 +1,29 @@
+name = "GPT 6 Sol Pro"
+description = "GPT-6 Sol Pro uses the same underlying model as GPT-6 Sol with Pro reasoning mode enabled for higher-quality responses on complex tasks."
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-6-sol.toml b/providers/nano-gpt/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..109a0a67fce
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-sol.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-6-sol"
+name = "GPT 6 Sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[limit]
+input = 1_050_000

From 7120b4559017612dd0db8e2babe965bbdfc09021 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:33 +0000
Subject: [PATCH 1924/2012] chore(sync): update OpenRouter model catalog
 (#7791)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 +++---
 .../openrouter/models/openai/gpt-6-luna.toml  | 19 +------------------
 .../openrouter/models/openai/gpt-6-sol.toml   | 19 +------------------
 .../models/~deepseek/deepseek-pro-latest.toml |  8 ++++----
 .../~deepseek/deepseek-v4-flash-latest.toml   |  2 +-
 .../models/~moonshotai/kimi-latest.toml       |  6 +++---
 6 files changed, 13 insertions(+), 47 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 33415e50e77..74f20945d6e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.904104
-output = 1.808208
-cache_read = 0.075342
+input = 0.895578
+output = 1.791156
+cache_read = 0.074632
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/openai/gpt-6-luna.toml b/providers/openrouter/models/openai/gpt-6-luna.toml
index d4a1dfaba03..be8a0fb9e1b 100644
--- a/providers/openrouter/models/openai/gpt-6-luna.toml
+++ b/providers/openrouter/models/openai/gpt-6-luna.toml
@@ -1,14 +1,5 @@
-name = "GPT-6 Luna"
+base_model = "openai/gpt-6-luna"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-09-22"
-last_updated = "2026-09-22"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -26,11 +17,3 @@ input = 0.2
 output = 0.75
 cache_read = 0.02
 cache_write = 0.25
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]
diff --git a/providers/openrouter/models/openai/gpt-6-sol.toml b/providers/openrouter/models/openai/gpt-6-sol.toml
index 0994e9d8867..83f598548a3 100644
--- a/providers/openrouter/models/openai/gpt-6-sol.toml
+++ b/providers/openrouter/models/openai/gpt-6-sol.toml
@@ -1,14 +1,5 @@
-name = "GPT-6 Sol"
+base_model = "openai/gpt-6-sol"
 description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2026-09-22"
-last_updated = "2026-09-22"
-attachment = true
-reasoning = true
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = false
 
 [[reasoning_options]]
 type = "effort"
@@ -26,11 +17,3 @@ input = 4
 output = 15
 cache_read = 0.4
 cache_write = 5
-
-[limit]
-context = 1_050_000
-output = 128_000
-
-[modalities]
-input = ["pdf", "image", "text"]
-output = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 786d9386078..3112dca3caa 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.54384
-output = 1.63152
-cache_read = 0.017304
+input = 0.4
+output = 4.3
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 8c5f6c380c1..6c267e8f600 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -21,7 +21,7 @@ values = ["low", "high", "max"]
 
 [cost]
 input = 0.03
-output = 1
+output = 0.8
 cache_read = 0.008
 
 [limit]
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 0bb8650abd5..0aebd435d56 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
+input = 1.4989
+output = 10.758
+cache_read = 0.3
 
 [limit]
 context = 1_048_576

From 672754931632e7d652e517606bdf0f1804fbc4ba Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 19:23:39 +0000
Subject: [PATCH 1925/2012] chore(sync): update LLM Gateway model catalog
 (#7786)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-5-5.toml           | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/anthropic/claude-opus-5-5.toml

diff --git a/providers/llmgateway-providers/models/anthropic/claude-opus-5-5.toml b/providers/llmgateway-providers/models/anthropic/claude-opus-5-5.toml
new file mode 100644
index 00000000000..090b539ee3e
--- /dev/null
+++ b/providers/llmgateway-providers/models/anthropic/claude-opus-5-5.toml
@@ -0,0 +1,13 @@
+base_model = "anthropic/claude-opus-5-5"
+name = "Claude Opus 5.5 (Anthropic)"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5

From 3d6dcd13b178feed787d98a6c3e5efb9cc4f4c9d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 14:59:55 -0500
Subject: [PATCH 1926/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7785)

* chore(sync): update Vercel AI Gateway model catalog

* fix(vercel): list GPT-6 reasoning efforts

---------

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>
---
 .../vercel/models/openai/gpt-6-luna-fast.toml | 19 +++++++++++++++++++
 .../vercel/models/openai/gpt-6-luna.toml      | 18 ++++++++++++++++++
 .../vercel/models/openai/gpt-6-sol-fast.toml  | 19 +++++++++++++++++++
 providers/vercel/models/openai/gpt-6-sol.toml | 18 ++++++++++++++++++
 4 files changed, 74 insertions(+)
 create mode 100644 providers/vercel/models/openai/gpt-6-luna-fast.toml
 create mode 100644 providers/vercel/models/openai/gpt-6-luna.toml
 create mode 100644 providers/vercel/models/openai/gpt-6-sol-fast.toml
 create mode 100644 providers/vercel/models/openai/gpt-6-sol.toml

diff --git a/providers/vercel/models/openai/gpt-6-luna-fast.toml b/providers/vercel/models/openai/gpt-6-luna-fast.toml
new file mode 100644
index 00000000000..5087080071f
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-luna-fast.toml
@@ -0,0 +1,19 @@
+base_model = "openai/gpt-6-luna"
+name = "GPT-6 Luna (Fast)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.2
+output = 1
+cache_read = 0.02
+cache_write = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 0.4
+output = 1.5
+cache_read = 0.04
+cache_write = 0.5
diff --git a/providers/vercel/models/openai/gpt-6-luna.toml b/providers/vercel/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..c9115379e71
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-luna.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/vercel/models/openai/gpt-6-sol-fast.toml b/providers/vercel/models/openai/gpt-6-sol-fast.toml
new file mode 100644
index 00000000000..8b33bf6106a
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-sol-fast.toml
@@ -0,0 +1,19 @@
+base_model = "openai/gpt-6-sol"
+name = "GPT-6 Sol (Fast)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.4
+cache_write = 5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 8
+output = 30
+cache_read = 0.8
+cache_write = 10
diff --git a/providers/vercel/models/openai/gpt-6-sol.toml b/providers/vercel/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..39b8ede6517
--- /dev/null
+++ b/providers/vercel/models/openai/gpt-6-sol.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_001 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5

From 38ed0d71b6573ae3423d0a944a40ff50b1278a9d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 20:25:49 +0000
Subject: [PATCH 1927/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7798)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/gpt-6-luna.toml | 11 +++++++++++
 providers/llmgateway/models/gpt-6-sol.toml  | 11 +++++++++++
 2 files changed, 22 insertions(+)
 create mode 100644 providers/llmgateway/models/gpt-6-luna.toml
 create mode 100644 providers/llmgateway/models/gpt-6-sol.toml

diff --git a/providers/llmgateway/models/gpt-6-luna.toml b/providers/llmgateway/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..f058ea629ce
--- /dev/null
+++ b/providers/llmgateway/models/gpt-6-luna.toml
@@ -0,0 +1,11 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
diff --git a/providers/llmgateway/models/gpt-6-sol.toml b/providers/llmgateway/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..e4694368a0c
--- /dev/null
+++ b/providers/llmgateway/models/gpt-6-sol.toml
@@ -0,0 +1,11 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5

From 60b05473b5f603bb3c3e5803b4fc46d525a39b50 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 20:25:54 +0000
Subject: [PATCH 1928/2012] chore(sync): update LLM Gateway model catalog
 (#7799)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/openai/gpt-6-luna.toml                    | 12 ++++++++++++
 .../models/openai/gpt-6-sol.toml                     | 12 ++++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 providers/llmgateway-providers/models/openai/gpt-6-luna.toml
 create mode 100644 providers/llmgateway-providers/models/openai/gpt-6-sol.toml

diff --git a/providers/llmgateway-providers/models/openai/gpt-6-luna.toml b/providers/llmgateway-providers/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..6383b613669
--- /dev/null
+++ b/providers/llmgateway-providers/models/openai/gpt-6-luna.toml
@@ -0,0 +1,12 @@
+base_model = "openai/gpt-6-luna"
+name = "GPT-6 Luna (OpenAI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
diff --git a/providers/llmgateway-providers/models/openai/gpt-6-sol.toml b/providers/llmgateway-providers/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..88bcf456342
--- /dev/null
+++ b/providers/llmgateway-providers/models/openai/gpt-6-sol.toml
@@ -0,0 +1,12 @@
+base_model = "openai/gpt-6-sol"
+name = "GPT-6 Sol (OpenAI)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5

From 95bc3dad0012393187f48814cb1e09f95534c756 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 20:25:57 +0000
Subject: [PATCH 1929/2012] chore(sync): update NanoGPT model catalog (#7802)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/{ => nano}/lumen-stealth.toml      |  0
 .../models/openai/gpt-6-luna-pro.toml         | 29 +++++++++++++++++++
 .../nano-gpt/models/openai/gpt-6-luna.toml    | 15 ++++++++++
 .../nano-gpt/models/openai/gpt-6-sol-pro.toml |  2 +-
 .../nano-gpt/models/openai/gpt-6-sol.toml     |  2 +-
 5 files changed, 46 insertions(+), 2 deletions(-)
 rename providers/nano-gpt/models/{ => nano}/lumen-stealth.toml (100%)
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-luna-pro.toml
 create mode 100644 providers/nano-gpt/models/openai/gpt-6-luna.toml

diff --git a/providers/nano-gpt/models/lumen-stealth.toml b/providers/nano-gpt/models/nano/lumen-stealth.toml
similarity index 100%
rename from providers/nano-gpt/models/lumen-stealth.toml
rename to providers/nano-gpt/models/nano/lumen-stealth.toml
diff --git a/providers/nano-gpt/models/openai/gpt-6-luna-pro.toml b/providers/nano-gpt/models/openai/gpt-6-luna-pro.toml
new file mode 100644
index 00000000000..cc8bc7d726f
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-luna-pro.toml
@@ -0,0 +1,29 @@
+name = "GPT 6 Luna Pro"
+description = "GPT-6 Luna Pro uses the same underlying model as GPT-6 Luna with Pro reasoning mode enabled for higher-quality responses on complex tasks."
+family = "gpt"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.05
+output = 0.25
+cache_read = 0.005
+cache_write = 0.0625
+
+[limit]
+context = 1_050_000
+input = 1_050_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/openai/gpt-6-luna.toml b/providers/nano-gpt/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..23f8b252eec
--- /dev/null
+++ b/providers/nano-gpt/models/openai/gpt-6-luna.toml
@@ -0,0 +1,15 @@
+base_model = "openai/gpt-6-luna"
+name = "GPT 6 Luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.05
+output = 0.25
+cache_read = 0.005
+cache_write = 0.0625
+
+[limit]
+input = 1_050_000
diff --git a/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml b/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
index daa2b93f111..1720ef7ae10 100644
--- a/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
+++ b/providers/nano-gpt/models/openai/gpt-6-sol-pro.toml
@@ -11,7 +11,7 @@ open_weights = false
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 2
diff --git a/providers/nano-gpt/models/openai/gpt-6-sol.toml b/providers/nano-gpt/models/openai/gpt-6-sol.toml
index 109a0a67fce..ee3ffe25af3 100644
--- a/providers/nano-gpt/models/openai/gpt-6-sol.toml
+++ b/providers/nano-gpt/models/openai/gpt-6-sol.toml
@@ -3,7 +3,7 @@ name = "GPT 6 Sol"
 
 [[reasoning_options]]
 type = "effort"
-values = ["low", "medium", "high", "xhigh", "max"]
+values = ["none", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 2

From c1785c7e6c7be95e3067611da0458fd38644bbb4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 20:26:00 +0000
Subject: [PATCH 1930/2012] chore(sync): update Kilo model catalog (#7801)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/aion-labs/aion-2.0.toml       |  2 +-
 providers/kilo/models/aion-labs/aion-3.0-mini.toml  |  2 +-
 providers/kilo/models/aion-labs/aion-3.0.toml       |  2 +-
 .../kilo/models/deepseek/deepseek-v4.1-flash.toml   |  1 +
 providers/kilo/models/qwen/qwen3.8-omni-flash.toml  | 13 +++++++++++++
 .../models/~deepseek/deepseek-flash-latest.toml     |  6 +++---
 .../kilo/models/~deepseek/deepseek-pro-latest.toml  |  8 ++++----
 7 files changed, 24 insertions(+), 10 deletions(-)
 create mode 100644 providers/kilo/models/qwen/qwen3.8-omni-flash.toml

diff --git a/providers/kilo/models/aion-labs/aion-2.0.toml b/providers/kilo/models/aion-labs/aion-2.0.toml
index 46cc9f272f7..4bbbba9076b 100644
--- a/providers/kilo/models/aion-labs/aion-2.0.toml
+++ b/providers/kilo/models/aion-labs/aion-2.0.toml
@@ -19,7 +19,7 @@ output = 1.6
 cache_read = 0.2
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/aion-labs/aion-3.0-mini.toml b/providers/kilo/models/aion-labs/aion-3.0-mini.toml
index 84f2979d62c..4987c065dc7 100644
--- a/providers/kilo/models/aion-labs/aion-3.0-mini.toml
+++ b/providers/kilo/models/aion-labs/aion-3.0-mini.toml
@@ -19,7 +19,7 @@ output = 1.4
 cache_read = 0.18
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/aion-labs/aion-3.0.toml b/providers/kilo/models/aion-labs/aion-3.0.toml
index ac0a56ff9f9..caa24ee8921 100644
--- a/providers/kilo/models/aion-labs/aion-3.0.toml
+++ b/providers/kilo/models/aion-labs/aion-3.0.toml
@@ -19,7 +19,7 @@ output = 6
 cache_read = 0.75
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index 2d43303aad0..d755c215bcf 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,3 +12,4 @@ cache_read = 0.006
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/qwen/qwen3.8-omni-flash.toml b/providers/kilo/models/qwen/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..574800249bc
--- /dev/null
+++ b/providers/kilo/models/qwen/qwen3.8-omni-flash.toml
@@ -0,0 +1,13 @@
+base_model = "alibaba/qwen3.8-omni-flash"
+description = "Qwen3.8 Omni Flash is an omni-modal reasoning model from Alibaba, the first Qwen model built around agentic capabilities with native audio-video understanding. It is suited for audio-video analysis and summarization,..."
+temperature = true
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 2193a017b17..9e869daead9 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.12
-output = 0.48
-cache_read = 0.0036
+input = 0.1
+output = 0.8
+cache_read = 0.1
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 7666171face..6f1d7199aea 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4
-output = 4.3
-cache_read = 0.033
+input = 0.39996
+output = 1.19988
+cache_read = 0.012726
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From dadd2f0d0b534dc1652296a4f325287c2814ef36 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 20:26:13 +0000
Subject: [PATCH 1931/2012] chore(sync): update OpenRouter model catalog
 (#7800)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/aion-labs/aion-2.0.toml    |  2 +-
 .../models/aion-labs/aion-3.0-mini.toml          |  2 +-
 .../openrouter/models/aion-labs/aion-3.0.toml    |  2 +-
 .../models/deepseek/deepseek-v4-pro.toml         |  6 +++---
 .../models/deepseek/deepseek-v4.1-flash.toml     |  7 ++++---
 .../models/qwen/qwen3.8-omni-flash.toml          | 16 ++++++++++++++++
 .../models/~deepseek/deepseek-flash-latest.toml  |  6 +++---
 .../models/~deepseek/deepseek-pro-latest.toml    |  8 ++++----
 8 files changed, 33 insertions(+), 16 deletions(-)
 create mode 100644 providers/openrouter/models/qwen/qwen3.8-omni-flash.toml

diff --git a/providers/openrouter/models/aion-labs/aion-2.0.toml b/providers/openrouter/models/aion-labs/aion-2.0.toml
index 8e499e073cf..b54b64899f3 100644
--- a/providers/openrouter/models/aion-labs/aion-2.0.toml
+++ b/providers/openrouter/models/aion-labs/aion-2.0.toml
@@ -16,7 +16,7 @@ output = 1.6
 cache_read = 0.2
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/aion-labs/aion-3.0-mini.toml b/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
index 33c5d843701..af968c32457 100644
--- a/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
+++ b/providers/openrouter/models/aion-labs/aion-3.0-mini.toml
@@ -16,7 +16,7 @@ output = 1.4
 cache_read = 0.18
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/aion-labs/aion-3.0.toml b/providers/openrouter/models/aion-labs/aion-3.0.toml
index f16de215471..e25d916b2b3 100644
--- a/providers/openrouter/models/aion-labs/aion-3.0.toml
+++ b/providers/openrouter/models/aion-labs/aion-3.0.toml
@@ -16,7 +16,7 @@ output = 6
 cache_read = 0.75
 
 [limit]
-context = 1_048_576
+context = 131_072
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 74f20945d6e..b98c0880f9f 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.895578
-output = 1.791156
-cache_read = 0.074632
+input = 0.887226
+output = 1.774452
+cache_read = 0.073936
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..f8e214df2a7 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.1
+output = 0.8
+cache_read = 0.1
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/openrouter/models/qwen/qwen3.8-omni-flash.toml b/providers/openrouter/models/qwen/qwen3.8-omni-flash.toml
new file mode 100644
index 00000000000..d5769de2059
--- /dev/null
+++ b/providers/openrouter/models/qwen/qwen3.8-omni-flash.toml
@@ -0,0 +1,16 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+base_model = "alibaba/qwen3.8-omni-flash"
+temperature = true
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.15
+output = 0.47
+cache_read = 0.016
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 56426182daf..9c090dba1b3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.12
-output = 0.48
-cache_read = 0.0036
+input = 0.1
+output = 0.8
+cache_read = 0.1
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 3112dca3caa..5a32de51cb5 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.4
-output = 4.3
-cache_read = 0.033
+input = 0.39996
+output = 1.19988
+cache_read = 0.012726
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]

From 276e02f0972a6b572e9fbacfdea7f7af5b374734 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 21:23:32 +0000
Subject: [PATCH 1932/2012] chore(sync): update Venice model catalog (#7812)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../venice/models/openai-gpt-6-luna.toml      | 22 +++++++++++++++++++
 providers/venice/models/openai-gpt-6-sol.toml | 22 +++++++++++++++++++
 2 files changed, 44 insertions(+)
 create mode 100644 providers/venice/models/openai-gpt-6-luna.toml
 create mode 100644 providers/venice/models/openai-gpt-6-sol.toml

diff --git a/providers/venice/models/openai-gpt-6-luna.toml b/providers/venice/models/openai-gpt-6-luna.toml
new file mode 100644
index 00000000000..df056dfdc5c
--- /dev/null
+++ b/providers/venice/models/openai-gpt-6-luna.toml
@@ -0,0 +1,22 @@
+base_model = "openai/gpt-6-luna"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.125
+output = 0.625
+cache_read = 0.0125
+cache_write = 0.15625
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.25
+output = 0.9375
+cache_read = 0.025
+cache_write = 0.3125
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/venice/models/openai-gpt-6-sol.toml b/providers/venice/models/openai-gpt-6-sol.toml
new file mode 100644
index 00000000000..5a130d788ba
--- /dev/null
+++ b/providers/venice/models/openai-gpt-6-sol.toml
@@ -0,0 +1,22 @@
+base_model = "openai/gpt-6-sol"
+description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2.5
+output = 12.5
+cache_read = 0.25
+cache_write = 3.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 5
+output = 18.75
+cache_read = 0.5
+cache_write = 6.25
+
+[modalities]
+input = ["text", "image"]

From 3701e91036b401d801b8e860d836179c2e288d5e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 21:23:37 +0000
Subject: [PATCH 1933/2012] chore(sync): update OpenRouter model catalog
 (#7811)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml    | 7 +++----
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml   | 4 ++--
 4 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index b98c0880f9f..ead71f4358c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.887226
-output = 1.774452
-cache_read = 0.073936
+input = 0.877134
+output = 1.754268
+cache_read = 0.073095
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index f8e214df2a7..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,10 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.8
-cache_read = 0.1
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 9c090dba1b3..56426182daf 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.8
-cache_read = 0.1
+input = 0.12
+output = 0.48
+cache_read = 0.0036
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 0aebd435d56..9cb7f42fe0a 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4989
-output = 10.758
+input = 1.4
+output = 13
 cache_read = 0.3
 
 [limit]

From 2bc6cf3c6766209f70f0504df50da0b76cc55f90 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 21:24:23 +0000
Subject: [PATCH 1934/2012] chore(sync): update Kilo model catalog (#7809)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml    | 1 -
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml         | 4 ++--
 3 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index d755c215bcf..2d43303aad0 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,4 +12,3 @@ cache_read = 0.006
 
 [limit]
 context = 1_048_576
-output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 9e869daead9..2193a017b17 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.8
-cache_read = 0.1
+input = 0.12
+output = 0.48
+cache_read = 0.0036
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 366b401a83c..70e7e6ec8c8 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.4989
-output = 10.758
+input = 1.4
+output = 13
 cache_read = 0.3
 
 [limit]

From be826a62bb6bbb563d249e00ef1d12557c43c286 Mon Sep 17 00:00:00 2001
From: fwang <83515+fwang@users.noreply.github.com>
Date: Tue, 22 Sep 2026 21:53:31 +0000
Subject: [PATCH 1935/2012] feat(opencode): add Grok 4.7 with 30% off pricing

---
 providers/opencode/models/grok-4.7.toml | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 providers/opencode/models/grok-4.7.toml

diff --git a/providers/opencode/models/grok-4.7.toml b/providers/opencode/models/grok-4.7.toml
new file mode 100644
index 00000000000..e121d9b0cd5
--- /dev/null
+++ b/providers/opencode/models/grok-4.7.toml
@@ -0,0 +1,21 @@
+# Pricing: 30% off OpenCode Go Grok 4.7 rates, including long-context tiers.
+base_model = "xai/grok-4.7"
+name = "Grok 4.7 (30% Off)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1.4
+output = 4.2
+cache_read = 0.35
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2.8
+output = 8.4
+cache_read = 0.7
+
+[provider]
+npm = "@ai-sdk/openai"

From f1b0b8c791c2ea32eff83c242cc27c4415c22bb0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:24:37 +0000
Subject: [PATCH 1936/2012] chore(sync): update Kilo model catalog (#7817)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml    | 1 +
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml         | 6 +++---
 providers/kilo/models/~z-ai/glm-latest.toml                | 6 +++---
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index 2d43303aad0..d755c215bcf 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,3 +12,4 @@ cache_read = 0.006
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 2193a017b17..0295282190d 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.12
-output = 0.48
-cache_read = 0.0036
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 70e7e6ec8c8..afff7105d92 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 13
-cache_read = 0.3
+input = 1.5
+output = 7.5
+cache_read = 0.15
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 0facfa991b5..2ce00d3e342 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.6538
-output = 2.0548
-cache_read = 0.12142
+input = 0.5625
+output = 2.5
+cache_read = 0.125
 
 [limit]
 context = 1_048_576

From e84899e0363bb0e84a6ee04ecaf61883acde210f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:24:48 +0000
Subject: [PATCH 1937/2012] chore(sync): update OpenRouter model catalog
 (#7818)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml  | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml    | 7 ++++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml   | 6 +++---
 providers/openrouter/models/~z-ai/glm-latest.toml          | 6 +++---
 5 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index ead71f4358c..9472d4dda50 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.877134
-output = 1.754268
-cache_read = 0.073095
+input = 0.865302
+output = 1.730604
+cache_read = 0.072109
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..ee75b6ad088 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 56426182daf..a3bff918037 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.12
-output = 0.48
-cache_read = 0.0036
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 9cb7f42fe0a..0bb8650abd5 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4
-output = 13
-cache_read = 0.3
+input = 1.5
+output = 7.5
+cache_read = 0.15
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index e0e1e1ca329..9bbf416c3f6 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.6538
-output = 2.0548
-cache_read = 0.12142
+input = 0.5625
+output = 2.5
+cache_read = 0.125
 
 [limit]
 context = 1_310_720

From 414bfa0469d94a427a86279773518479fb45b6fd Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 23:23:38 +0000
Subject: [PATCH 1938/2012] chore(sync): update OpenRouter model catalog
 (#7822)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/cohere/command-a-plus.toml         | 29 +++++++++++++++++++
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 providers/openrouter/models/z-ai/glm-5.3.toml |  6 ++--
 .../models/~moonshotai/kimi-latest.toml       |  6 ++--
 .../models/~z-ai/glm-flash-latest.toml        |  2 +-
 .../openrouter/models/~z-ai/glm-latest.toml   |  6 ++--
 6 files changed, 42 insertions(+), 13 deletions(-)
 create mode 100644 providers/openrouter/models/cohere/command-a-plus.toml

diff --git a/providers/openrouter/models/cohere/command-a-plus.toml b/providers/openrouter/models/cohere/command-a-plus.toml
new file mode 100644
index 00000000000..a3ee649020e
--- /dev/null
+++ b/providers/openrouter/models/cohere/command-a-plus.toml
@@ -0,0 +1,29 @@
+# Toggle: reasoning.enabled = true|false
+# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
+name = "Command A+"
+description = "Cohere command model for multilingual enterprise agents, tools, and chat"
+family = "command-a"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "toggle"
+
+[cost]
+input = 0.3
+output = 1.5
+cache_read = 0.15
+
+[limit]
+context = 192_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 9472d4dda50..19122a2a8b0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.865302
-output = 1.730604
-cache_read = 0.072109
+input = 0.856776
+output = 1.713552
+cache_read = 0.071398
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 7c1154ec1bd..5422c468088 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.6538
-output = 2.0548
-cache_read = 0.12142
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 0bb8650abd5..c97ad1a78e9 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
+input = 1.35
+output = 14.33
+cache_read = 0.3
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~z-ai/glm-flash-latest.toml b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
index 8f90bbaaf93..c32fca67d4b 100644
--- a/providers/openrouter/models/~z-ai/glm-flash-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_310_720
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 9bbf416c3f6..42058ee6fca 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5625
-output = 2.5
-cache_read = 0.125
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_310_720

From ead2794b307369126f84794e9d355a05c0ace6fa Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 23:27:59 +0000
Subject: [PATCH 1939/2012] chore(sync): update Kilo model catalog (#7821)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/cohere/command-a-plus.toml    | 28 +++++++++++++++++++
 .../kilo/models/~moonshotai/kimi-latest.toml  |  6 ++--
 .../kilo/models/~z-ai/glm-flash-latest.toml   |  2 +-
 providers/kilo/models/~z-ai/glm-latest.toml   |  6 ++--
 4 files changed, 35 insertions(+), 7 deletions(-)
 create mode 100644 providers/kilo/models/cohere/command-a-plus.toml

diff --git a/providers/kilo/models/cohere/command-a-plus.toml b/providers/kilo/models/cohere/command-a-plus.toml
new file mode 100644
index 00000000000..dcee2ea1432
--- /dev/null
+++ b/providers/kilo/models/cohere/command-a-plus.toml
@@ -0,0 +1,28 @@
+name = "Cohere: Command A+"
+description = "Command A+ is Cohere's flagship model for enterprise agentic workflows. It accepts text and image inputs with a 192K context window, supports native tool calling with strict tool schemas, structured..."
+family = "command-a"
+release_date = "2026-09-22"
+last_updated = "2026-09-22"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "high"]
+
+[cost]
+input = 0.3
+output = 1.5
+cache_read = 0.15
+
+[limit]
+context = 192_000
+output = 64_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index afff7105d92..5677dcd7e7e 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.5
-output = 7.5
-cache_read = 0.15
+input = 1.35
+output = 14.33
+cache_read = 0.3
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~z-ai/glm-flash-latest.toml b/providers/kilo/models/~z-ai/glm-flash-latest.toml
index aee70deebb0..e4ce4e5dfe6 100644
--- a/providers/kilo/models/~z-ai/glm-flash-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-flash-latest.toml
@@ -21,7 +21,7 @@ cache_read = 0.015
 
 [limit]
 context = 1_048_576
-output = 943_718
+output = 131_072
 
 [modalities]
 input = ["text", "image", "video"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2ce00d3e342..d891278d7ef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5625
-output = 2.5
-cache_read = 0.125
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_048_576

From 90cc958ca96f1bd0691f0b554eefd6355fc5744b Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 00:58:11 +0000
Subject: [PATCH 1940/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7824)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/gemini-2.5-flash-image.toml       |  3 +--
 .../vercel/models/google/gemini-2.5-flash-lite.toml | 13 +++++++++++--
 .../vercel/models/google/gemini-3.7-flash.toml      |  1 +
 .../vercel/models/google/gemini-3.8-flash.toml      |  1 +
 4 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/providers/vercel/models/google/gemini-2.5-flash-image.toml b/providers/vercel/models/google/gemini-2.5-flash-image.toml
index 1ea71885c13..115f8bd626c 100644
--- a/providers/vercel/models/google/gemini-2.5-flash-image.toml
+++ b/providers/vercel/models/google/gemini-2.5-flash-image.toml
@@ -2,7 +2,6 @@ base_model = "google/gemini-2.5-flash-image"
 name = "Nano Banana (Gemini 2.5 Flash Image)"
 attachment = false
 reasoning = false
-knowledge = "2025-01"
 
 [cost]
 input = 0.3
@@ -10,4 +9,4 @@ output = 2.5
 cache_read = 0.03
 
 [limit]
-output = 65_536
+output = 65_535
diff --git a/providers/vercel/models/google/gemini-2.5-flash-lite.toml b/providers/vercel/models/google/gemini-2.5-flash-lite.toml
index 9dbd105b8d4..c77c8d9fb11 100644
--- a/providers/vercel/models/google/gemini-2.5-flash-lite.toml
+++ b/providers/vercel/models/google/gemini-2.5-flash-lite.toml
@@ -1,12 +1,21 @@
 base_model = "google/gemini-2.5-flash-lite"
-reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 512, max = 24_576 }]
 name = "Gemini 2.5 Flash Lite"
 
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 512
+max = 24_576
+
 [cost]
 input = 0.1
 output = 0.4
 cache_read = 0.01
 
+[limit]
+output = 65_535
+
 [modalities]
 input = ["text", "image", "pdf"]
-output = ["text"]
diff --git a/providers/vercel/models/google/gemini-3.7-flash.toml b/providers/vercel/models/google/gemini-3.7-flash.toml
index 9a978ed0a3b..c75cf2fdf40 100644
--- a/providers/vercel/models/google/gemini-3.7-flash.toml
+++ b/providers/vercel/models/google/gemini-3.7-flash.toml
@@ -12,6 +12,7 @@ cache_read = 0.075
 
 [limit]
 context = 1_000_000
+output = 65_535
 
 [modalities]
 input = ["text", "image", "pdf"]
diff --git a/providers/vercel/models/google/gemini-3.8-flash.toml b/providers/vercel/models/google/gemini-3.8-flash.toml
index 32fddda9360..39eb1bf1658 100644
--- a/providers/vercel/models/google/gemini-3.8-flash.toml
+++ b/providers/vercel/models/google/gemini-3.8-flash.toml
@@ -8,6 +8,7 @@ cache_read = 0.075
 
 [limit]
 context = 1_000_000
+output = 65_535
 
 [modalities]
 input = ["text", "image", "pdf"]

From 0d39179af05da0f714c4a6024092b18f0a502659 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 00:58:14 +0000
Subject: [PATCH 1941/2012] chore(sync): update OpenRouter model catalog
 (#7823)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml | 6 +++---
 providers/openrouter/models/tencent/hy3.toml              | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 6 +++---
 6 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 141d4b41ca6..6f817c5a0a0 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.049
-output = 0.098
-cache_read = 0.0098
+input = 0.088606
+output = 0.177212
+cache_read = 0.017721
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index 19122a2a8b0..e7683af4617 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.856776
-output = 1.713552
-cache_read = 0.071398
+input = 0.95526
+output = 1.91052
+cache_read = 0.079605
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index f61fa3175e9..80ecfdd3aa8 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.132
+output = 0.528
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 5422c468088..8b9af97ef26 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5614
-output = 1.7644
-cache_read = 0.10426
+input = 0.84
+output = 2.64
+cache_read = 0.156
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 5a32de51cb5..3112dca3caa 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.39996
-output = 1.19988
-cache_read = 0.012726
+input = 0.4
+output = 4.3
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 42058ee6fca..9bbf416c3f6 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5614
-output = 1.7644
-cache_read = 0.10426
+input = 0.5625
+output = 2.5
+cache_read = 0.125
 
 [limit]
 context = 1_310_720

From bd64c9c1d797ca6d681d46aa914cb7b1336444dc Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 00:58:22 +0000
Subject: [PATCH 1942/2012] chore(sync): update Kilo model catalog (#7825)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/tencent/hy3.toml                   | 6 +++---
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml              | 6 +++---
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index e96e3110ab1..c3469c1e9dc 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.0825
-output = 0.33
-cache_read = 0.020625
+input = 0.13
+output = 0.53
+cache_read = 0.033
 
 [limit]
 context = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 6f1d7199aea..7666171face 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.39996
-output = 1.19988
-cache_read = 0.012726
+input = 0.4
+output = 4.3
+cache_read = 0.033
 
 [limit]
 context = 1_048_576
-output = 393_216
+output = 384_000
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index d891278d7ef..2ce00d3e342 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5614
-output = 1.7644
-cache_read = 0.10426
+input = 0.5625
+output = 2.5
+cache_read = 0.125
 
 [limit]
 context = 1_048_576

From c4b04c6643cfa9235bff7bdd84f8aefda014eea5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 01:35:18 +0000
Subject: [PATCH 1943/2012] chore(sync): update OpenRouter model catalog
 (#7826)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index c97ad1a78e9..599b9319e64 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.35
-output = 14.33
+input = 1.45
+output = 11.7
 cache_read = 0.3
 
 [limit]

From cd9335ab37c70978becb7ca96a0da6e4749f9d88 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 01:35:22 +0000
Subject: [PATCH 1944/2012] chore(sync): update NanoGPT model catalog (#7827)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

From faa9cd00f2e52ab6b00210d2637350061fa3b83e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 01:35:28 +0000
Subject: [PATCH 1945/2012] chore(sync): update Kilo model catalog (#7828)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 5677dcd7e7e..7113c49b465 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.35
-output = 14.33
+input = 1.45
+output = 11.7
 cache_read = 0.3
 
 [limit]

From 123647e6f9daf344871d6540cd0516629f598b1a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 02:34:55 +0000
Subject: [PATCH 1946/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7829)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/anthropic/claude-opus-4.toml   | 3 ---
 providers/vercel/models/anthropic/claude-sonnet-4.toml | 1 -
 2 files changed, 4 deletions(-)

diff --git a/providers/vercel/models/anthropic/claude-opus-4.toml b/providers/vercel/models/anthropic/claude-opus-4.toml
index a80b0442665..2553fa1f326 100644
--- a/providers/vercel/models/anthropic/claude-opus-4.toml
+++ b/providers/vercel/models/anthropic/claude-opus-4.toml
@@ -6,6 +6,3 @@ input = 15
 output = 75
 cache_read = 1.5
 cache_write = 18.75
-
-[limit]
-output = 8_192
diff --git a/providers/vercel/models/anthropic/claude-sonnet-4.toml b/providers/vercel/models/anthropic/claude-sonnet-4.toml
index 61e522dfd97..7e189ae395a 100644
--- a/providers/vercel/models/anthropic/claude-sonnet-4.toml
+++ b/providers/vercel/models/anthropic/claude-sonnet-4.toml
@@ -16,4 +16,3 @@ cache_write = 7.5
 
 [limit]
 context = 1_000_000
-output = 8_192

From 98852586f2e5f1802c345abebd8772c12fd31a8f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 02:34:58 +0000
Subject: [PATCH 1947/2012] chore(sync): update OpenRouter model catalog
 (#7830)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 599b9319e64..e135fc2e21e 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.45
-output = 11.7
+input = 1.05
+output = 13
 cache_read = 0.3
 
 [limit]

From 3d93b844f404d7378880a4ef6e2491d3a11b1f38 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 02:35:00 +0000
Subject: [PATCH 1948/2012] chore(sync): update Kilo model catalog (#7831)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 7113c49b465..daf2852a1ed 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.45
-output = 11.7
+input = 1.05
+output = 13
 cache_read = 0.3
 
 [limit]

From 1dc15b2429516045fae614efcba4c3388e9a7729 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 03:32:01 +0000
Subject: [PATCH 1949/2012] chore(sync): update CrossModel model catalog
 (#7834)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic/claude-opus-5-5.toml     | 14 +++++++++++++
 .../crossmodel/models/openai/gpt-6-luna.toml  | 21 +++++++++++++++++++
 .../crossmodel/models/openai/gpt-6-sol.toml   | 21 +++++++++++++++++++
 3 files changed, 56 insertions(+)
 create mode 100644 providers/crossmodel/models/anthropic/claude-opus-5-5.toml
 create mode 100644 providers/crossmodel/models/openai/gpt-6-luna.toml
 create mode 100644 providers/crossmodel/models/openai/gpt-6-sol.toml

diff --git a/providers/crossmodel/models/anthropic/claude-opus-5-5.toml b/providers/crossmodel/models/anthropic/claude-opus-5-5.toml
new file mode 100644
index 00000000000..d9abb20946e
--- /dev/null
+++ b/providers/crossmodel/models/anthropic/claude-opus-5-5.toml
@@ -0,0 +1,14 @@
+base_model = "anthropic/claude-opus-5-5"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/crossmodel/models/openai/gpt-6-luna.toml b/providers/crossmodel/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..e91e64b7330
--- /dev/null
+++ b/providers/crossmodel/models/openai/gpt-6-luna.toml
@@ -0,0 +1,21 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/crossmodel/models/openai/gpt-6-sol.toml b/providers/crossmodel/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..56437cf29c3
--- /dev/null
+++ b/providers/crossmodel/models/openai/gpt-6-sol.toml
@@ -0,0 +1,21 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5
+
+[modalities]
+input = ["text", "image"]

From 446239c4f17c3d6b441d04c5462ff2bcb687a65a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 03:32:09 +0000
Subject: [PATCH 1950/2012] chore(sync): update Kilo model catalog (#7833)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~moonshotai/kimi-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index daf2852a1ed..366b401a83c 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,8 +15,8 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.05
-output = 13
+input = 1.4989
+output = 10.758
 cache_read = 0.3
 
 [limit]

From f6f06181d6887d7c7cd445aafafe4ae900f19299 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 03:32:14 +0000
Subject: [PATCH 1951/2012] chore(sync): update OpenRouter model catalog
 (#7835)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/~moonshotai/kimi-latest.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index e135fc2e21e..0aebd435d56 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,8 +20,8 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.05
-output = 13
+input = 1.4989
+output = 10.758
 cache_read = 0.3
 
 [limit]

From 1f9dbe7e9eb58e7eb82405fd039159362bb91134 Mon Sep 17 00:00:00 2001
From: Vladimir Glafirov <vglafirov@gitlab.com>
Date: Wed, 23 Sep 2026 05:54:28 +0200
Subject: [PATCH 1952/2012] feat: add gitlab duo-chat-opus-5-5 model (#7794)

* feat: add gitlab duo-chat-opus-5-5 model

* fix: use base_model inheritance for duo-chat-opus-5-5

Rewrite as base_model = "anthropic/claude-opus-5-5" plus only the
provider-specific delta (name, reasoning_options, zero cost), instead
of a full standalone duplicate. This inherits description, family,
dates, knowledge (2026-06), booleans, [limit], and [modalities] from
the existing lab entry, matching the pattern used by the sibling
duo-chat-opus-5 / duo-chat-sonnet-5 files.
---
 providers/gitlab/models/duo-chat-opus-5-5.toml | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100644 providers/gitlab/models/duo-chat-opus-5-5.toml

diff --git a/providers/gitlab/models/duo-chat-opus-5-5.toml b/providers/gitlab/models/duo-chat-opus-5-5.toml
new file mode 100644
index 00000000000..acfba463a94
--- /dev/null
+++ b/providers/gitlab/models/duo-chat-opus-5-5.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-5-5"
+name = "Agentic Chat (Claude Opus 5.5)"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

From 912dc39011be121800c239a18805f8d568ac7cc9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:55:15 -0500
Subject: [PATCH 1953/2012] chore(sync): update Requesty model catalog (#7810)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/requesty/models/gpt-6-luna.toml    | 19 +++++++++++++++++++
 providers/requesty/models/gpt-6-luna@eu.toml | 20 ++++++++++++++++++++
 providers/requesty/models/gpt-6-sol.toml     | 19 +++++++++++++++++++
 providers/requesty/models/gpt-6-sol@eu.toml  | 20 ++++++++++++++++++++
 4 files changed, 78 insertions(+)
 create mode 100644 providers/requesty/models/gpt-6-luna.toml
 create mode 100644 providers/requesty/models/gpt-6-luna@eu.toml
 create mode 100644 providers/requesty/models/gpt-6-sol.toml
 create mode 100644 providers/requesty/models/gpt-6-sol@eu.toml

diff --git a/providers/requesty/models/gpt-6-luna.toml b/providers/requesty/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..a11e192297d
--- /dev/null
+++ b/providers/requesty/models/gpt-6-luna.toml
@@ -0,0 +1,19 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
diff --git a/providers/requesty/models/gpt-6-luna@eu.toml b/providers/requesty/models/gpt-6-luna@eu.toml
new file mode 100644
index 00000000000..b0d9e112fcc
--- /dev/null
+++ b/providers/requesty/models/gpt-6-luna@eu.toml
@@ -0,0 +1,20 @@
+base_model = "openai/gpt-6-luna"
+name = "GPT-6 Luna (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.12
+output = 0.6
+cache_read = 0.012
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.24
+output = 0.9
+cache_read = 0.024
diff --git a/providers/requesty/models/gpt-6-sol.toml b/providers/requesty/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..fccceb9056e
--- /dev/null
+++ b/providers/requesty/models/gpt-6-sol.toml
@@ -0,0 +1,19 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
diff --git a/providers/requesty/models/gpt-6-sol@eu.toml b/providers/requesty/models/gpt-6-sol@eu.toml
new file mode 100644
index 00000000000..357ef21ce6c
--- /dev/null
+++ b/providers/requesty/models/gpt-6-sol@eu.toml
@@ -0,0 +1,20 @@
+base_model = "openai/gpt-6-sol"
+name = "GPT-6 Sol (EU)"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "max"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 2.4
+output = 12
+cache_read = 0.24
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4.8
+output = 18
+cache_read = 0.48

From c35bce144cb4afd55edc104b7313b0abb06c0137 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:56:31 -0500
Subject: [PATCH 1954/2012] fix: [missing-model] github-copilot: gpt-6-luna
 (#7813)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../github-copilot/models/gpt-6-luna.toml     | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/github-copilot/models/gpt-6-luna.toml

diff --git a/providers/github-copilot/models/gpt-6-luna.toml b/providers/github-copilot/models/gpt-6-luna.toml
new file mode 100644
index 00000000000..bd1d21a407f
--- /dev/null
+++ b/providers/github-copilot/models/gpt-6-luna.toml
@@ -0,0 +1,20 @@
+# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# GPT-6 Luna rates: ≤272K $0.10/$0.01/$0.125/$0.50; >272K $0.20/$0.02/$0.25/$0.75 per 1M (input/cached/cache write/output)
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25

From e68f03d3561839bf645aca27178eff95fb5fe16c Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:56:40 -0500
Subject: [PATCH 1955/2012] fix: [missing-model] github-copilot: gpt-6-sol
 (#7814)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../github-copilot/models/gpt-6-sol.toml      | 20 +++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/github-copilot/models/gpt-6-sol.toml

diff --git a/providers/github-copilot/models/gpt-6-sol.toml b/providers/github-copilot/models/gpt-6-sol.toml
new file mode 100644
index 00000000000..209b909b0be
--- /dev/null
+++ b/providers/github-copilot/models/gpt-6-sol.toml
@@ -0,0 +1,20 @@
+# Pricing: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# GPT-6 Sol rates: ≤272K $2/$0.20/$2.50/$10; >272K $4/$0.40/$5.00/$15 per 1M (input/cached/cache write/output)
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5

From c7e9597dc37fadc4a4398123b107539b03d818c3 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:57:25 -0500
Subject: [PATCH 1956/2012] fix: [missing-model] pioneer:
 fastino/gliner2.5-multi-v1 (#7793)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/fastino/gliner2.5-multi-v1.toml    | 35 +++++++++++++++++++
 1 file changed, 35 insertions(+)
 create mode 100644 providers/pioneer/models/fastino/gliner2.5-multi-v1.toml

diff --git a/providers/pioneer/models/fastino/gliner2.5-multi-v1.toml b/providers/pioneer/models/fastino/gliner2.5-multi-v1.toml
new file mode 100644
index 00000000000..932fc2e0d70
--- /dev/null
+++ b/providers/pioneer/models/fastino/gliner2.5-multi-v1.toml
@@ -0,0 +1,35 @@
+# Sources:
+# - Pioneer live catalog GET https://api.pioneer.ai/base-models (id fastino/gliner2.5-multi-v1: label "GLiNER 2.5 Multi", description "Multilingual boundary NER and span extraction; non-trainable encoder.", context_window 4096, input/output $0.15, cache $0.15, license Apache-2.0, release Aug 2026, is_chat_model true)
+# - Model card https://huggingface.co/fastino/gliner2.5-multi-v1 (287M mDeBERTa-v3-base multilingual boundary checkpoint, max_len 4096, Apache-2.0)
+# - Release announcement https://fastino.ai/blog/gliner2-5-span-free-information-extraction (released Aug 24, 2026; three Apache-2.0 variants)
+# Pioneer is the first-party Fastino inference API (https://pioneer.ai/, https://www.prnewswire.com/news-releases/fastino-launches-pioneer-the-first-agent-for-fine-tuning-and-inference-of-llms-302748105.html); full inline entry follows existing providers/pioneer/models/fastino/* pattern.
+name = "GLiNER 2.5 Multi"
+description = "Multilingual boundary NER and span extraction; non-trainable encoder."
+release_date = "2026-08-24"
+last_updated = "2026-08-24"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.15
+output = 0.15
+cache_read = 0.15
+cache_write = 0.15
+
+[limit]
+context = 4_096
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]

From c3fd2c89b4485817795b54c63c4859e8290374af Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:57:32 -0500
Subject: [PATCH 1957/2012] fix: [missing-model] ofox: x-ai/grok-4.7 (#7792)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/x-ai/grok-4.7.toml | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 providers/ofox/models/x-ai/grok-4.7.toml

diff --git a/providers/ofox/models/x-ai/grok-4.7.toml b/providers/ofox/models/x-ai/grok-4.7.toml
new file mode 100644
index 00000000000..b829098d460
--- /dev/null
+++ b/providers/ofox/models/x-ai/grok-4.7.toml
@@ -0,0 +1,16 @@
+# Ofox: x-ai/grok-4.7 → xAI Grok 4.7
+# Sources:
+# - https://api.ofox.ai/v1/models/x-ai/grok-4.7 (prompt $2/M, completion $6/M, cache_read $0.5/M; context 500k; max_completion 65536; supported_parameters includes reasoning)
+# - https://ofox.ai/models/x-ai/grok-4.7 (same pricing; max output 66K; reasoning + tools + prompt caching)
+# - https://docs.x.ai/developers/models/grok-4.7 (reasoning_effort low|medium|high|xhigh; cannot disable)
+# reasoning_options match first-party xAI + relay peers (no none)
+base_model = "xai/grok-4.7"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.5
+
+[limit]
+output = 65_536

From 3960b9bf712feb58b553e8045edfdfcd0a92efdb Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:57:41 -0500
Subject: [PATCH 1958/2012] fix: [missing-model] pioneer:
 fastino/gliguard-PII-multi (#7780)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../models/fastino/gliguard-PII-multi.toml    | 36 +++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 providers/pioneer/models/fastino/gliguard-PII-multi.toml

diff --git a/providers/pioneer/models/fastino/gliguard-PII-multi.toml b/providers/pioneer/models/fastino/gliguard-PII-multi.toml
new file mode 100644
index 00000000000..cec904e782a
--- /dev/null
+++ b/providers/pioneer/models/fastino/gliguard-PII-multi.toml
@@ -0,0 +1,36 @@
+# Sources:
+# - Pioneer live catalog GET https://api.pioneer.ai/v1/models (id fastino/gliguard-PII-multi: display_name, max_input_tokens 8192, max_tokens 8192, prices 0.15, capabilities all false)
+# - Pioneer base-models GET https://api.pioneer.ai/base-models (label GLiNER2-Guardrails-PII-Multi, context_window 8192, license Apache-2.0, release_month Jul 2026)
+# - Fastino release blog https://fastino.ai/blog/gliner2-guardrails-pii-multi-safety-moderation-privacy-filtering-small-language-model (released July 8 2026, 0.3B param, single forward pass, Apache 2.0)
+# - Fastino model page https://fastino.ai/models/gliner2-guardrails-pii-multi (300M-param multilingual safety moderation + PII detection, open weights)
+# - Hugging Face https://huggingface.co/fastino/GLiNER2-Guardrails-PII-Multi (Apache-2.0, 0.3B params, unified safety moderation + PII detection)
+name = "GLiNER2-Guardrails-PII-Multi"
+description = "A 300M-parameter multilingual model that runs LLM safety moderation and PII detection in a single forward pass."
+release_date = "2026-07-08"
+last_updated = "2026-07-08"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[cost]
+input = 0.15
+output = 0.15
+cache_read = 0.15
+cache_write = 0.15
+
+[limit]
+context = 8_192
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 7d4d66712d3208d832bd61f3b7b94b4ffc6eb775 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:57:50 -0500
Subject: [PATCH 1959/2012] fix: [missing-model] ofox: openai/gpt-6-luna
 (#7804)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/openai/gpt-6-luna.toml | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/ofox/models/openai/gpt-6-luna.toml

diff --git a/providers/ofox/models/openai/gpt-6-luna.toml b/providers/ofox/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..990db0bda79
--- /dev/null
+++ b/providers/ofox/models/openai/gpt-6-luna.toml
@@ -0,0 +1,20 @@
+# Sources:
+# - Ofox model page: https://ofox.ai/models/openai/gpt-6-luna
+# - Ofox catalog API: https://api.ofox.ai/v2/models/catalog?include=provider_price&search=gpt-6-luna
+# - OpenAI model docs (pricing/reasoning): https://developers.openai.com/api/docs/models/gpt-6-luna
+# Ofox provider_price override is flat $0.08/$0.40 (+ cache) with no context tier in the catalog, so no [[cost.tiers]].
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.08
+output = 0.4
+cache_read = 0.008
+cache_write = 0.1
+
+[provider]
+npm = "@ai-sdk/openai"
+api = "https://api.ofox.ai/v1"

From d53018c0bf6f9464dcfeb98eae2e948ae781a836 Mon Sep 17 00:00:00 2001
From: Yazan ALBaiz <106471410+yazan-albaiz@users.noreply.github.com>
Date: Wed, 23 Sep 2026 06:58:22 +0300
Subject: [PATCH 1960/2012] feat(amazon-bedrock): add GPT-6 Sol and Luna
 inference profiles (#7805)

---
 .../models/global.openai.gpt-6-luna.toml      | 24 +++++++++++++++++++
 .../models/global.openai.gpt-6-sol.toml       | 24 +++++++++++++++++++
 .../models/us.openai.gpt-6-luna.toml          | 24 +++++++++++++++++++
 .../models/us.openai.gpt-6-sol.toml           | 24 +++++++++++++++++++
 4 files changed, 96 insertions(+)
 create mode 100644 providers/amazon-bedrock/models/global.openai.gpt-6-luna.toml
 create mode 100644 providers/amazon-bedrock/models/global.openai.gpt-6-sol.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-6-luna.toml
 create mode 100644 providers/amazon-bedrock/models/us.openai.gpt-6-sol.toml

diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-luna.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-luna.toml
new file mode 100644
index 00000000000..d541b9a308b
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-luna.toml
@@ -0,0 +1,24 @@
+# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles (ACTIVE) on 2026-09-22; AWS has not published a GPT-6 Luna model card yet.
+# Global CRIS pricing = OpenAI API list rates: https://developers.openai.com/api/docs/models/gpt-6-luna
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max (Bedrock's own validation error lists these).
+# Live Converse check 2026-09-22 (eu-west-1, global profile): tool call + tool-result round trip; effort low/xhigh/max accepted.
+base_model = "openai/gpt-6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+name = "GPT-6 Luna (Global)"
+
+[cost]
+input = 0.10
+output = 0.50
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.20
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/global.openai.gpt-6-sol.toml b/providers/amazon-bedrock/models/global.openai.gpt-6-sol.toml
new file mode 100644
index 00000000000..fee63c0e68e
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.openai.gpt-6-sol.toml
@@ -0,0 +1,24 @@
+# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles (ACTIVE) on 2026-09-22; AWS has not published a GPT-6 Sol model card yet.
+# Global CRIS pricing = OpenAI API list rates: https://developers.openai.com/api/docs/models/gpt-6-sol
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max (Bedrock's validation error on the Luna twin lists these; matches OpenAI-native Sol).
+# Live Converse check 2026-09-22 (eu-west-1, global profile): tool call + tool-result round trip; effort low/xhigh/max accepted.
+base_model = "openai/gpt-6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+name = "GPT-6 Sol (Global)"
+
+[cost]
+input = 2.00
+output = 10.00
+cache_read = 0.20
+cache_write = 2.50
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 4.00
+output = 15.00
+cache_read = 0.40
+cache_write = 5.00
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-luna.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-luna.toml
new file mode 100644
index 00000000000..0d59417b8e6
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-luna.toml
@@ -0,0 +1,24 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles (ACTIVE) on 2026-09-22; AWS has not published a GPT-6 Luna model card yet.
+# Geo CRIS pricing = OpenAI API list rates + AWS's 10% fee (as stated on the GPT-6 Astra card): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max (Bedrock's own validation error lists these).
+# Live Converse check 2026-09-22 (eu-west-1, global profile): tool call + tool-result round trip; effort low/xhigh/max accepted.
+base_model = "openai/gpt-6-luna"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+name = "GPT-6 Luna (US)"
+
+[cost]
+input = 0.11
+output = 0.55
+cache_read = 0.011
+cache_write = 0.1375
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.22
+output = 0.825
+cache_read = 0.022
+cache_write = 0.275
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.openai.gpt-6-sol.toml b/providers/amazon-bedrock/models/us.openai.gpt-6-sol.toml
new file mode 100644
index 00000000000..9cb9cff8731
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.openai.gpt-6-sol.toml
@@ -0,0 +1,24 @@
+# US cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
+# Profile ID verified via ListInferenceProfiles (ACTIVE) on 2026-09-22; AWS has not published a GPT-6 Sol model card yet.
+# Geo CRIS pricing = OpenAI API list rates + AWS's 10% fee (as stated on the GPT-6 Astra card): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
+# Converse effort: additionalModelRequestFields.reasoning.effort = none|low|medium|high|xhigh|max (Bedrock's validation error on the Luna twin lists these; matches OpenAI-native Sol).
+# Live Converse check 2026-09-22 (eu-west-1, global profile): tool call + tool-result round trip; effort low/xhigh/max accepted.
+base_model = "openai/gpt-6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+name = "GPT-6 Sol (US)"
+
+[cost]
+input = 2.20
+output = 11.00
+cache_read = 0.22
+cache_write = 2.75
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 4.40
+output = 16.50
+cache_read = 0.44
+cache_write = 5.50
+
+[modalities]
+input = ["text", "image"]

From cc75f6ae0a9d7a5f26c79676201e1d1cb8f0e9ef Mon Sep 17 00:00:00 2001
From: David Strouk <davidstrouk@gmail.com>
Date: Wed, 23 Sep 2026 06:58:28 +0300
Subject: [PATCH 1961/2012] feat(kenari): add deepseek-v4-1-flash (#7755)

Live on https://kenari.id/v1/models as deepseek-v4-1-flash
(context 1M, input text+image, reasoning low/high/max, tool_call).
Uses base_model deepseek/deepseek-v4.1-flash, cost 0/0 per Kenari
subscription convention.

Co-authored-by: davidstrouk <david@davidstrouk.com>
---
 providers/kenari/models/deepseek-v4-1-flash.toml | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 providers/kenari/models/deepseek-v4-1-flash.toml

diff --git a/providers/kenari/models/deepseek-v4-1-flash.toml b/providers/kenari/models/deepseek-v4-1-flash.toml
new file mode 100644
index 00000000000..2a775a81e1f
--- /dev/null
+++ b/providers/kenari/models/deepseek-v4-1-flash.toml
@@ -0,0 +1,11 @@
+# Source: https://kenari.id/v1/models (id deepseek-v4-1-flash, reasoning low/high/max)
+# Kenari convention: subscription pricing, cost 0/0 like other Kenari entries.
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0
+output = 0

From 1ec5c576bb1c1a375a5ec4819c696cdb95afb6b2 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:58:45 -0500
Subject: [PATCH 1962/2012] fix: [missing-model] github-copilot:
 claude-opus-5.5 (#7815)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../github-copilot/models/claude-opus-5.5.toml      | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 providers/github-copilot/models/claude-opus-5.5.toml

diff --git a/providers/github-copilot/models/claude-opus-5.5.toml b/providers/github-copilot/models/claude-opus-5.5.toml
new file mode 100644
index 00000000000..6e3f31961a6
--- /dev/null
+++ b/providers/github-copilot/models/claude-opus-5.5.toml
@@ -0,0 +1,13 @@
+# Sources:
+# - https://github.blog/changelog/2026-09-22-claude-opus-5-5-is-now-available-in-github-copilot/
+# - https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing
+# - https://www.anthropic.com/claude-opus-5-5
+# - https://platform.claude.com/docs/en/models/opus-5-5/overview
+base_model = "anthropic/claude-opus-5-5"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5

From 80f9e2b41a4fde3e7bc39315ade9c53a75138793 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:59:18 -0500
Subject: [PATCH 1963/2012] fix: [missing-model] ofox: openai/gpt-6-sol (#7803)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 providers/ofox/models/openai/gpt-6-sol.toml | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
 create mode 100644 providers/ofox/models/openai/gpt-6-sol.toml

diff --git a/providers/ofox/models/openai/gpt-6-sol.toml b/providers/ofox/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..c686a7a19a1
--- /dev/null
+++ b/providers/ofox/models/openai/gpt-6-sol.toml
@@ -0,0 +1,20 @@
+# Sources:
+# - Ofox model page: https://ofox.ai/models/openai/gpt-6-sol
+# - Ofox catalog API: https://api.ofox.ai/v2/models/catalog?include=provider_price&search=gpt-6-sol
+# - OpenAI model docs (reasoning effort + list pricing): https://developers.openai.com/api/docs/models/gpt-6-sol
+# Ofox provider_price is the -20% promo override ($1.6/$8 + cache vs $2/$10 list); flat with no >=272k context tier in the catalog, so no [[cost.tiers]].
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1.6
+output = 8
+cache_read = 0.16
+cache_write = 2
+
+[provider]
+npm = "@ai-sdk/openai"
+api = "https://api.ofox.ai/v1"

From 3487ff2546c77c2ed30bfa677a8cbb72708aee7c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 22:59:25 -0500
Subject: [PATCH 1964/2012] chore(sync): update DigitalOcean model catalog
 (#7789)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/anthropic-claude-opus-5.5.toml     | 22 +++++++++++++++++++
 .../models/openai-gpt-6-luna.toml             | 20 +++++++++++++++++
 .../digitalocean/models/openai-gpt-6-sol.toml | 20 +++++++++++++++++
 3 files changed, 62 insertions(+)
 create mode 100644 providers/digitalocean/models/anthropic-claude-opus-5.5.toml
 create mode 100644 providers/digitalocean/models/openai-gpt-6-luna.toml
 create mode 100644 providers/digitalocean/models/openai-gpt-6-sol.toml

diff --git a/providers/digitalocean/models/anthropic-claude-opus-5.5.toml b/providers/digitalocean/models/anthropic-claude-opus-5.5.toml
new file mode 100644
index 00000000000..23c000bcf3d
--- /dev/null
+++ b/providers/digitalocean/models/anthropic-claude-opus-5.5.toml
@@ -0,0 +1,22 @@
+base_model = "anthropic/claude-opus-5-5"
+name = "Anthropic Claude Opus 5.5"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 8
+output = 30
+cache_read = 0.4
+cache_write = 10
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/digitalocean/models/openai-gpt-6-luna.toml b/providers/digitalocean/models/openai-gpt-6-luna.toml
new file mode 100644
index 00000000000..27a362b9081
--- /dev/null
+++ b/providers/digitalocean/models/openai-gpt-6-luna.toml
@@ -0,0 +1,20 @@
+base_model = "openai/gpt-6-luna"
+name = "OpenAI GPT-6 Luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/digitalocean/models/openai-gpt-6-sol.toml b/providers/digitalocean/models/openai-gpt-6-sol.toml
new file mode 100644
index 00000000000..8c434d8b143
--- /dev/null
+++ b/providers/digitalocean/models/openai-gpt-6-sol.toml
@@ -0,0 +1,20 @@
+base_model = "openai/gpt-6-sol"
+name = "OpenAI GPT-6 Sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+
+[modalities]
+input = ["text", "image"]

From 1c243c995f740d9b9c7c17608d678edb2367ef6e Mon Sep 17 00:00:00 2001
From: Yihan Yan <yanyihan@xiaomi.com>
Date: Wed, 23 Sep 2026 12:01:03 +0800
Subject: [PATCH 1965/2012] fix(xiaomi): MiMo V2.6 open weights and input
 modalities (#7719)

* fix(xiaomi): mark MiMo V2.6 Flash/Pro open-weight and drop unsupported pdf input

UltraSpeed remains closed-weight. Align input modalities to text/image/audio/video.

* docs(xiaomi): link MiMo V2.6 Flash/Pro Hugging Face weights

* fix(xiaomi): mark MiMo V2.6 Pro-UltraSpeed open-weight with Pro-RL weights

Same parameters as Pro; link the shared Hugging Face release.

* fix(xiaomi): drop incorrect MiMo V2.6 knowledge cutoff

The 2026-09-22 value was the release date, not a knowledge cutoff. Omit until a real cutoff is known.
---
 models/xiaomi/mimo-v2.6-flash.toml                    | 9 ++++++---
 models/xiaomi/mimo-v2.6-pro-ultraspeed.toml           | 8 ++++++--
 models/xiaomi/mimo-v2.6-pro.toml                      | 9 ++++++---
 providers/xiaomi/models/mimo-v2.6-flash.toml          | 5 ++---
 providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml | 4 ++--
 providers/xiaomi/models/mimo-v2.6-pro.toml            | 5 ++---
 6 files changed, 24 insertions(+), 16 deletions(-)

diff --git a/models/xiaomi/mimo-v2.6-flash.toml b/models/xiaomi/mimo-v2.6-flash.toml
index f065c84b627..2ac6276589a 100644
--- a/models/xiaomi/mimo-v2.6-flash.toml
+++ b/models/xiaomi/mimo-v2.6-flash.toml
@@ -7,13 +7,16 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2026-09-22"
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "audio", "video", "pdf"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL"
diff --git a/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml b/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
index a4ef4e05975..82d134ed28a 100644
--- a/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
+++ b/models/xiaomi/mimo-v2.6-pro-ultraspeed.toml
@@ -7,12 +7,16 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "video", "audio"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL"
diff --git a/models/xiaomi/mimo-v2.6-pro.toml b/models/xiaomi/mimo-v2.6-pro.toml
index fea580df360..fc53e943157 100644
--- a/models/xiaomi/mimo-v2.6-pro.toml
+++ b/models/xiaomi/mimo-v2.6-pro.toml
@@ -7,13 +7,16 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2026-09-22"
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "audio", "video", "pdf"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL"
diff --git a/providers/xiaomi/models/mimo-v2.6-flash.toml b/providers/xiaomi/models/mimo-v2.6-flash.toml
index 338e97b8d17..8a98ed174da 100644
--- a/providers/xiaomi/models/mimo-v2.6-flash.toml
+++ b/providers/xiaomi/models/mimo-v2.6-flash.toml
@@ -9,8 +9,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2026-09-22"
-open_weights = false
+open_weights = true
 
 [[reasoning_options]]
 type = "toggle"
@@ -25,7 +24,7 @@ context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "audio", "video", "pdf"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
 
 [interleaved]
diff --git a/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml b/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
index f1d91260627..400e80cc7c5 100644
--- a/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
+++ b/providers/xiaomi/models/mimo-v2.6-pro-ultraspeed.toml
@@ -9,7 +9,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 
 [[reasoning_options]]
 type = "toggle"
@@ -24,7 +24,7 @@ context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "video", "audio"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
 
 [interleaved]
diff --git a/providers/xiaomi/models/mimo-v2.6-pro.toml b/providers/xiaomi/models/mimo-v2.6-pro.toml
index 8debc33d06f..d3b71b59d4c 100644
--- a/providers/xiaomi/models/mimo-v2.6-pro.toml
+++ b/providers/xiaomi/models/mimo-v2.6-pro.toml
@@ -9,8 +9,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-knowledge = "2026-09-22"
-open_weights = false
+open_weights = true
 
 [[reasoning_options]]
 type = "toggle"
@@ -25,7 +24,7 @@ context = 1_048_576
 output = 131_072
 
 [modalities]
-input = ["text", "image", "audio", "video", "pdf"]
+input = ["text", "image", "audio", "video"]
 output = ["text"]
 
 [interleaved]

From 975556366c20f5980a197a7a7da84e781314dd53 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 23:01:55 -0500
Subject: [PATCH 1966/2012] chore(sync): update Cloudflare Workers AI model
 catalog (#7707)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../cloudflare-workers-ai/models/@cf/openai/gpt-oss-20b.toml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/providers/cloudflare-workers-ai/models/@cf/openai/gpt-oss-20b.toml b/providers/cloudflare-workers-ai/models/@cf/openai/gpt-oss-20b.toml
index fc4739a9fbe..0dcc22de116 100644
--- a/providers/cloudflare-workers-ai/models/@cf/openai/gpt-oss-20b.toml
+++ b/providers/cloudflare-workers-ai/models/@cf/openai/gpt-oss-20b.toml
@@ -3,7 +3,10 @@
 # https://developers.cloudflare.com/workers-ai/models/gpt-oss-20b/sync-input.json (accessed 2026-06-25)
 base_model = "openai/gpt-oss-20b"
 description = "Open-weight GPT model for self-hosted reasoning and instruction-following workloads"
-reasoning_options = []
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
 
 [cost]
 input = 0.2

From 83ca66bff6a341b30720836510d4b905b8bc766c Mon Sep 17 00:00:00 2001
From: Abdul Munim <iam@munim.net>
Date: Wed, 23 Sep 2026 06:02:01 +0200
Subject: [PATCH 1967/2012] cline-pass: add missing mimo-v2.6-flash,
 mimo-v2.6-pro, muse-spark-1.3-contributor (#7729)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(cline-pass): add mimo-v2.6 and muse-spark-1.3 models

Live clinePass catalog lists 14 models; three slugs were absent from
providers/cline-pass/models/cline-pass/. All three serve requests on
api.cline.bot (probe-verified 2026-09-22); nonexistent slugs 404.
muse-spark-1.3-contributor carries Meta reference rates because Cline
publishes no ClinePass rate for it.

* fix(cline-pass): follow review — wire-cited effort, drop muse cost

mimo-v2.6-* reasoning_options now list exactly the seven levels the
ClinePass endpoint accepts (probe 2026-09-22: seven levels -> 200,
unknown level -> 500), cited in a leading wire comment.
muse-spark-1.3-contributor drops [cost]: Cline publishes no
ClinePass rate for the slug.

* fix(cline-pass): revert mimo effort to same-host peer set

Effect probe (reasoning_tokens, 3 trials/level) shows minimal/max
do not differentiate on this host, so declare the mimo-v2.5* peer
set per review; wire comment keeps the full acceptance evidence.
---
 .../models/cline-pass/mimo-v2.6-flash.toml       | 16 ++++++++++++++++
 .../models/cline-pass/mimo-v2.6-pro.toml         | 16 ++++++++++++++++
 .../cline-pass/muse-spark-1.3-contributor.toml   |  8 ++++++++
 3 files changed, 40 insertions(+)
 create mode 100644 providers/cline-pass/models/cline-pass/mimo-v2.6-flash.toml
 create mode 100644 providers/cline-pass/models/cline-pass/mimo-v2.6-pro.toml
 create mode 100644 providers/cline-pass/models/cline-pass/muse-spark-1.3-contributor.toml

diff --git a/providers/cline-pass/models/cline-pass/mimo-v2.6-flash.toml b/providers/cline-pass/models/cline-pass/mimo-v2.6-flash.toml
new file mode 100644
index 00000000000..612d086a7f4
--- /dev/null
+++ b/providers/cline-pass/models/cline-pass/mimo-v2.6-flash.toml
@@ -0,0 +1,16 @@
+# Model ID cline-pass/mimo-v2.6-flash; reference pricing from the Cline model
+# catalog https://api.cline.bot/api/v1/ai/cline/models (accessed 2026-09-22):
+# $0.14 / $0.28 / $0.0028 per 1M tokens (input/output/cache read).
+# Wire evidence (2026-09-22): POST /api/v1/chat/completions accepts
+# reasoning_effort = none|low|medium|high|xhigh|minimal|max (HTTP 200) and
+# rejects unknown levels (HTTP 500). An effect probe (3 trials per level,
+# reasoning_tokens per response) shows no measurable difference for
+# minimal/max, so this file declares the established same-host peer set
+# used by cline-pass/mimo-v2.5* (none, low, medium, high, xhigh).
+base_model = "xiaomi/mimo-v2.6-flash"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
diff --git a/providers/cline-pass/models/cline-pass/mimo-v2.6-pro.toml b/providers/cline-pass/models/cline-pass/mimo-v2.6-pro.toml
new file mode 100644
index 00000000000..d42aa0606d7
--- /dev/null
+++ b/providers/cline-pass/models/cline-pass/mimo-v2.6-pro.toml
@@ -0,0 +1,16 @@
+# Model ID cline-pass/mimo-v2.6-pro; reference pricing from the Cline model
+# catalog https://api.cline.bot/api/v1/ai/cline/models (accessed 2026-09-22):
+# $0.435 / $0.87 / $0.0036 per 1M tokens (input/output/cache read).
+# Wire evidence (2026-09-22): POST /api/v1/chat/completions accepts
+# reasoning_effort = none|low|medium|high|xhigh|minimal|max (HTTP 200) and
+# rejects unknown levels (HTTP 500). An effect probe (reasoning_tokens per
+# response) shows no measurable difference for minimal/max, so this file
+# declares the established same-host peer set used by cline-pass/mimo-v2.5-pro
+# (none, low, medium, high, xhigh).
+base_model = "xiaomi/mimo-v2.6-pro"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.0036
diff --git a/providers/cline-pass/models/cline-pass/muse-spark-1.3-contributor.toml b/providers/cline-pass/models/cline-pass/muse-spark-1.3-contributor.toml
new file mode 100644
index 00000000000..a5f354380fd
--- /dev/null
+++ b/providers/cline-pass/models/cline-pass/muse-spark-1.3-contributor.toml
@@ -0,0 +1,8 @@
+# Model ID cline-pass/muse-spark-1.3-contributor. Cline publishes no
+# ClinePass reference rate for this model: docs.cline.bot/getting-started/
+# clinepass has no row for it (accessed 2026-09-22) and the model is absent
+# from api.cline.bot/api/v1/ai/cline/models, so no [cost] block is declared.
+base_model = "meta/muse-spark-1.3"
+name = "Muse Spark 1.3 Contributor"
+
+reasoning_options = [{ type = "effort", values = ["minimal", "low", "medium", "high", "xhigh"] }]

From 552f3b99c6595cd6dc579be2a19e5faef9be5e57 Mon Sep 17 00:00:00 2001
From: Tianning Li <51225681+lit26@users.noreply.github.com>
Date: Wed, 23 Sep 2026 12:03:44 +0800
Subject: [PATCH 1968/2012] feat(stepfun): add step-5-preview to StepFun
 providers (#7535)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Adds the Step 5 Preview lab entry and serves it from all four StepFun
hosts: StepFun (China), StepFun (Global), and both Step Plan channels.

Specs from the official model page: 1M context / input / output, text +
image + video input, text output, tool calling, JSON Mode + JSON Schema,
prompt caching, and low/medium/high reasoning effort via Chat
`reasoning_effort`, Messages `output_config.effort`, and Responses
`reasoning.effort`.

Pricing: China converts the CNY list price (7 / 20 / 0.35 per 1M) at the
same 7.2973 CNY/USD rate the other StepFun entries use; Global uses the
published USD price (1.00 / 2.70 / 0.05) directly, so that entry is a
real file rather than a symlink into providers/stepfun.

Step Plan entries carry no cost block, matching their siblings — that
channel bills against plan credit, not per token.

Provider comments also updated: step-5-preview added to the effort
matrix, and the Responses API now serves step-5-preview alongside
step-3.7-flash.

Sources:
https://platform.stepfun.com/docs/zh/guides/models/step-5-preview
https://platform.stepfun.ai/docs/en/guides/models/step-5-preview
https://platform.stepfun.com/docs/zh/guides/pricing/details
https://platform.stepfun.ai/docs/en/guides/pricing/details
https://platform.stepfun.com/docs/zh/step-plan/integrations/reasoning-api

Co-authored-by: Tianning Li <litianning@stepfun.com>
---
 .../models/step-5-preview.toml                | 11 ++++++++++
 providers/stepfun-ai-step-plan/provider.toml  |  8 ++++----
 providers/stepfun/models/step-5-preview.toml  | 20 +++++++++++++++++++
 providers/stepfun/provider.toml               |  7 ++++---
 4 files changed, 39 insertions(+), 7 deletions(-)
 create mode 100644 providers/stepfun-ai-step-plan/models/step-5-preview.toml
 create mode 100644 providers/stepfun/models/step-5-preview.toml

diff --git a/providers/stepfun-ai-step-plan/models/step-5-preview.toml b/providers/stepfun-ai-step-plan/models/step-5-preview.toml
new file mode 100644
index 00000000000..674a616d307
--- /dev/null
+++ b/providers/stepfun-ai-step-plan/models/step-5-preview.toml
@@ -0,0 +1,11 @@
+# Chat `reasoning_effort` and Messages `output_config.effort` accept
+# low/medium/high (accessed 2026-09-20).
+# https://platform.stepfun.ai/docs/en/step-plan/integrations/reasoning-api
+base_model = "stepfun/step-5-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[interleaved]
+field = "reasoning_content"
diff --git a/providers/stepfun-ai-step-plan/provider.toml b/providers/stepfun-ai-step-plan/provider.toml
index 15e9fb3b704..f357f0a530f 100644
--- a/providers/stepfun-ai-step-plan/provider.toml
+++ b/providers/stepfun-ai-step-plan/provider.toml
@@ -1,12 +1,12 @@
 name = "StepFun Step Plan (Global)"
 env = ["STEPFUN_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
-# Reasoning HTTP format (accessed 2026-06-25):
+# Reasoning HTTP format (accessed 2026-09-20):
 # Step Plan exposes POST /step_plan/v1/chat/completions with top-level
 # `reasoning_effort` and POST /step_plan/v1/messages with
-# `output_config.effort`. step-3.7-flash accepts low/medium/high;
-# step-3.5-flash and step-3.5-flash-2603 accept low/high. No plan Responses
-# endpoint is listed.
+# `output_config.effort`. step-5-preview and step-3.7-flash accept
+# low/medium/high; step-3.5-flash and step-3.5-flash-2603 accept low/high.
+# No plan Responses endpoint is listed.
 # Source:
 # https://platform.stepfun.ai/docs/en/step-plan/integrations/reasoning-api
 doc = "https://platform.stepfun.ai/docs/en/step-plan/integrations/reasoning-api"
diff --git a/providers/stepfun/models/step-5-preview.toml b/providers/stepfun/models/step-5-preview.toml
new file mode 100644
index 00000000000..94aa0f9f5f1
--- /dev/null
+++ b/providers/stepfun/models/step-5-preview.toml
@@ -0,0 +1,20 @@
+# Chat `reasoning_effort`, Messages `output_config.effort`, and Responses
+# `reasoning.effort` accept low/medium/high (accessed 2026-09-20).
+# https://platform.stepfun.com/docs/zh/guides/models/step-5-preview
+# Cost converted from CNY list price (¥7 / ¥20 / ¥0.35 per 1M tokens) at the
+# same 7.2973 CNY/USD rate used by the other stepfun entries (accessed
+# 2026-09-20).
+# https://platform.stepfun.com/docs/zh/guides/pricing/details
+base_model = "stepfun/step-5-preview"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.959
+output = 2.741
+cache_read = 0.048
diff --git a/providers/stepfun/provider.toml b/providers/stepfun/provider.toml
index 0de8764f300..b687e0750f1 100644
--- a/providers/stepfun/provider.toml
+++ b/providers/stepfun/provider.toml
@@ -1,11 +1,12 @@
 name = "StepFun (China)"
 env = ["STEPFUN_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
-# Reasoning HTTP format (accessed 2026-06-25):
+# Reasoning HTTP format (accessed 2026-09-20):
 # POST /v1/chat/completions uses top-level `reasoning_effort`; POST /v1/messages
 # uses `output_config.effort`; POST /v1/responses uses `reasoning.effort`.
-# Values are low/medium/high for step-3.7-flash; step-3.5-flash-2603 accepts
-# low/high. Responses supports only step-3.7-flash. Chat returns reasoning at
+# Values are low/medium/high for step-5-preview and step-3.7-flash;
+# step-3.5-flash-2603 accepts low/high. Responses supports only
+# step-5-preview and step-3.7-flash. Chat returns reasoning at
 # `choices[].message.reasoning` or streamed `choices[].delta.reasoning`;
 # `reasoning_format` is general (default) or deepseek-style, the latter using
 # `reasoning_content`. Responses streams `response.reasoning_text.delta`.

From 1222d431c02455357c5b2cd31e613e41e0f06a9f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 23:09:32 -0500
Subject: [PATCH 1969/2012] fix: add supported Kimi K3 Bedrock profiles (#7795)

Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>
---
 .../models/global.moonshotai.kimi-k3.toml     | 22 +++++++++++++++++++
 .../models/us.moonshotai.kimi-k3.toml         | 22 +++++++++++++++++++
 2 files changed, 44 insertions(+)
 create mode 100644 providers/amazon-bedrock/models/global.moonshotai.kimi-k3.toml
 create mode 100644 providers/amazon-bedrock/models/us.moonshotai.kimi-k3.toml

diff --git a/providers/amazon-bedrock/models/global.moonshotai.kimi-k3.toml b/providers/amazon-bedrock/models/global.moonshotai.kimi-k3.toml
new file mode 100644
index 00000000000..05556e9c377
--- /dev/null
+++ b/providers/amazon-bedrock/models/global.moonshotai.kimi-k3.toml
@@ -0,0 +1,22 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k3.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html
+# Global CRIS standard pricing; cache_write is the 30-minute rate.
+# Reasoning is always on; no Bedrock Converse effort control is documented or verified.
+# Live Converse 2026-09-23: reasoning_effort accepted invalid strings, numbers,
+# and objects exactly like an unrelated unknown field; low/max probes showed no
+# verified effect. The bare model ID was rejected; an inference profile is required.
+# Global and US profiles returned reasoningContent.
+# Bedrock supports text and image input, but not the base model's video input.
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (Global)"
+reasoning_options = []
+interleaved = true
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/amazon-bedrock/models/us.moonshotai.kimi-k3.toml b/providers/amazon-bedrock/models/us.moonshotai.kimi-k3.toml
new file mode 100644
index 00000000000..9c724cee821
--- /dev/null
+++ b/providers/amazon-bedrock/models/us.moonshotai.kimi-k3.toml
@@ -0,0 +1,22 @@
+# Sources: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k3.html
+# https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html
+# US CRIS standard pricing; cache_write is the 30-minute rate.
+# Reasoning is always on; no Bedrock Converse effort control is documented or verified.
+# Live Converse 2026-09-23: reasoning_effort accepted invalid strings, numbers,
+# and objects exactly like an unrelated unknown field; low/max probes showed no
+# verified effect. The bare model ID was rejected; an inference profile is required.
+# Global and US profiles returned reasoningContent.
+# Bedrock supports text and image input, but not the base model's video input.
+base_model = "moonshotai/kimi-k3"
+name = "Kimi K3 (US)"
+reasoning_options = []
+interleaved = true
+
+[cost]
+input = 3.30
+output = 16.50
+cache_read = 0.33
+cache_write = 4.125
+
+[modalities]
+input = ["text", "image"]

From 07fb48aae70f62ae6d63b7a7121347aa88171380 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Tue, 22 Sep 2026 23:13:28 -0500
Subject: [PATCH 1970/2012] fix: ClinePass thinking-effort levels are wrong for
 all models: registry shows none/low/medium/high/xhigh, but DeepSeek V4.1
 Flash officially supports off/low/high/max (#7534)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 .../cline-pass/models/cline-pass/deepseek-v4-flash.toml   | 6 +++++-
 .../cline-pass/models/cline-pass/deepseek-v4-pro.toml     | 6 +++++-
 .../cline-pass/models/cline-pass/deepseek-v4.1-flash.toml | 6 ++++--
 providers/cline-pass/models/cline-pass/glm-5.2.toml       | 6 +++++-
 providers/cline-pass/models/cline-pass/kimi-k2.6.toml     | 5 ++++-
 .../cline-pass/models/cline-pass/kimi-k2.7-code.toml      | 5 ++++-
 providers/cline-pass/models/cline-pass/kimi-k3.toml       | 4 +++-
 providers/cline-pass/models/cline-pass/mimo-v2.5-pro.toml | 5 ++++-
 providers/cline-pass/models/cline-pass/mimo-v2.5.toml     | 5 ++++-
 providers/cline-pass/models/cline-pass/minimax-m3.toml    | 5 ++++-
 providers/cline-pass/models/cline-pass/qwen3.7-max.toml   | 6 +++++-
 providers/cline-pass/models/cline-pass/qwen3.7-plus.toml  | 6 +++++-
 providers/cline-pass/models/cline-pass/qwen3.8-max.toml   | 8 ++++----
 13 files changed, 56 insertions(+), 17 deletions(-)

diff --git a/providers/cline-pass/models/cline-pass/deepseek-v4-flash.toml b/providers/cline-pass/models/cline-pass/deepseek-v4-flash.toml
index a402b4d533b..044e118d9af 100644
--- a/providers/cline-pass/models/cline-pass/deepseek-v4-flash.toml
+++ b/providers/cline-pass/models/cline-pass/deepseek-v4-flash.toml
@@ -1,5 +1,9 @@
+# https://docs.cline.bot/getting-started/clinepass
+# https://api-docs.deepseek.com/guides/thinking_mode/ (OpenAI: thinking.type + reasoning_effort low|high|max)
+# Effort: reasoning_effort = none|low|high|max (none disables thinking; DeepSeek Flash graded set)
+# Legacy Flash slug; lab serves V4.1-Flash. Same reasoning surface as deepseek-v4.1-flash.
 base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
 
 [cost]
 input = 0.14
diff --git a/providers/cline-pass/models/cline-pass/deepseek-v4-pro.toml b/providers/cline-pass/models/cline-pass/deepseek-v4-pro.toml
index 3fb0bd31762..191909977a6 100644
--- a/providers/cline-pass/models/cline-pass/deepseek-v4-pro.toml
+++ b/providers/cline-pass/models/cline-pass/deepseek-v4-pro.toml
@@ -1,5 +1,9 @@
+# https://docs.cline.bot/getting-started/clinepass
+# https://api-docs.deepseek.com/guides/thinking_mode/ (OpenAI: thinking.type + reasoning_effort)
+# First-party DeepSeek Pro: effort high|max (lab providers/deepseek/models/deepseek-v4-pro.toml)
+# Effort: reasoning_effort = none|high|max (none disables thinking)
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
 
 [cost]
 input = 1.74
diff --git a/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml b/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml
index cdc87207c83..453a58da360 100644
--- a/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml
+++ b/providers/cline-pass/models/cline-pass/deepseek-v4.1-flash.toml
@@ -1,9 +1,11 @@
 # https://api.cline.bot/api/v1/ai/cline/recommended-models (clinePass id cline-pass/deepseek-v4.1-flash, accessed 2026-09-11)
 # https://docs.cline.bot/getting-started/clinepass (ClinePass reference pricing cites DeepSeek API pricing for Flash quota)
 # https://api-docs.deepseek.com/quick_start/pricing (DeepSeek-V4.1-Flash off-peak USD/MTok; ClinePass docs table still lists deepseek-v4-flash only)
-# Reasoning effort values match other ClinePass DeepSeek entries on this host.
+# https://api-docs.deepseek.com/guides/thinking_mode/ (OpenAI: thinking.type + reasoning_effort low|high|max; medium/xhigh map to high)
+# Effort: reasoning_effort = none|low|high|max (none disables thinking; matches DeepSeek Flash graded set + off)
+# ClinePass is an OpenAI-compatible relay; do not advertise GPT-style medium/xhigh.
 base_model = "deepseek/deepseek-v4.1-flash"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
 
 [cost]
 input = 0.15
diff --git a/providers/cline-pass/models/cline-pass/glm-5.2.toml b/providers/cline-pass/models/cline-pass/glm-5.2.toml
index b3cac6301a6..7458eed5caa 100644
--- a/providers/cline-pass/models/cline-pass/glm-5.2.toml
+++ b/providers/cline-pass/models/cline-pass/glm-5.2.toml
@@ -1,5 +1,9 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Zhipu GLM-5.2: effective effort high|max; none|minimal skip thinking
+# (providers/zhipuai/models/glm-5.2.toml; https://docs.bigmodel.cn/cn/guide/capabilities/thinking)
+# Effort: reasoning_effort = none|high|max
 base_model = "zhipuai/glm-5.2"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
 
 [cost]
 input = 1.4
diff --git a/providers/cline-pass/models/cline-pass/kimi-k2.6.toml b/providers/cline-pass/models/cline-pass/kimi-k2.6.toml
index 373fb552c69..fc3fb04b0db 100644
--- a/providers/cline-pass/models/cline-pass/kimi-k2.6.toml
+++ b/providers/cline-pass/models/cline-pass/kimi-k2.6.toml
@@ -1,5 +1,8 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Moonshot K2.6: thinking on/off only (providers/moonshotai/models/kimi-k2.6.toml)
+# Toggle: thinking / reasoning enabled|disabled (OpenAI-compatible relay; no graded effort on lab)
 base_model = "moonshotai/kimi-k2.6"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.95
diff --git a/providers/cline-pass/models/cline-pass/kimi-k2.7-code.toml b/providers/cline-pass/models/cline-pass/kimi-k2.7-code.toml
index 7340bf79c7a..9fc9a2f9a5a 100644
--- a/providers/cline-pass/models/cline-pass/kimi-k2.7-code.toml
+++ b/providers/cline-pass/models/cline-pass/kimi-k2.7-code.toml
@@ -1,5 +1,8 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Moonshot K2.7 Code: always-on reasoning, no caller control
+# (providers/moonshotai/models/kimi-k2.7-code.toml; openrouter peer also [])
 base_model = "moonshotai/kimi-k2.7-code"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = []
 
 [cost]
 input = 0.95
diff --git a/providers/cline-pass/models/cline-pass/kimi-k3.toml b/providers/cline-pass/models/cline-pass/kimi-k3.toml
index 856a9963ae3..2b9dee1d87c 100644
--- a/providers/cline-pass/models/cline-pass/kimi-k3.toml
+++ b/providers/cline-pass/models/cline-pass/kimi-k3.toml
@@ -1,7 +1,9 @@
 # https://docs.cline.bot/getting-started/clinepass (accessed 2026-07-22)
 # Model ID cline-pass/kimi-k3; reference pricing $3 / $15 / $0.30 per 1M tokens (input/output/cache read).
+# Lab Moonshot K3: thinking.type + output_config.effort low|high|max (providers/moonshotai/models/kimi-k3.toml)
+# Effort: reasoning_effort = none|low|high|max (none disables; no medium/xhigh)
 base_model = "moonshotai/kimi-k3"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
 
 [cost]
 input = 3
diff --git a/providers/cline-pass/models/cline-pass/mimo-v2.5-pro.toml b/providers/cline-pass/models/cline-pass/mimo-v2.5-pro.toml
index 4b15e595d0b..8ff3d62b6dc 100644
--- a/providers/cline-pass/models/cline-pass/mimo-v2.5-pro.toml
+++ b/providers/cline-pass/models/cline-pass/mimo-v2.5-pro.toml
@@ -1,5 +1,8 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Xiaomi MiMo-V2.5-Pro: thinking on/off only (providers/xiaomi/models/mimo-v2.5-pro.toml)
+# Toggle: thinking / reasoning enabled|disabled (no graded effort on lab)
 base_model = "xiaomi/mimo-v2.5-pro"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 1.74
diff --git a/providers/cline-pass/models/cline-pass/mimo-v2.5.toml b/providers/cline-pass/models/cline-pass/mimo-v2.5.toml
index a05c72ca039..3a9f90ed0bc 100644
--- a/providers/cline-pass/models/cline-pass/mimo-v2.5.toml
+++ b/providers/cline-pass/models/cline-pass/mimo-v2.5.toml
@@ -1,5 +1,8 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Xiaomi MiMo-V2.5: thinking on/off only (providers/xiaomi/models/mimo-v2.5.toml)
+# Toggle: thinking / reasoning enabled|disabled (no graded effort on lab)
 base_model = "xiaomi/mimo-v2.5"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.14
diff --git a/providers/cline-pass/models/cline-pass/minimax-m3.toml b/providers/cline-pass/models/cline-pass/minimax-m3.toml
index 1367e07bc0e..599f9fc9ec1 100644
--- a/providers/cline-pass/models/cline-pass/minimax-m3.toml
+++ b/providers/cline-pass/models/cline-pass/minimax-m3.toml
@@ -1,5 +1,8 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab MiniMax-M3: thinking on/off only (providers/minimax/models/MiniMax-M3.toml)
+# Toggle: thinking / reasoning enabled|disabled (no graded effort on lab)
 base_model = "minimax/MiniMax-M3"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.3
diff --git a/providers/cline-pass/models/cline-pass/qwen3.7-max.toml b/providers/cline-pass/models/cline-pass/qwen3.7-max.toml
index 7ec1495bb20..302e667ede8 100644
--- a/providers/cline-pass/models/cline-pass/qwen3.7-max.toml
+++ b/providers/cline-pass/models/cline-pass/qwen3.7-max.toml
@@ -1,5 +1,9 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Alibaba Qwen3.7 Max: enable_thinking + thinking_budget (providers/alibaba/models/qwen3.7-max.toml)
+# ClinePass OpenAI-compatible Chat Completions path: peers (OpenRouter) expose on/off only, not budget_tokens.
+# Toggle: enable_thinking / reasoning enabled|disabled
 base_model = "alibaba/qwen3.7-max"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 2.5
diff --git a/providers/cline-pass/models/cline-pass/qwen3.7-plus.toml b/providers/cline-pass/models/cline-pass/qwen3.7-plus.toml
index 55e28b335eb..4c0d4b2c041 100644
--- a/providers/cline-pass/models/cline-pass/qwen3.7-plus.toml
+++ b/providers/cline-pass/models/cline-pass/qwen3.7-plus.toml
@@ -1,5 +1,9 @@
+# https://docs.cline.bot/getting-started/clinepass
+# Lab Alibaba Qwen3.7 Plus: enable_thinking + thinking_budget (providers/alibaba/models/qwen3.7-plus.toml)
+# ClinePass OpenAI-compatible Chat Completions path: peers (OpenRouter) expose on/off only, not budget_tokens.
+# Toggle: enable_thinking / reasoning enabled|disabled
 base_model = "alibaba/qwen3.7-plus"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.4
diff --git a/providers/cline-pass/models/cline-pass/qwen3.8-max.toml b/providers/cline-pass/models/cline-pass/qwen3.8-max.toml
index a72991d846b..6a9d910be85 100644
--- a/providers/cline-pass/models/cline-pass/qwen3.8-max.toml
+++ b/providers/cline-pass/models/cline-pass/qwen3.8-max.toml
@@ -1,11 +1,11 @@
 # https://docs.cline.bot/getting-started/clinepass (accessed 2026-08-24)
 # Model ID cline-pass/qwen3.8-max; reference pricing $2 / $6 / $0.25 per 1M tokens (input/output/cache read), cache write $2.5.
+# Lab Alibaba Qwen3.8 Max: enable_thinking + reasoning_effort low|medium|xhigh
+# (providers/alibaba/models/qwen3.8-max.toml; https://docs.qwencloud.com/developer-guides/text-generation/thinking)
+# Effort includes none for off (hybrid model); no high (lab maps high→xhigh; native graded set is low|medium|xhigh)
 base_model = "alibaba/qwen3.8-max"
 structured_output = true
-
-[[reasoning_options]]
-type = "effort"
-values = ["minimal", "low", "medium", "high", "xhigh"]
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
 
 [cost]
 input = 2

From 76689e21b85fc1b87fe6fbcac0a5d4aeb61b0c23 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 04:31:20 +0000
Subject: [PATCH 1971/2012] chore(sync): update NanoGPT model catalog (#7839)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml       | 3 ---
 providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml         | 3 ---
 3 files changed, 3 insertions(+), 9 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
index bf3082ff91f..79ac140f7b6 100644
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.6-flash.toml
@@ -14,6 +14,3 @@ cache_write = 0
 
 [limit]
 input = 1_048_576
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml
index ea8e65a904c..a99c234165d 100644
--- a/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml
+++ b/providers/nano-gpt/models/xiaomi/mimo-v2.6-pro.toml
@@ -14,6 +14,3 @@ cache_write = 0
 
 [limit]
 input = 1_048_576
-
-[modalities]
-input = ["text", "image", "video", "audio"]

From 7d4b261ad404a5211432478df74e9e1a01a917e4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 04:31:24 +0000
Subject: [PATCH 1972/2012] chore(sync): update OpenRouter model catalog
 (#7840)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml     | 3 ---
 providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml       | 3 ---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 5 files changed, 9 insertions(+), 15 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..7c5828b16e2 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.66
+output = 1.98
+cache_read = 0.022
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index ee75b6ad088..c689935f095 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.5
-cache_read = 0.01
+input = 0.094
+output = 0.6
+cache_read = 0.05
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
index 3e52e783fa7..e1bb054f013 100644
--- a/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-flash.toml
@@ -11,6 +11,3 @@ type = "toggle"
 input = 0.14
 output = 0.28
 cache_read = 0.0028
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml b/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml
index 2f40e81f221..7f938318372 100644
--- a/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml
+++ b/providers/openrouter/models/xiaomi/mimo-v2.6-pro.toml
@@ -11,6 +11,3 @@ type = "toggle"
 input = 0.435
 output = 0.87
 cache_read = 0.0036
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index a3bff918037..2ca0c23e6a3 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.5
-cache_read = 0.01
+input = 0.094
+output = 0.6
+cache_read = 0.05
 
 [limit]
 context = 1_048_576

From e4eb305bc0faa591cbf275f65001629e85e0211a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 04:31:32 +0000
Subject: [PATCH 1973/2012] chore(sync): update EmpirioLabs AI model catalog
 (#7837)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/empiriolabs/models/mimo-v2-6-flash.toml | 3 ---
 providers/empiriolabs/models/mimo-v2-6-pro.toml   | 3 ---
 2 files changed, 6 deletions(-)

diff --git a/providers/empiriolabs/models/mimo-v2-6-flash.toml b/providers/empiriolabs/models/mimo-v2-6-flash.toml
index ab796f4d7ef..14d640b4db5 100644
--- a/providers/empiriolabs/models/mimo-v2-6-flash.toml
+++ b/providers/empiriolabs/models/mimo-v2-6-flash.toml
@@ -12,6 +12,3 @@ cache_read = 0.7
 
 [limit]
 context = 1_000_000
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/empiriolabs/models/mimo-v2-6-pro.toml b/providers/empiriolabs/models/mimo-v2-6-pro.toml
index 04dcc7ceb0e..54b2f7069c3 100644
--- a/providers/empiriolabs/models/mimo-v2-6-pro.toml
+++ b/providers/empiriolabs/models/mimo-v2-6-pro.toml
@@ -12,6 +12,3 @@ cache_read = 2.175
 
 [limit]
 context = 1_000_000
-
-[modalities]
-input = ["text", "image", "video", "audio"]

From 59eda218803092a288b792a51e5e1ce081480f3c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 04:31:35 +0000
Subject: [PATCH 1974/2012] chore(sync): update Kilo model catalog (#7838)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/xiaomi/mimo-v2.6-flash.toml          | 3 ---
 providers/kilo/models/xiaomi/mimo-v2.6-pro.toml            | 3 ---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 3 files changed, 3 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml b/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
index b4165fb24ad..0a1c6e4fe63 100644
--- a/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-flash.toml
@@ -10,6 +10,3 @@ values = ["none", "high"]
 input = 0.14
 output = 0.28
 cache_read = 0.0028
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml b/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml
index e3e0a47689b..1285f109afd 100644
--- a/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml
+++ b/providers/kilo/models/xiaomi/mimo-v2.6-pro.toml
@@ -10,6 +10,3 @@ values = ["none", "high"]
 input = 0.435
 output = 0.87
 cache_read = 0.0036
-
-[modalities]
-input = ["text", "image", "video", "audio"]
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 0295282190d..48b2c256919 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.5
-cache_read = 0.01
+input = 0.094
+output = 0.6
+cache_read = 0.05
 
 [limit]
 context = 1_048_576

From 7d87c4fc34b195d8d5fd449903054f666fbb3d24 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 04:31:48 +0000
Subject: [PATCH 1975/2012] chore(sync): update CrossModel model catalog
 (#7841)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml | 3 ---
 providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml   | 3 ---
 2 files changed, 6 deletions(-)

diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml b/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
index 51fd77c6aa8..5fdd8d29025 100644
--- a/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.6-flash.toml
@@ -9,6 +9,3 @@ input = 0.16
 output = 0.32
 cache_read = 0.004
 cache_write = 0.16
-
-[modalities]
-input = ["text", "image", "audio", "video"]
diff --git a/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml b/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml
index e0fca43e997..1dbc817d0fa 100644
--- a/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml
+++ b/providers/crossmodel/models/xiaomi/mimo-v2.6-pro.toml
@@ -9,6 +9,3 @@ input = 0.47
 output = 0.94
 cache_read = 0.005
 cache_write = 0.47
-
-[modalities]
-input = ["text", "image", "audio", "video"]

From 1c94dc631474bfb5c37251758f2a280ae3a3a91a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 05:27:40 +0000
Subject: [PATCH 1976/2012] chore(sync): update Kilo model catalog (#7845)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml          | 1 +
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 4 ++--
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index 39ad2aa079f..c79bdfce589 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -16,3 +16,4 @@ cache_write = 2.5
 
 [limit]
 context = 1_000_000
+output = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 48b2c256919..8923039f0f4 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.094
+input = 0.079
 output = 0.6
-cache_read = 0.05
+cache_read = 0.06
 
 [limit]
 context = 1_048_576

From b3318a67977cea8231db5fa772f1ba1900ded7f9 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 05:27:44 +0000
Subject: [PATCH 1977/2012] chore(sync): update Eden AI model catalog (#7847)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/amazon/moonshotai.kimi-k2.5.toml    |  1 -
 .../models/anthropic/claude-opus-5-5.toml      | 12 ++++++++++++
 .../databricks-deepseek-v4-flash-0731.toml     |  2 +-
 .../databricks-deepseek-v4-pro-0813.toml       |  6 +++---
 .../models/databricks/databricks-inkling.toml  |  6 +++---
 providers/edenai/models/openai/gpt-6-luna.toml | 18 ++++++++++++++++++
 providers/edenai/models/openai/gpt-6-sol.toml  | 18 ++++++++++++++++++
 7 files changed, 55 insertions(+), 8 deletions(-)
 create mode 100644 providers/edenai/models/anthropic/claude-opus-5-5.toml
 create mode 100644 providers/edenai/models/openai/gpt-6-luna.toml
 create mode 100644 providers/edenai/models/openai/gpt-6-sol.toml

diff --git a/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml b/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
index 45d5d0a17cb..c64d911d778 100644
--- a/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
+++ b/providers/edenai/models/amazon/moonshotai.kimi-k2.5.toml
@@ -1,6 +1,5 @@
 base_model = "moonshotai/kimi-k2.5"
 name = "Kimi K2.5 (Amazon Bedrock)"
-structured_output = false
 reasoning_options = []
 
 [cost]
diff --git a/providers/edenai/models/anthropic/claude-opus-5-5.toml b/providers/edenai/models/anthropic/claude-opus-5-5.toml
new file mode 100644
index 00000000000..8d8ff11a7d2
--- /dev/null
+++ b/providers/edenai/models/anthropic/claude-opus-5-5.toml
@@ -0,0 +1,12 @@
+base_model = "anthropic/claude-opus-5-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 4
+output = 20
+cache_read = 0.2
+cache_write = 5
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
index d4d7d35c658..ec6f7cacc88 100644
--- a/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-flash-0731.toml
@@ -9,5 +9,5 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.14
 output = 0.28
-cache_read = 0.028
+cache_read = 0.014
 cache_write = 0.14
diff --git a/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
index a25d6b3e95b..a293c8ac08b 100644
--- a/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
+++ b/providers/edenai/models/databricks/databricks-deepseek-v4-pro-0813.toml
@@ -7,7 +7,7 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.31999
-output = 3.95997
-cache_read = 0.13202
+input = 1.32
+output = 3.959999
+cache_read = 0.132
 cache_write = 1.31999
diff --git a/providers/edenai/models/databricks/databricks-inkling.toml b/providers/edenai/models/databricks/databricks-inkling.toml
index 177f6bd3f2d..eef08eb01d0 100644
--- a/providers/edenai/models/databricks/databricks-inkling.toml
+++ b/providers/edenai/models/databricks/databricks-inkling.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "max"]
 
 [cost]
-input = 1.00002
-output = 4.04999
-cache_read = 0.17003
+input = 1
+output = 4.05
+cache_read = 0.1
 cache_write = 1.00002
 
 [limit]
diff --git a/providers/edenai/models/openai/gpt-6-luna.toml b/providers/edenai/models/openai/gpt-6-luna.toml
new file mode 100644
index 00000000000..9a9537cbf02
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-6-luna.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-luna"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.5
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 0.2
+output = 0.75
+cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/edenai/models/openai/gpt-6-sol.toml b/providers/edenai/models/openai/gpt-6-sol.toml
new file mode 100644
index 00000000000..94f694d64de
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-6-sol.toml
@@ -0,0 +1,18 @@
+base_model = "openai/gpt-6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5

From d149193b3a0397cdb37b4105a26a543ad79ea2a7 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 05:27:46 +0000
Subject: [PATCH 1978/2012] chore(sync): update OpenRouter model catalog
 (#7846)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml | 4 ++--
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml       | 1 +
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml    | 4 ++--
 3 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index c689935f095..e0411f29de8 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.094
+input = 0.079
 output = 0.6
-cache_read = 0.05
+cache_read = 0.06
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index 74b5d1276f6..abacdd68c88 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -18,3 +18,4 @@ cache_read = 0.25
 
 [limit]
 context = 1_048_576
+output = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 2ca0c23e6a3..aeddf9446b1 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.094
+input = 0.079
 output = 0.6
-cache_read = 0.05
+cache_read = 0.06
 
 [limit]
 context = 1_048_576

From f380ec9a313fa04cacf315c2d78402aea21a416e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 06:41:58 +0000
Subject: [PATCH 1979/2012] chore(sync): update NanoGPT model catalog (#7848)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 providers/nano-gpt/models/nano/lumen-stealth.toml           | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index b4cce19afb3..a62995b0beb 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576
diff --git a/providers/nano-gpt/models/nano/lumen-stealth.toml b/providers/nano-gpt/models/nano/lumen-stealth.toml
index 13c7ea58999..9b27918c763 100644
--- a/providers/nano-gpt/models/nano/lumen-stealth.toml
+++ b/providers/nano-gpt/models/nano/lumen-stealth.toml
@@ -16,7 +16,7 @@ output = 0
 [limit]
 context = 200_000
 input = 200_000
-output = 16_000
+output = 100_000
 
 [modalities]
 input = ["text", "image"]

From ceec55e82ddeb5028bf335903b8d85b8833b180c Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 06:42:04 +0000
Subject: [PATCH 1980/2012] chore(sync): update OpenRouter model catalog
 (#7849)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml    | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml     | 1 -
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 4 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 7c5828b16e2..b8d53b0809c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.66
-output = 1.98
-cache_read = 0.022
+input = 1.32
+output = 3.96
+cache_read = 0.044
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index e0411f29de8..0e2589390ee 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.079
-output = 0.6
-cache_read = 0.06
+input = 0.04
+output = 0.64
+cache_read = 0.016
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
index abacdd68c88..74b5d1276f6 100644
--- a/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/openrouter/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -18,4 +18,3 @@ cache_read = 0.25
 
 [limit]
 context = 1_048_576
-output = 262_144
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index aeddf9446b1..893b093ef16 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.079
-output = 0.6
-cache_read = 0.06
+input = 0.04
+output = 0.64
+cache_read = 0.016
 
 [limit]
 context = 1_048_576

From 812f7a1be29cb6da43494bae51c0ab3fcb517ad3 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 06:42:17 +0000
Subject: [PATCH 1981/2012] chore(sync): update Kilo model catalog (#7850)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml          | 1 -
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
index c79bdfce589..39ad2aa079f 100644
--- a/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
+++ b/providers/kilo/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -16,4 +16,3 @@ cache_write = 2.5
 
 [limit]
 context = 1_000_000
-output = 262_144
diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 8923039f0f4..4173b146120 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.079
-output = 0.6
-cache_read = 0.06
+input = 0.04
+output = 0.64
+cache_read = 0.016
 
 [limit]
 context = 1_048_576

From 519f233069fc0749386b97d41c3373c9b7de470d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 07:31:19 +0000
Subject: [PATCH 1982/2012] chore(sync): update OpenRouter model catalog
 (#7855)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 0e2589390ee..ffb1c4c680d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.64
-cache_read = 0.016
+input = 0.039
+output = 0.6
+cache_read = 0.038
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index 893b093ef16..c649cf1995c 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.64
-cache_read = 0.016
+input = 0.039
+output = 0.6
+cache_read = 0.038
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 0aebd435d56..4c30be833f9 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.4989
-output = 10.758
-cache_read = 0.3
+input = 1.49
+output = 14.5
+cache_read = 0.21
 
 [limit]
 context = 1_048_576

From e3468982b4d22c9bf86e739d606de38f4fbf55f4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 07:31:28 +0000
Subject: [PATCH 1983/2012] chore(sync): update Kilo model catalog (#7854)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml         | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 4173b146120..6b3302e301e 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.04
-output = 0.64
-cache_read = 0.016
+input = 0.039
+output = 0.6
+cache_read = 0.038
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 366b401a83c..163a53ae425 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.4989
-output = 10.758
-cache_read = 0.3
+input = 1.49
+output = 14.5
+cache_read = 0.21
 
 [limit]
 context = 1_048_576

From 0c7e2f19d5159d239e6b0a33086f4cf35fa11786 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 08:34:44 +0000
Subject: [PATCH 1984/2012] chore(sync): update Eden AI model catalog (#7859)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/gpt-oss-20b.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/edenai/models/flexai/gpt-oss-20b.toml b/providers/edenai/models/flexai/gpt-oss-20b.toml
index cf8291af93e..da4bdfa9352 100644
--- a/providers/edenai/models/flexai/gpt-oss-20b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-20b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.03
-output = 0.13
+input = 0.018
+output = 0.09

From 19bf42d193128dd0ef1104e8cc0036bc919eb7f0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 08:34:49 +0000
Subject: [PATCH 1985/2012] chore(sync): update Kilo model catalog (#7858)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 .../kilo/models/~deepseek/deepseek-v4-flash-latest.toml     | 6 +++---
 providers/kilo/models/~moonshotai/kimi-latest.toml          | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
index 6b3302e301e..0295282190d 100644
--- a/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.039
-output = 0.6
-cache_read = 0.038
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
index 0412502ff5d..095da1d4acb 100644
--- a/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.8
-cache_read = 0.008
+input = 0.038
+output = 0.55
+cache_read = 0.0228
 
 [limit]
 context = 1_048_576
diff --git a/providers/kilo/models/~moonshotai/kimi-latest.toml b/providers/kilo/models/~moonshotai/kimi-latest.toml
index 163a53ae425..366b401a83c 100644
--- a/providers/kilo/models/~moonshotai/kimi-latest.toml
+++ b/providers/kilo/models/~moonshotai/kimi-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 1.49
-output = 14.5
-cache_read = 0.21
+input = 1.4989
+output = 10.758
+cache_read = 0.3
 
 [limit]
 context = 1_048_576

From 8200e1330cf4e3a30b38acffd9e4af74d3d3b046 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 08:35:14 +0000
Subject: [PATCH 1986/2012] chore(sync): update OpenRouter model catalog
 (#7857)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml     | 6 +++---
 .../openrouter/models/~deepseek/deepseek-flash-latest.toml  | 6 +++---
 .../models/~deepseek/deepseek-v4-flash-latest.toml          | 6 +++---
 providers/openrouter/models/~moonshotai/kimi-latest.toml    | 6 +++---
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index ffb1c4c680d..ee75b6ad088 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.039
-output = 0.6
-cache_read = 0.038
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
index c649cf1995c..a3bff918037 100644
--- a/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.039
-output = 0.6
-cache_read = 0.038
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
index 6c267e8f600..f4f9f511585 100644
--- a/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.03
-output = 0.8
-cache_read = 0.008
+input = 0.038
+output = 0.55
+cache_read = 0.0228
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~moonshotai/kimi-latest.toml b/providers/openrouter/models/~moonshotai/kimi-latest.toml
index 4c30be833f9..0aebd435d56 100644
--- a/providers/openrouter/models/~moonshotai/kimi-latest.toml
+++ b/providers/openrouter/models/~moonshotai/kimi-latest.toml
@@ -20,9 +20,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.49
-output = 14.5
-cache_read = 0.21
+input = 1.4989
+output = 10.758
+cache_read = 0.3
 
 [limit]
 context = 1_048_576

From f0c6380414cd56cb4be5bed1f869882815cbf7e0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 09:30:03 +0000
Subject: [PATCH 1987/2012] chore(sync): update OpenRouter model catalog
 (#7861)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/nvidia/nemotron-3.5-lightning.toml      | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
index 842e5779095..3b251679339 100644
--- a/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/openrouter/models/nvidia/nemotron-3.5-lightning.toml
@@ -7,9 +7,9 @@ description = "Nemotron model for efficient reasoning, coding, and specialized A
 type = "toggle"
 
 [cost]
-input = 0.07
+input = 0.08
 output = 0.2
 cache_read = 0.04
 
 [limit]
-output = 235_929
+output = 131_072

From 1eb046d7bb7ad975f07872090052b4dbd65c4668 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 09:30:22 +0000
Subject: [PATCH 1988/2012] chore(sync): update Kilo model catalog (#7860)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/nvidia/nemotron-3.5-lightning.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
index 5244dc1ac02..78104d7319f 100644
--- a/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
+++ b/providers/kilo/models/nvidia/nemotron-3.5-lightning.toml
@@ -10,4 +10,4 @@ input = 0.065
 output = 0.18
 
 [limit]
-output = 235_929
+output = 131_072

From 7b48252dccd08028f2a3a9fb7b767356df978d46 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 10:27:54 +0000
Subject: [PATCH 1989/2012] chore(sync): update OpenRouter model catalog
 (#7866)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml   | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4.1-flash.toml    | 7 +++----
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index b8d53b0809c..4aa5d8a215d 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
+input = 0.498696
+output = 1.496088
+cache_read = 0.016623
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index ee75b6ad088..16f058ed841 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,10 +11,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.1
-output = 0.5
-cache_read = 0.01
+input = 0.15
+output = 0.6
+cache_read = 0.003
 
 [limit]
 context = 1_048_576
-output = 943_718

From e50d3893c7e3fa5084ba37dcbaad7cf24c7cb69a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 10:28:10 +0000
Subject: [PATCH 1990/2012] chore(sync): update Kilo model catalog (#7865)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml | 2 +-
 providers/kilo/models/deepseek/deepseek-v4.1-flash.toml  | 1 -
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
index 007b654034e..01800e18158 100644
--- a/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4-pro-0813.toml
@@ -11,4 +11,4 @@ output = 3.96
 cache_read = 0.044
 
 [limit]
-context = 1_048_576
+context = 1_024_000
diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index d755c215bcf..2d43303aad0 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,4 +12,3 @@ cache_read = 0.006
 
 [limit]
 context = 1_048_576
-output = 943_718

From 31a4a598f94ab1818e8eb5551b560c078593f68e Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 10:28:40 +0000
Subject: [PATCH 1991/2012] chore(sync): update NanoGPT model catalog (#7867)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-flash-vision-exp.toml       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
index a62995b0beb..b4cce19afb3 100644
--- a/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
+++ b/providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.44
-output = 1.32
-cache_read = 0.014
+input = 0.22
+output = 0.66
+cache_read = 0.007
 
 [limit]
 context = 1_048_576

From fb4e56d8fda4de2de32ed9bddfa561a3e7cb7771 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 11:25:59 +0000
Subject: [PATCH 1992/2012] chore(sync): update OpenRouter model catalog
 (#7870)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/upstage/solar-mini4.toml           | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/openrouter/models/upstage/solar-mini4.toml

diff --git a/providers/openrouter/models/upstage/solar-mini4.toml b/providers/openrouter/models/upstage/solar-mini4.toml
new file mode 100644
index 00000000000..e64d7840ac6
--- /dev/null
+++ b/providers/openrouter/models/upstage/solar-mini4.toml
@@ -0,0 +1,28 @@
+name = "Solar Mini 4"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+family = "solar"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.05
+output = 0.2
+cache_read = 0.005
+
+[limit]
+context = 524_288
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 137601bdeb8acf653d7b169774cd7f34d1802539 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 11:26:10 +0000
Subject: [PATCH 1993/2012] chore(sync): update Kilo model catalog (#7869)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/upstage/solar-mini4.toml      | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 providers/kilo/models/upstage/solar-mini4.toml

diff --git a/providers/kilo/models/upstage/solar-mini4.toml b/providers/kilo/models/upstage/solar-mini4.toml
new file mode 100644
index 00000000000..94538bf4b0a
--- /dev/null
+++ b/providers/kilo/models/upstage/solar-mini4.toml
@@ -0,0 +1,28 @@
+name = "Upstage: Solar Mini 4"
+description = "Solar Mini 4 is Upstage's compact, cost-efficient language model, a 35B-parameter mixture-of-experts with 3B active parameters and a 524K context window. It is built for agentic use cases where response..."
+family = "solar"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.1
+output = 0.4
+cache_read = 0.01
+
+[limit]
+context = 524_288
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 3beb7f45e258202502b57bff7d08e51737e61da4 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 13:27:45 +0000
Subject: [PATCH 1994/2012] chore(sync): update Kilo model catalog (#7880)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/kilo-auto/free.toml                | 2 +-
 providers/kilo/models/~deepseek/deepseek-pro-latest.toml | 8 ++++----
 providers/kilo/models/~z-ai/glm-latest.toml              | 6 +++---
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/kilo/models/kilo-auto/free.toml b/providers/kilo/models/kilo-auto/free.toml
index 826dc9f64ed..9b52762995f 100644
--- a/providers/kilo/models/kilo-auto/free.toml
+++ b/providers/kilo/models/kilo-auto/free.toml
@@ -20,7 +20,7 @@ cache_write = 0
 
 [limit]
 context = 256_000
-output = 10_000
+output = 32_768
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
index 7666171face..6f1d7199aea 100644
--- a/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/kilo/models/~deepseek/deepseek-pro-latest.toml
@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.4
-output = 4.3
-cache_read = 0.033
+input = 0.39996
+output = 1.19988
+cache_read = 0.012726
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/kilo/models/~z-ai/glm-latest.toml b/providers/kilo/models/~z-ai/glm-latest.toml
index 2ce00d3e342..d891278d7ef 100644
--- a/providers/kilo/models/~z-ai/glm-latest.toml
+++ b/providers/kilo/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5625
-output = 2.5
-cache_read = 0.125
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_048_576

From c2cb3793df951a771ed043f803c214bc66f55c93 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 13:27:49 +0000
Subject: [PATCH 1995/2012] chore(sync): update OpenRouter model catalog
 (#7881)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../openrouter/models/deepseek/deepseek-v4-flash.toml     | 6 +++---
 .../openrouter/models/deepseek/deepseek-v4-pro-0813.toml  | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml             | 6 +++---
 .../openrouter/models/~deepseek/deepseek-pro-latest.toml  | 8 ++++----
 providers/openrouter/models/~z-ai/glm-latest.toml         | 6 +++---
 5 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 6f817c5a0a0..5f20b283e5c 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.088606
-output = 0.177212
-cache_read = 0.017721
+input = 0.08708
+output = 0.17416
+cache_read = 0.017416
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
index 4aa5d8a215d..7d6b9d6f45e 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.498696
-output = 1.496088
-cache_read = 0.016623
+input = 0.462
+output = 1.386
+cache_read = 0.0154
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 8b9af97ef26..5422c468088 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.84
-output = 2.64
-cache_read = 0.156
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_310_720
diff --git a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
index 3112dca3caa..5a32de51cb5 100644
--- a/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
+++ b/providers/openrouter/models/~deepseek/deepseek-pro-latest.toml
@@ -20,13 +20,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.4
-output = 4.3
-cache_read = 0.033
+input = 0.39996
+output = 1.19988
+cache_read = 0.012726
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 393_216
 
 [modalities]
 input = ["text"]
diff --git a/providers/openrouter/models/~z-ai/glm-latest.toml b/providers/openrouter/models/~z-ai/glm-latest.toml
index 9bbf416c3f6..42058ee6fca 100644
--- a/providers/openrouter/models/~z-ai/glm-latest.toml
+++ b/providers/openrouter/models/~z-ai/glm-latest.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5625
-output = 2.5
-cache_read = 0.125
+input = 0.5614
+output = 1.7644
+cache_read = 0.10426
 
 [limit]
 context = 1_310_720

From 6280e4b87388cca7d048d672242404a324a5ee56 Mon Sep 17 00:00:00 2001
From: Jack <jiangyoubob@gmail.com>
Date: Wed, 23 Sep 2026 22:18:09 +0800
Subject: [PATCH 1996/2012] feat(opencode): add Space Bunny Free

---
 .../opencode-go/models/space-bunny-free.toml  | 30 +++++++++++++++++++
 .../opencode/models/space-bunny-free.toml     | 30 +++++++++++++++++++
 2 files changed, 60 insertions(+)
 create mode 100644 providers/opencode-go/models/space-bunny-free.toml
 create mode 100644 providers/opencode/models/space-bunny-free.toml

diff --git a/providers/opencode-go/models/space-bunny-free.toml b/providers/opencode-go/models/space-bunny-free.toml
new file mode 100644
index 00000000000..6fb0ca2a5eb
--- /dev/null
+++ b/providers/opencode-go/models/space-bunny-free.toml
@@ -0,0 +1,30 @@
+name = "Space Bunny Free"
+description = "Anonymous preview reasoning model for coding, agentic tasks, tool use, and multimodal input"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+
+attachment = true
+reasoning = true
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_048_576
+input = 524_288
+output = 524_288
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
\ No newline at end of file
diff --git a/providers/opencode/models/space-bunny-free.toml b/providers/opencode/models/space-bunny-free.toml
new file mode 100644
index 00000000000..6fb0ca2a5eb
--- /dev/null
+++ b/providers/opencode/models/space-bunny-free.toml
@@ -0,0 +1,30 @@
+name = "Space Bunny Free"
+description = "Anonymous preview reasoning model for coding, agentic tasks, tool use, and multimodal input"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+
+attachment = true
+reasoning = true
+reasoning_options = [
+  { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] },
+]
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_048_576
+input = 524_288
+output = 524_288
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
\ No newline at end of file

From 8cf8b71dc75c408e54baf9623676317d9d9092c0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 14:29:20 +0000
Subject: [PATCH 1997/2012] chore(sync): update Kilo model catalog (#7884)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../kilo/models/aion-labs/aion-3.5-mini.toml  | 27 +++++++++++++++++++
 providers/kilo/models/aion-labs/aion-3.5.toml | 27 +++++++++++++++++++
 .../models/deepseek/deepseek-v4.1-flash.toml  |  1 +
 .../kilo/models/moonshotai/kimi-k2.5.toml     |  1 -
 .../kilo/models/moonshotai/kimi-k2.6.toml     |  4 +--
 .../models/moonshotai/kimi-k2.7-code.toml     |  6 ++---
 6 files changed, 60 insertions(+), 6 deletions(-)
 create mode 100644 providers/kilo/models/aion-labs/aion-3.5-mini.toml
 create mode 100644 providers/kilo/models/aion-labs/aion-3.5.toml

diff --git a/providers/kilo/models/aion-labs/aion-3.5-mini.toml b/providers/kilo/models/aion-labs/aion-3.5-mini.toml
new file mode 100644
index 00000000000..13d926e0dd0
--- /dev/null
+++ b/providers/kilo/models/aion-labs/aion-3.5-mini.toml
@@ -0,0 +1,27 @@
+name = "AionLabs: Aion 3.5 Mini"
+description = "Aion 3.5 Mini is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM family of models. It is the smaller, lower-cost sibling of Aion 3.5 and uses..."
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.7
+output = 1.4
+cache_read = 0.18
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/aion-labs/aion-3.5.toml b/providers/kilo/models/aion-labs/aion-3.5.toml
new file mode 100644
index 00000000000..bf909dcbb8e
--- /dev/null
+++ b/providers/kilo/models/aion-labs/aion-3.5.toml
@@ -0,0 +1,27 @@
+name = "AionLabs: Aion 3.5"
+description = "Aion 3.5 is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM family of models. It uses a collaborative generation process in which multiple specialized models each..."
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 3
+output = 6
+cache_read = 0.75
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
index 2d43303aad0..d755c215bcf 100644
--- a/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/kilo/models/deepseek/deepseek-v4.1-flash.toml
@@ -12,3 +12,4 @@ cache_read = 0.006
 
 [limit]
 context = 1_048_576
+output = 943_718
diff --git a/providers/kilo/models/moonshotai/kimi-k2.5.toml b/providers/kilo/models/moonshotai/kimi-k2.5.toml
index c11fd52e611..7631b0fcb16 100644
--- a/providers/kilo/models/moonshotai/kimi-k2.5.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2.5.toml
@@ -9,7 +9,6 @@ values = ["none", "high"]
 [cost]
 input = 0.6
 output = 3
-cache_read = 0.1
 
 [limit]
 output = 235_929
diff --git a/providers/kilo/models/moonshotai/kimi-k2.6.toml b/providers/kilo/models/moonshotai/kimi-k2.6.toml
index 6c37950cf82..de82e76dbba 100644
--- a/providers/kilo/models/moonshotai/kimi-k2.6.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2.6.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.8
-output = 3.4
+input = 0.95
+output = 4
 cache_read = 0.16
 
 [limit]
diff --git a/providers/kilo/models/moonshotai/kimi-k2.7-code.toml b/providers/kilo/models/moonshotai/kimi-k2.7-code.toml
index a3ec51be6ac..647938cad59 100644
--- a/providers/kilo/models/moonshotai/kimi-k2.7-code.toml
+++ b/providers/kilo/models/moonshotai/kimi-k2.7-code.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["high"]
 
 [cost]
-input = 0.95
-output = 4
-cache_read = 0.19
+input = 0.7062
+output = 3.3
+cache_read = 0.18
 
 [limit]
 output = 235_929

From f42d1b149437946de728e186436a7edd4449701d Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 14:29:29 +0000
Subject: [PATCH 1998/2012] chore(sync): update OpenRouter model catalog
 (#7885)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aion-labs/aion-3.5-mini.toml       | 27 +++++++++++++++++++
 .../openrouter/models/aion-labs/aion-3.5.toml | 27 +++++++++++++++++++
 .../models/deepseek/deepseek-v4-flash.toml    |  6 ++---
 .../models/deepseek/deepseek-v4.1-flash.toml  |  7 ++---
 4 files changed, 61 insertions(+), 6 deletions(-)
 create mode 100644 providers/openrouter/models/aion-labs/aion-3.5-mini.toml
 create mode 100644 providers/openrouter/models/aion-labs/aion-3.5.toml

diff --git a/providers/openrouter/models/aion-labs/aion-3.5-mini.toml b/providers/openrouter/models/aion-labs/aion-3.5-mini.toml
new file mode 100644
index 00000000000..a7b06af9a33
--- /dev/null
+++ b/providers/openrouter/models/aion-labs/aion-3.5-mini.toml
@@ -0,0 +1,27 @@
+name = "Aion 3.5 Mini"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.7
+output = 1.4
+cache_read = 0.18
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/aion-labs/aion-3.5.toml b/providers/openrouter/models/aion-labs/aion-3.5.toml
new file mode 100644
index 00000000000..1dfb66964e7
--- /dev/null
+++ b/providers/openrouter/models/aion-labs/aion-3.5.toml
@@ -0,0 +1,27 @@
+name = "Aion 3.5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 3
+output = 6
+cache_read = 0.75
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index 5f20b283e5c..a2be37136f3 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08708
-output = 0.17416
-cache_read = 0.017416
+input = 0.08554
+output = 0.17108
+cache_read = 0.017108
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
index 16f058ed841..ee75b6ad088 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4.1-flash.toml
@@ -11,9 +11,10 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.15
-output = 0.6
-cache_read = 0.003
+input = 0.1
+output = 0.5
+cache_read = 0.01
 
 [limit]
 context = 1_048_576
+output = 943_718

From 15a35f2268660aa6b1e1d39c184129c13353bd27 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 14:29:49 +0000
Subject: [PATCH 1999/2012] chore(sync): update Eden AI model catalog (#7887)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml   | 3 +--
 providers/edenai/models/flexai/Muse-Glimmer-30B.toml         | 2 +-
 providers/edenai/models/flexai/Step-3.7-Flash.toml           | 4 ++--
 providers/edenai/models/flexai/gpt-oss-120b.toml             | 3 +--
 providers/edenai/models/flexai/gpt-oss-20b.toml              | 3 +--
 providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml | 1 +
 6 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
index 1e1e8029798..4ad5f4e9eb7 100644
--- a/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
+++ b/providers/edenai/models/flexai/DeepSeek-V4-Flash-0731.toml
@@ -1,7 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash-0731"
 name = "DeepSeek V4 Flash 0731 (FlexAI)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -10,6 +8,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.065
 output = 0.18
+cache_read = 0.0097
 
 [limit]
 context = 1_048_576
diff --git a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
index 2585cc19725..667e84be7c2 100644
--- a/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
+++ b/providers/edenai/models/flexai/Muse-Glimmer-30B.toml
@@ -1,6 +1,5 @@
 base_model = "meta/muse-glimmer-30b"
 name = "Muse Glimmer 30B (FlexAI)"
-tool_call = false
 structured_output = false
 
 [[reasoning_options]]
@@ -10,3 +9,4 @@ values = ["low", "medium", "high", "xhigh"]
 [cost]
 input = 0.3
 output = 1.1
+cache_read = 0.045
diff --git a/providers/edenai/models/flexai/Step-3.7-Flash.toml b/providers/edenai/models/flexai/Step-3.7-Flash.toml
index 85009ab8b76..ea10ac487a7 100644
--- a/providers/edenai/models/flexai/Step-3.7-Flash.toml
+++ b/providers/edenai/models/flexai/Step-3.7-Flash.toml
@@ -1,8 +1,7 @@
 base_model = "stepfun/step-3.7-flash"
 base_model_omit = ["limit.input"]
 name = "Step 3.7 Flash (FlexAI)"
-tool_call = false
-structured_output = false
+structured_output = true
 
 [[reasoning_options]]
 type = "effort"
@@ -11,6 +10,7 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.2
 output = 1.15
+cache_read = 0.03
 
 [limit]
 context = 262_144
diff --git a/providers/edenai/models/flexai/gpt-oss-120b.toml b/providers/edenai/models/flexai/gpt-oss-120b.toml
index 63d2eff1fa3..ee8b1fd793f 100644
--- a/providers/edenai/models/flexai/gpt-oss-120b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-120b.toml
@@ -1,7 +1,5 @@
 base_model = "openai/gpt-oss-120b"
 name = "GPT OSS 120B (FlexAI)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -10,3 +8,4 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.037
 output = 0.17
+cache_read = 0.0055
diff --git a/providers/edenai/models/flexai/gpt-oss-20b.toml b/providers/edenai/models/flexai/gpt-oss-20b.toml
index da4bdfa9352..7a4c54cadec 100644
--- a/providers/edenai/models/flexai/gpt-oss-20b.toml
+++ b/providers/edenai/models/flexai/gpt-oss-20b.toml
@@ -1,7 +1,5 @@
 base_model = "openai/gpt-oss-20b"
 name = "GPT OSS 20B (FlexAI)"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
@@ -10,3 +8,4 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.018
 output = 0.09
+cache_read = 0.0027
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index d2f9f4d763d..44b84926f82 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -9,6 +9,7 @@ values = ["none", "low", "high", "max"]
 [cost]
 input = 0.45852
 output = 0.91704
+cache_read = 0.091704
 
 [limit]
 context = 256_000

From 4e7a144e41ba91d37eb750dcce68714cb2d68355 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 14:29:53 +0000
Subject: [PATCH 2000/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7886)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/vercel/models/openai/gpt-oss-120b.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/vercel/models/openai/gpt-oss-120b.toml b/providers/vercel/models/openai/gpt-oss-120b.toml
index e6cb3565f32..12e8f41b034 100644
--- a/providers/vercel/models/openai/gpt-oss-120b.toml
+++ b/providers/vercel/models/openai/gpt-oss-120b.toml
@@ -1,5 +1,4 @@
 base_model = "openai/gpt-oss-120b"
-knowledge = "2024-10"
 
 [[reasoning_options]]
 type = "effort"
@@ -8,6 +7,7 @@ values = ["low", "medium", "high"]
 [cost]
 input = 0.1
 output = 0.5
+cache_read = 0.1
 
 [limit]
 output = 131_072

From ad2ee378b470240df55d91f0604652fd3e9b6851 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 14:30:28 +0000
Subject: [PATCH 2001/2012] chore(sync): update NanoGPT model catalog (#7888)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aion-labs/aion-3.5-mini.toml       | 24 ++++++++++++++++
 .../nano-gpt/models/aion-labs/aion-3.5.toml   | 24 ++++++++++++++++
 .../nano-gpt/models/upstage/solar-mini4.toml  | 24 ++++++++++++++++
 .../models/upstage/solar-mini4:thinking.toml  | 28 +++++++++++++++++++
 4 files changed, 100 insertions(+)
 create mode 100644 providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
 create mode 100644 providers/nano-gpt/models/aion-labs/aion-3.5.toml
 create mode 100644 providers/nano-gpt/models/upstage/solar-mini4.toml
 create mode 100644 providers/nano-gpt/models/upstage/solar-mini4:thinking.toml

diff --git a/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml b/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
new file mode 100644
index 00000000000..1759f0b258a
--- /dev/null
+++ b/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
@@ -0,0 +1,24 @@
+name = "AionLabs: Aion 3.5 Mini"
+description = "A GLM-family collaborative generation model tuned for immersive roleplay and storytelling, with stronger narrative structure, tension, conflict, and nuanced mature themes."
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+tool_call = false
+structured_output = false
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.7
+output = 1.4
+cache_read = 0.18
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/aion-labs/aion-3.5.toml b/providers/nano-gpt/models/aion-labs/aion-3.5.toml
new file mode 100644
index 00000000000..fb349612f03
--- /dev/null
+++ b/providers/nano-gpt/models/aion-labs/aion-3.5.toml
@@ -0,0 +1,24 @@
+name = "AionLabs: Aion 3.5"
+description = "A GLM-family collaborative generation model tuned for immersive roleplay and storytelling, with stronger narrative structure, tension, conflict, and nuanced mature themes."
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+tool_call = false
+structured_output = false
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3
+output = 6
+cache_read = 0.75
+
+[limit]
+context = 262_144
+input = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/upstage/solar-mini4.toml b/providers/nano-gpt/models/upstage/solar-mini4.toml
new file mode 100644
index 00000000000..6ac86e046ec
--- /dev/null
+++ b/providers/nano-gpt/models/upstage/solar-mini4.toml
@@ -0,0 +1,24 @@
+name = "Solar Mini 4"
+description = "Upstage's compact 35B-parameter mixture-of-experts model with 3B active parameters and a 524K context window. Built for fast, cost-efficient agentic tasks, with strong Korean and English support."
+family = "solar"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.05
+output = 0.2
+cache_read = 0.005
+
+[limit]
+context = 524_288
+input = 524_288
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/nano-gpt/models/upstage/solar-mini4:thinking.toml b/providers/nano-gpt/models/upstage/solar-mini4:thinking.toml
new file mode 100644
index 00000000000..10528b0f392
--- /dev/null
+++ b/providers/nano-gpt/models/upstage/solar-mini4:thinking.toml
@@ -0,0 +1,28 @@
+name = "Solar Mini 4 Thinking"
+description = "Solar Mini 4 with reasoning enabled for agentic tasks and harder analysis across a 524K-token context window."
+family = "solar"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.05
+output = 0.2
+cache_read = 0.005
+
+[limit]
+context = 524_288
+input = 524_288
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 1edbb1d0f997d3eaae9e77249f71c936c6a82a1a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 15:28:14 +0000
Subject: [PATCH 2002/2012] chore(sync): update Eden AI model catalog (#7892)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml | 4 ++--
 .../models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml     | 4 ++--
 providers/edenai/models/ionos/openai/gpt-oss-120b.toml      | 4 ++--
 .../edenai/models/scaleway/deepseek-v4-flash-0731.toml      | 6 +++---
 providers/edenai/models/scaleway/gpt-oss-120b.toml          | 4 ++--
 .../edenai/models/scaleway/llama-3.3-70b-instruct.toml      | 4 ++--
 6 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
index 73ebbfe4b80..3dbec673b70 100644
--- a/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/edenai/models/infomaniak/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -5,8 +5,8 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.34389
-output = 0.45852
+input = 0.34233
+output = 0.45644
 
 [limit]
 context = 100_000
diff --git a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
index 92757e9fa56..0cb3a8f6a33 100644
--- a/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
+++ b/providers/edenai/models/ionos/meta-llama/Llama-3.3-70B-Instruct.toml
@@ -4,5 +4,5 @@ tool_call = false
 structured_output = false
 
 [cost]
-input = 0.745095
-output = 0.745095
+input = 0.741715
+output = 0.741715
diff --git a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
index cad05c7d393..84c064335a1 100644
--- a/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
+++ b/providers/edenai/models/ionos/openai/gpt-oss-120b.toml
@@ -8,5 +8,5 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.171945
-output = 0.745095
+input = 0.171165
+output = 0.741715
diff --git a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
index 44b84926f82..d9d452ba506 100644
--- a/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
+++ b/providers/edenai/models/scaleway/deepseek-v4-flash-0731.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.45852
-output = 0.91704
-cache_read = 0.091704
+input = 0.45644
+output = 0.91288
+cache_read = 0.091288
 
 [limit]
 context = 256_000
diff --git a/providers/edenai/models/scaleway/gpt-oss-120b.toml b/providers/edenai/models/scaleway/gpt-oss-120b.toml
index 8f735ccfed6..ffa042bc61e 100644
--- a/providers/edenai/models/scaleway/gpt-oss-120b.toml
+++ b/providers/edenai/models/scaleway/gpt-oss-120b.toml
@@ -7,8 +7,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.171945
-output = 0.68778
+input = 0.171165
+output = 0.68466
 
 [limit]
 context = 128_000
diff --git a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
index 55662b4313f..be3eaa9445a 100644
--- a/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
+++ b/providers/edenai/models/scaleway/llama-3.3-70b-instruct.toml
@@ -3,5 +3,5 @@ name = "Llama-3.3-70B-Instruct (Scaleway)"
 structured_output = false
 
 [cost]
-input = 1.03167
-output = 1.03167
+input = 1.02699
+output = 1.02699

From b62670aa873d8f3be6457e74ad44dd3dfac306a0 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 15:28:22 +0000
Subject: [PATCH 2003/2012] chore(sync): update OpenRouter model catalog
 (#7889)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/deepseek/deepseek-v4-flash.toml | 6 +++---
 providers/openrouter/models/deepseek/deepseek-v4-pro.toml   | 6 +++---
 providers/openrouter/models/z-ai/glm-5.3.toml               | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
index a2be37136f3..b4a73a16375 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-flash.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.08554
-output = 0.17108
-cache_read = 0.017108
+input = 0.08246
+output = 0.16492
+cache_read = 0.016492
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e7683af4617..e57f7b0b838 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.95526
-output = 1.91052
-cache_read = 0.079605
+input = 0.953172
+output = 1.906344
+cache_read = 0.079431
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/z-ai/glm-5.3.toml b/providers/openrouter/models/z-ai/glm-5.3.toml
index 5422c468088..8b9af97ef26 100644
--- a/providers/openrouter/models/z-ai/glm-5.3.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.5614
-output = 1.7644
-cache_read = 0.10426
+input = 0.84
+output = 2.64
+cache_read = 0.156
 
 [limit]
 context = 1_310_720

From f2036449f734cb90c595335421ff91828ec50df1 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 15:28:28 +0000
Subject: [PATCH 2004/2012] chore(sync): update DevPass (LLM Gateway) model
 catalog (#7891)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/llmgateway/models/glm-5.3-flash.toml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/llmgateway/models/glm-5.3-flash.toml b/providers/llmgateway/models/glm-5.3-flash.toml
index 5aecfa64069..d46987e8acc 100644
--- a/providers/llmgateway/models/glm-5.3-flash.toml
+++ b/providers/llmgateway/models/glm-5.3-flash.toml
@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.07
-output = 0.19
-cache_read = 0.015
+input = 0.088
+output = 0.25
+cache_read = 0.025
 
 [limit]
 context = 1_048_576

From 59f2e562921d6d8c7c398890f544aeaafc14207f Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 15:28:31 +0000
Subject: [PATCH 2005/2012] chore(sync): update LLM Gateway model catalog
 (#7890)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/gonka24/deepseek-v4-flash.toml                   | 6 +++---
 .../llmgateway-providers/models/gonka24/glm-5.3-flash.toml  | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/providers/llmgateway-providers/models/gonka24/deepseek-v4-flash.toml b/providers/llmgateway-providers/models/gonka24/deepseek-v4-flash.toml
index 6abdece9001..3497ea65828 100644
--- a/providers/llmgateway-providers/models/gonka24/deepseek-v4-flash.toml
+++ b/providers/llmgateway-providers/models/gonka24/deepseek-v4-flash.toml
@@ -10,9 +10,9 @@ type = "effort"
 values = ["none", "low", "medium", "high"]
 
 [cost]
-input = 0.051
-output = 0.104
-cache_read = 0.0097
+input = 0.065
+output = 0.116
+cache_read = 0.012
 
 [limit]
 context = 390_000
diff --git a/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml b/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
index 921d7dc252d..8c7a30a6719 100644
--- a/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
+++ b/providers/llmgateway-providers/models/gonka24/glm-5.3-flash.toml
@@ -8,9 +8,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
-input = 0.07
-output = 0.19
-cache_read = 0.015
+input = 0.15
+output = 0.3
+cache_read = 0.035
 
 [limit]
 context = 200_000

From d1f1da9c972eb7a60ed50eddb08f50e4f63a0532 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 16:30:37 +0000
Subject: [PATCH 2006/2012] chore(sync): update NanoGPT model catalog (#7896)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml | 2 +-
 providers/nano-gpt/models/aion-labs/aion-3.5.toml      | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml b/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
index 1759f0b258a..0142fb07219 100644
--- a/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
+++ b/providers/nano-gpt/models/aion-labs/aion-3.5-mini.toml
@@ -4,7 +4,7 @@ release_date = "2026-09-23"
 last_updated = "2026-09-23"
 attachment = false
 reasoning = true
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = false
 reasoning_options = []
diff --git a/providers/nano-gpt/models/aion-labs/aion-3.5.toml b/providers/nano-gpt/models/aion-labs/aion-3.5.toml
index fb349612f03..30e04d9c3c8 100644
--- a/providers/nano-gpt/models/aion-labs/aion-3.5.toml
+++ b/providers/nano-gpt/models/aion-labs/aion-3.5.toml
@@ -4,7 +4,7 @@ release_date = "2026-09-23"
 last_updated = "2026-09-23"
 attachment = false
 reasoning = true
-tool_call = false
+tool_call = true
 structured_output = false
 open_weights = false
 reasoning_options = []

From a0d3c7c7f5841b48b51df7ffaa01a1a1da83a1d5 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 16:30:52 +0000
Subject: [PATCH 2007/2012] chore(sync): update Kilo model catalog (#7897)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/google/gemini-2.5-flash-image.toml |  8 +++---
 .../google/gemini-3.1-flash-lite-preview.toml | 10 +++----
 .../inclusionai/ling-3.0-flash-vl:free.toml   | 27 -------------------
 .../kilo/models/mistralai/devstral-2512.toml  | 15 -----------
 providers/kilo/models/tencent/hy3.toml        |  6 ++---
 5 files changed, 12 insertions(+), 54 deletions(-)
 delete mode 100644 providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml
 delete mode 100644 providers/kilo/models/mistralai/devstral-2512.toml

diff --git a/providers/kilo/models/google/gemini-2.5-flash-image.toml b/providers/kilo/models/google/gemini-2.5-flash-image.toml
index 1ac49c4d12d..cec711d9189 100644
--- a/providers/kilo/models/google/gemini-2.5-flash-image.toml
+++ b/providers/kilo/models/google/gemini-2.5-flash-image.toml
@@ -4,10 +4,10 @@ reasoning = false
 structured_output = true
 
 [cost]
-input = 0.15
-output = 1.25
-cache_read = 0.015
-cache_write = 0.041667
+input = 0.3
+output = 2.5
+cache_read = 0.03
+cache_write = 0.083333
 
 [limit]
 output = 8_192
diff --git a/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml b/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
index 3cc639df076..3ed3174dbdc 100644
--- a/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
+++ b/providers/kilo/models/google/gemini-3.1-flash-lite-preview.toml
@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.125
-output = 0.75
-reasoning = 0.75
-cache_read = 0.0125
-cache_write = 0.041667
+input = 0.25
+output = 1.5
+reasoning = 1.5
+cache_read = 0.025
+cache_write = 0.083333
diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml
deleted file mode 100644
index 25f5cdff910..00000000000
--- a/providers/kilo/models/inclusionai/ling-3.0-flash-vl:free.toml
+++ /dev/null
@@ -1,27 +0,0 @@
-name = "inclusionAI: Ling 3.0 Flash VL (free)"
-description = "Ling 3.0 Flash VL builds on Ling 3.0 Flash (124B total / 5.5B active MoE from InclusionAI), further strengthening its language capabilities while adding native visual perception and advanced visual..."
-family = "ling"
-release_date = "2026-09-10"
-last_updated = "2026-09-10"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["none", "high"]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
diff --git a/providers/kilo/models/mistralai/devstral-2512.toml b/providers/kilo/models/mistralai/devstral-2512.toml
deleted file mode 100644
index 33109af1642..00000000000
--- a/providers/kilo/models/mistralai/devstral-2512.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "mistral/devstral-2512"
-description = "Devstral 2 is a state-of-the-art open-source model by Mistral AI specializing in agentic coding. It is a 123B-parameter dense transformer model supporting a 256K context window. Devstral 2 supports exploring..."
-attachment = true
-structured_output = true
-
-[cost]
-input = 0.4
-output = 2
-cache_read = 0.04
-
-[limit]
-output = 209_715
-
-[modalities]
-input = ["text", "pdf"]
diff --git a/providers/kilo/models/tencent/hy3.toml b/providers/kilo/models/tencent/hy3.toml
index c3469c1e9dc..e96e3110ab1 100644
--- a/providers/kilo/models/tencent/hy3.toml
+++ b/providers/kilo/models/tencent/hy3.toml
@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.13
-output = 0.53
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From 806fdff2535df1121f3764c4ddf3b2d9b39a9163 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 16:30:56 +0000
Subject: [PATCH 2008/2012] chore(sync): update OpenRouter model catalog
 (#7895)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/deepseek/deepseek-v4-pro.toml      |  6 ++--
 .../models/inclusionai/ling-3.0-flash-vl.toml |  2 +-
 .../inclusionai/ling-3.0-flash-vl:free.toml   | 28 -------------------
 .../models/mistralai/devstral-2512.toml       | 15 ----------
 providers/openrouter/models/tencent/hy3.toml  |  6 ++--
 5 files changed, 7 insertions(+), 50 deletions(-)
 delete mode 100644 providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml
 delete mode 100644 providers/openrouter/models/mistralai/devstral-2512.toml

diff --git a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
index e57f7b0b838..2a431cd5648 100644
--- a/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
+++ b/providers/openrouter/models/deepseek/deepseek-v4-pro.toml
@@ -13,9 +13,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.953172
-output = 1.906344
-cache_read = 0.079431
+input = 0.951432
+output = 1.902864
+cache_read = 0.079286
 
 [limit]
 context = 1_048_576
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
index 0b0d75e3590..b7c7c3951b1 100644
--- a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
+++ b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl.toml
@@ -21,7 +21,7 @@ output = 0.18
 cache_read = 0.012
 
 [limit]
-context = 131_072
+context = 262_144
 output = 32_768
 
 [modalities]
diff --git a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml b/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml
deleted file mode 100644
index b0955907412..00000000000
--- a/providers/openrouter/models/inclusionai/ling-3.0-flash-vl:free.toml
+++ /dev/null
@@ -1,28 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://openrouter.ai/docs/guides/best-practices/reasoning-tokens
-name = "Ling 3.0 Flash VL (free)"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "ling"
-release_date = "2026-09-10"
-last_updated = "2026-09-10"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 262_144
-output = 32_768
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
diff --git a/providers/openrouter/models/mistralai/devstral-2512.toml b/providers/openrouter/models/mistralai/devstral-2512.toml
deleted file mode 100644
index 91948161c41..00000000000
--- a/providers/openrouter/models/mistralai/devstral-2512.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-base_model = "mistral/devstral-2512"
-description = "Mistral coding agent model for repository tasks and software engineering workflows"
-attachment = true
-structured_output = true
-
-[cost]
-input = 0.4
-output = 2
-cache_read = 0.04
-
-[limit]
-output = 209_715
-
-[modalities]
-input = ["text", "pdf"]
diff --git a/providers/openrouter/models/tencent/hy3.toml b/providers/openrouter/models/tencent/hy3.toml
index 80ecfdd3aa8..f61fa3175e9 100644
--- a/providers/openrouter/models/tencent/hy3.toml
+++ b/providers/openrouter/models/tencent/hy3.toml
@@ -6,9 +6,9 @@ type = "effort"
 values = ["none", "low", "high"]
 
 [cost]
-input = 0.132
-output = 0.528
-cache_read = 0.033
+input = 0.0825
+output = 0.33
+cache_read = 0.020625
 
 [limit]
 context = 262_144

From f5f8f455f3cfc4a4ddaa307fd8b4938a65b7ba5a Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 16:30:59 +0000
Subject: [PATCH 2009/2012] chore(sync): update Venice model catalog (#7894)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../models/aion-labs-aion-3-0-mini.toml       |  5 +---
 .../venice/models/aion-labs-aion-3-0.toml     |  5 +---
 .../models/aion-labs-aion-3-5-mini.toml       | 23 +++++++++++++++++++
 .../venice/models/aion-labs-aion-3-5.toml     | 23 +++++++++++++++++++
 4 files changed, 48 insertions(+), 8 deletions(-)
 create mode 100644 providers/venice/models/aion-labs-aion-3-5-mini.toml
 create mode 100644 providers/venice/models/aion-labs-aion-3-5.toml

diff --git a/providers/venice/models/aion-labs-aion-3-0-mini.toml b/providers/venice/models/aion-labs-aion-3-0-mini.toml
index 5194d9671bc..bec7ef74857 100644
--- a/providers/venice/models/aion-labs-aion-3-0-mini.toml
+++ b/providers/venice/models/aion-labs-aion-3-0-mini.toml
@@ -7,10 +7,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+reasoning_options = []
 
 [cost]
 input = 0.875
diff --git a/providers/venice/models/aion-labs-aion-3-0.toml b/providers/venice/models/aion-labs-aion-3-0.toml
index 3b990a1546e..857dfc722ac 100644
--- a/providers/venice/models/aion-labs-aion-3-0.toml
+++ b/providers/venice/models/aion-labs-aion-3-0.toml
@@ -7,10 +7,7 @@ reasoning = true
 tool_call = true
 structured_output = true
 open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "medium", "high"]
+reasoning_options = []
 
 [cost]
 input = 3.75
diff --git a/providers/venice/models/aion-labs-aion-3-5-mini.toml b/providers/venice/models/aion-labs-aion-3-5-mini.toml
new file mode 100644
index 00000000000..1c7a4904c02
--- /dev/null
+++ b/providers/venice/models/aion-labs-aion-3-5-mini.toml
@@ -0,0 +1,23 @@
+name = "Aion 3.5 Mini"
+description = "Efficient model for low-latency assistance, extraction, and routine automation"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 0.875
+output = 1.75
+cache_read = 0.225
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/venice/models/aion-labs-aion-3-5.toml b/providers/venice/models/aion-labs-aion-3-5.toml
new file mode 100644
index 00000000000..1a4e0e48aa7
--- /dev/null
+++ b/providers/venice/models/aion-labs-aion-3-5.toml
@@ -0,0 +1,23 @@
+name = "Aion 3.5"
+description = "Reasoning model for deliberate analysis, multi-step problem solving, and tool use"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+open_weights = false
+reasoning_options = []
+
+[cost]
+input = 3.75
+output = 7.5
+cache_read = 0.9375
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 963b85b3fc6023259d6e983defac46cba6145bff Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 17:25:41 +0000
Subject: [PATCH 2010/2012] chore(sync): update Kilo model catalog (#7898)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml | 6 +++---
 providers/kilo/models/z-ai/glm-5.3-flashx.toml           | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml b/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
index 4ea1e77c01f..7231d1ddd68 100644
--- a/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
+++ b/providers/kilo/models/inclusionai/ling-3.0-flash-vl.toml
@@ -15,9 +15,9 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.06
-output = 0.18
-cache_read = 0.012
+input = 0.075
+output = 0.22
+cache_read = 0.015
 
 [limit]
 context = 131_072
diff --git a/providers/kilo/models/z-ai/glm-5.3-flashx.toml b/providers/kilo/models/z-ai/glm-5.3-flashx.toml
index a730a63619c..1ea88a6ac56 100644
--- a/providers/kilo/models/z-ai/glm-5.3-flashx.toml
+++ b/providers/kilo/models/z-ai/glm-5.3-flashx.toml
@@ -17,7 +17,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.37
 output = 1.25
-cache_read = 0.075
+cache_read = 0.09
 
 [limit]
 context = 1_048_576

From f2fee37102c04056cc6778de96ad920d0fc86869 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 17:25:45 +0000
Subject: [PATCH 2011/2012] chore(sync): update OpenRouter model catalog
 (#7899)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 providers/openrouter/models/z-ai/glm-5.3-flashx.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/providers/openrouter/models/z-ai/glm-5.3-flashx.toml b/providers/openrouter/models/z-ai/glm-5.3-flashx.toml
index ea5b2ee8747..4e45607c192 100644
--- a/providers/openrouter/models/z-ai/glm-5.3-flashx.toml
+++ b/providers/openrouter/models/z-ai/glm-5.3-flashx.toml
@@ -17,7 +17,7 @@ values = ["low", "high", "max"]
 [cost]
 input = 0.37
 output = 1.25
-cache_read = 0.075
+cache_read = 0.09
 
 [limit]
 context = 1_048_576

From bf47e10309b0beb39e68c76d046955feb19ff379 Mon Sep 17 00:00:00 2001
From: "opencode-agent[bot]"
 <219766164+opencode-agent[bot]@users.noreply.github.com>
Date: Wed, 23 Sep 2026 17:25:50 +0000
Subject: [PATCH 2012/2012] chore(sync): update Vercel AI Gateway model catalog
 (#7900)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
---
 .../google/gemini-3.8-flash-lite-tts.toml     | 22 +++++++++++++++++
 .../models/google/gemini-3.8-flash-tts.toml   | 22 +++++++++++++++++
 .../inclusionai/ling-3.0-flash-vl-free.toml   | 24 -------------------
 .../models/inclusionai/ling-3.0-flash-vl.toml |  9 ++++---
 .../models/recraft/recraft-v4.1-flash.toml    | 17 +++++++++++++
 5 files changed, 67 insertions(+), 27 deletions(-)
 create mode 100644 providers/vercel/models/google/gemini-3.8-flash-lite-tts.toml
 create mode 100644 providers/vercel/models/google/gemini-3.8-flash-tts.toml
 delete mode 100644 providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
 create mode 100644 providers/vercel/models/recraft/recraft-v4.1-flash.toml

diff --git a/providers/vercel/models/google/gemini-3.8-flash-lite-tts.toml b/providers/vercel/models/google/gemini-3.8-flash-lite-tts.toml
new file mode 100644
index 00000000000..00a0a5f1313
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.8-flash-lite-tts.toml
@@ -0,0 +1,22 @@
+name = "Gemini 3.8 Flash-Lite TTS"
+description = "Speech generation model for controllable voice, narration, and audio delivery"
+family = "gemini"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.5
+output = 6
+cache_read = 0.125
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["audio"]
diff --git a/providers/vercel/models/google/gemini-3.8-flash-tts.toml b/providers/vercel/models/google/gemini-3.8-flash-tts.toml
new file mode 100644
index 00000000000..7b51ef7d3af
--- /dev/null
+++ b/providers/vercel/models/google/gemini-3.8-flash-tts.toml
@@ -0,0 +1,22 @@
+name = "Gemini 3.8 Flash TTS"
+description = "Speech generation model for controllable voice, narration, and audio delivery"
+family = "gemini"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[cost]
+input = 0.5
+output = 9
+cache_read = 0.125
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["audio"]
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
deleted file mode 100644
index 63c96c847cf..00000000000
--- a/providers/vercel/models/inclusionai/ling-3.0-flash-vl-free.toml
+++ /dev/null
@@ -1,24 +0,0 @@
-# Toggle: reasoning.enabled = true|false
-# https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
-name = "Ling 3.0 Flash VL (Free)"
-description = "Multimodal reasoning model for visual analysis, planning, and tool use"
-family = "ling"
-release_date = "2026-09-08"
-last_updated = "2026-09-08"
-attachment = true
-reasoning = true
-tool_call = true
-open_weights = false
-reasoning_options = [{ type = "toggle" }]
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 256_000
-output = 32_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml b/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
index 3cf827af685..1084acba3f0 100644
--- a/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
+++ b/providers/vercel/models/inclusionai/ling-3.0-flash-vl.toml
@@ -9,11 +9,14 @@ attachment = true
 reasoning = true
 tool_call = true
 open_weights = false
-reasoning_options = [{ type = "toggle" }]
+
+[[reasoning_options]]
+type = "toggle"
 
 [cost]
-input = 0
-output = 0
+input = 0.075
+output = 0.015
+cache_read = 0.22
 
 [limit]
 context = 256_000
diff --git a/providers/vercel/models/recraft/recraft-v4.1-flash.toml b/providers/vercel/models/recraft/recraft-v4.1-flash.toml
new file mode 100644
index 00000000000..9d20cd0531c
--- /dev/null
+++ b/providers/vercel/models/recraft/recraft-v4.1-flash.toml
@@ -0,0 +1,17 @@
+name = "Recraft V4.1 Flash"
+description = "Image model for prompt-driven generation, editing, and visual design workflows"
+family = "recraft"
+release_date = "2026-09-23"
+last_updated = "2026-09-23"
+attachment = false
+reasoning = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["image"]